Plain-English definitions of the search engine marketing and B2B website marketing terms used across this site.
3
301 redirect
An HTTP status code indicating a permanent move to a new URL. Google classifies 301 and 308 as permanent, showing the target in results and treating the redirect as a strong canonicalisation signal toward it, and classifies 302, 303, and 307 as temporary. Google documents keeping redirects in place for as long as possible, generally at least a year. Two things not to assert: that Google documents no loss of PageRank through redirects, a claim traceable to a 2016 remark by a Google employee on social media rather than to documentation; and that redirecting a batch of retired pages to the homepage preserves anything. Map page to nearest equivalent page.
302 redirect
An HTTP status code indicating a temporary move. Google treats 302, along with 303 and 307, as temporary, which means it continues to show the original URL in results and does not consolidate signals onto the target the way a permanent redirect does. The problem in practice is rarely a deliberate choice. 302 is the default in many frameworks, load balancers, and CMS plugins, so permanent moves ship as temporary ones and a migration quietly underperforms for months. Check actual response codes after any migration instead of trusting the configuration. A genuinely temporary situation — a maintenance page, a short campaign URL — is what the code is for.
9
95-5 rule
A heuristic holding that at any moment only around 5% of business buyers in a category are actively in a buying process, and 95% are not — so most advertising reaches people who will buy later, if at all. It originates with Professor John Dawes of the Ehrenberg-Bass Institute, published through the LinkedIn B2B Institute in May 2021, and is derived from interpurchase intervals rather than from asking buyers. Dawes is explicit that it is not a precise rule but a heuristic, and the real split depends on how frequently the category is bought. Its useful implication is that being remembered when a need appears matters more than persuading anyone today.
A
A/B test
A controlled experiment in which visitors are randomly split between two versions and the difference in conversion rate is measured. Randomisation is what makes it evidence: a before-and-after comparison is not an A/B test, because seasonality, campaign changes, and the news all move the numbers independently. The B2B constraint is arithmetic. Detecting a realistic improvement needs hundreds of conversions per variant, which most business sites accumulate over months, and stretching a test across a quarter invites contamination from everything else that changes. Calculate the required sample before starting. If it exceeds what you can plausibly gather, split testing is simply not a tool available to you.
Accessibility
Designing and building so that people with disabilities can use a website — keyboard-only navigation, screen reader compatibility, sufficient colour contrast, captions, and form fields with properly associated labels. Two reasons it belongs in a marketing glossary. Corporate procurement increasingly asks suppliers for an accessibility conformance statement, so an inaccessible site can lose a deal directly. And the properties that help assistive technology — semantic headings, real text rather than text in images, labelled controls, meaningful alternative text — are the same properties search engines parse. Automated checkers catch only a fraction of failures; keyboard and screen reader testing catches the rest. Overlay widgets are not a remedy.
Account-based marketing (ABM)
Marketing run against a finite list of named target companies, with the account rather than the individual lead as the unit of planning and measurement. The term was introduced by ITSMA in 2003, with Bev Burgess most often credited, and the canonical tiers are strategic ABM (one-to-one), ABM Lite (one-to-few), and programmatic ABM (one-to-many). Target account lists are keyed on company domains in practice, because domains are what advertising platforms and personalisation tools can match against. ABM is not a technology purchase and does not survive sales disagreeing with the list. Vendor uplift figures for ABM are self-reported and uncontrolled; read them as marketing material, not evidence.
Ad Rank
The value Google calculates at auction time to decide whether an ad shows, in what order, and with which formats. It is computed per auction from the bid, auction-time quality of the ad and landing page, Ad Rank thresholds, the context of the search including device and location, and the expected impact of assets and extensions. Ad Rank, not the reported 1 to 10 Quality Score, is what determines position and price. The distinction matters when someone proposes a project to raise Quality Score: the reported score is a historical diagnostic averaged over past activity, while the auction is evaluated fresh every time somebody types a query.
AI Mode
A conversational search experience where Google answers with generated text and supports follow-up turns rather than returning a page of links. Google documents AI Mode alongside AI Overviews as grounded in its core Search ranking systems, using retrieval and query fan-out, and states there are no additional requirements to appear in either. On measurement, Google launched generative AI performance reporting in Search Console in June 2026, and traffic from AI features also rolls into the Performance report under the Web search type. What nobody outside Google can currently quantify is the click behaviour, so plan on the assumption that some informational demand is now answered without a visit.
AI Overviews
Google's AI-generated summary shown above or among traditional results for some queries, grounded in Google's core Search ranking systems using retrieval plus a technique Google calls query fan-out. Google documents that there are no additional requirements to appear: no new files, no special markup, no Markdown, and structured data is not required for generative AI search. There is also no dedicated opt-out that preserves ordinary snippets — the documented controls are nosnippet, data-nosnippet, max-snippet, and noindex, and nosnippet removes your normal search snippet as well. Treat any published figure for traffic lost to or gained from AI Overviews with suspicion; Google publishes none, and vendor studies measure different things.
Algorithm update
A change to how a search engine ranks results. Google confirms some updates by name and ships many more without announcement, which is why attributing a traffic change to one requires care. Check whether the change is site-wide or confined to a page group, whether impressions moved or only clicks, and whether the result pages for your queries changed shape by adding advertising or AI features. There is no lever to pull for recovery from a broad core update, and no consultant holds reversal instructions. The productive response is to identify which query groups lost visibility and ask honestly whether your pages are the best available answers to those queries.
Alt text
The text alternative for an image, supplied in the alt attribute, read aloud by screen readers and shown when an image fails to load. Alt text is an accessibility requirement first — WCAG requires a text alternative for non-text content — and a modest SEO input second, mainly for image search and for context. The rule most sites get wrong is that decorative images should carry an empty alt attribute rather than a description; announcing a background flourish to a screen reader user is worse than silence. Describe what the image conveys in its context rather than cataloguing what is depicted, and do not use the attribute to repeat keywords.
Attribution
Assigning credit for a conversion or a deal to the marketing that preceded it. Attribution is a modelling exercise rather than a measurement: it distributes credit according to rules somebody chose, and different rules produce different answers from identical data. In B2B the difficulties compound — several people from one account touch several channels across several months, much influential activity is never clickable, and consent gaps and cookie lifetimes break the trail. The defensible use is directional: compare channels under one consistent rule and watch for large changes. Treating a model's channel-level output as a measurement of causation is the most expensive habit in B2B reporting.
Average contract value (ACV)
The annualised value of a customer contract, used to normalise deals with different terms so they can be compared and forecast. A three-year agreement has an ACV of one third of its total contract value, and quoting the wrong one — in either direction — is a common source of inflated reporting and broken forecasts. ACV should govern marketing investment decisions because it sets how much acquisition effort a deal can carry. It also determines which channels are viable at all: a programme that depends on individual human follow-up is only affordable above a certain contract value, and below that threshold the buying process has to be self-serve or the unit economics never close.
B
B2B marketing
Creating and capturing demand among organisations that buy from other organisations. In practice the term covers four different jobs that get conflated: building awareness in a market where most companies are not currently buying anything, capturing demand that already exists in search and on review sites, supporting a long sales process with material a salesperson can actually send, and retaining and expanding existing accounts. Most B2B marketing teams are funded only for the second job and judged only on lead volume, which is why so many produce efficient reports and flat revenue. The term is routinely used as a synonym for lead generation. Lead generation is one activity inside it, not the whole discipline.
Backlink
A link from another website to yours. Links remain a real ranking input and are also the most misrepresented commodity in the industry, because a count is easy to sell and quality is hard to verify. What matters is the source rather than the number: one link from a trade publication, standards body, university, or supplier that a human would plausibly follow is worth an unbounded quantity of directory and comment links. Two practical points for B2B. The most durable links come from things worth citing — original data, tools, technical documentation, specifications — and paid link schemes create a liability that outlives whichever employee agreed to them.
Bounce rate
Traditionally, the proportion of sessions consisting of a single interaction with no further activity. The definition changed materially with GA4, where bounce rate is reported as the inverse of engagement rate rather than as the old single-page-session measure, so figures from before and after a migration are not comparable and should not be charted together. Bounce rate is also a poor quality signal on informational pages: a visitor who read a definition, got the answer, and left has been served well. Treat it as a diagnostic on pages where a next step was the entire point, such as a landing page behind an advertisement, and ignore it on reference content.
Business-to-business (B2B)
Selling products or services to other organisations rather than to individuals buying for themselves. The definition turns on the buyer, not the product: the same laptop is a consumer sale to a household and a B2B sale to a procurement department. What follows from that distinction is what matters in practice. Purchases are made by groups rather than individuals, the money belongs to the organisation, the buyer is accountable to colleagues for the choice, and the decision usually has to survive review by IT, legal, finance, and procurement. B2B is often used loosely as a synonym for enterprise software marketing. It also covers manufacturers, distributors, contractors, and professional services firms, whose buyers behave differently again.
Business-to-consumer (B2C)
Selling to individuals buying for themselves or their household. The mechanics that make consumer marketing work — impulse, short consideration, a single decision maker, large addressable audiences — are precisely the mechanics B2B lacks, which is why consumer playbooks imported wholesale into B2B disappoint. An audience of four thousand realistic buyers cannot be optimised the way an audience of four million can, and a campaign judged on same-month conversions will always look like a failure against a nine-month sales cycle. The distinction is about the buying context, not the channel. Plenty of B2B companies advertise successfully on consumer platforms; what has to change is the measurement window and the definition of a result.
Buying committee
The group inside a prospective customer who must agree before a purchase happens — typically an economic buyer, the people who will use the thing, and functional gatekeepers in IT, security, finance, legal, and procurement. Gartner's May 2025 research, based on a survey of 632 B2B buyers, described buying teams ranging from five to 16 people across as many as four functions, and found roughly 74% of buying groups showing unhealthy conflict. Also called the buying centre. The practical consequence is that content aimed at one persona can actively hurt: the same research found individual-level relevance reduced consensus by 59%, while group-level relevance improved it by 20%.
C
Call to action (CTA)
The instruction telling a visitor what to do next, and the element they act on. CTA quality is mostly a function of what is offered rather than how it is styled: a specific, low-commitment, plainly described next step will outperform an urgent verb on a brighter button. The common B2B error is offering only the highest-commitment action available — contact sales, book a demonstration — to visitors still working out whether they have a problem, then concluding the traffic was poor quality. Give a page one primary action, word it so it describes what happens after the click, and never let a form be the only option.
Canonical URL
The URL a search engine treats as the definitive version of a page when several URLs serve the same or very similar content. You express a preference with a rel=canonical element, but Google documents plainly that indicating a canonical preference is a hint, not a rule, and that it may choose a different page as canonical than you do. The documented signals it weighs are HTTP versus HTTPS, redirects, sitemap inclusion, and the rel=canonical itself — redirects and rel=canonical being strong signals and sitemap inclusion a weak one. Conflicting signals are the usual cause of unexpected canonicalisation: a canonical pointing one way while internal links and the sitemap point another.
Category entry points
The situations, needs, and cues that cause a buyer to start looking for a category — a contract renewal, a failed audit, a new regulation, a system that broke, a person who left. The framework is Professor Jenni Romaniuk's, developed at the Ehrenberg-Bass Institute and published with the LinkedIn B2B Institute: identify buying situations using the W questions, then prioritise them. For B2B marketing the application is direct and unglamorous. Entry points are what to build pages about and what to name in advertising, because they are what a buyer recognises as their own circumstances — which product features, however genuinely good, are not.
CCPA
The California Consumer Privacy Act, as amended by the CPRA — the most consequential US state privacy law for B2B marketers, for a specific reason that is routinely missed. The CCPA's exemptions for B2B and HR data expired on 1 January 2023 and were not renewed, so business contact information about California residents now sits fully within scope. The practical obligations include notice at collection, honouring opt-outs of sale and sharing including browser preference signals, and holding the required contract terms with advertising vendors. Californian enforcement through 2025 and 2026 has concentrated on non-functional opt-outs, ignored preference signals, verification friction, and missing vendor contract terms.
Churn rate
The proportion of customers, or of recurring revenue, lost in a period. Customer churn and revenue churn are different numbers and are constantly confused: losing many small accounts and losing one large one produce very different revenue outcomes from the same customer-count figure. Net revenue retention, which offsets churn against expansion inside remaining accounts, is the measure most B2B boards now look at, and it can exceed 100%. For marketing the relevance is that churn sets the value of everything acquisition produces. Where churn concentrates in a segment your campaigns target well, the efficient response is to stop targeting that segment, not to acquire more of it.
Close variant
Google's own paraphrases and synonyms of your keyword, which it will match your ads against. Close variants apply to every match type, exact match included, and Google documents that there is no way to opt out. For exact match, close variants include synonyms, paraphrases, and queries Google judges to have the same search intent. This is the most expensive single fact in B2B paid search, because it means an exact-match keyword is not literal, and a technical term can be matched to a consumer query that happens to share vocabulary. The only real controls are the search terms report and a maintained negative keyword list, worked as routine rather than at audit time.
Consent mode
Google's mechanism for adjusting how its tags behave according to the consent a visitor has given. It carries seven parameters; the four that govern marketing measurement are ad_storage, ad_user_data, ad_personalization, and analytics_storage, the last two added in version 2. Basic mode blocks tags until the visitor interacts with the banner. Advanced mode loads tags with consent defaulted to denied and sends cookieless pings, which is what enables Google's advertiser-specific modelling. One caution for B2B sites: GA4 behavioural modelling has documented daily traffic minimums for both denied and granted consent that most business sites do not meet, so advanced mode does not necessarily recover the missing data.
Conversion
Any action you have decided to count as a result — a form submitted, a document downloaded, a call started, a demonstration booked. A conversion is a definition you choose rather than a thing that exists in the world, which is why conversion counts mean nothing without the list of what is being counted. Two rules keep the number honest in B2B. Report by type instead of summing newsletter signups with demonstration requests; and record which conversions later became qualified opportunities, so the difference between activity and revenue stays visible. Any report where the conversion definition changed mid-period without annotation should be treated as unusable rather than argued about.
Conversion rate
Conversions divided by a denominator — sessions, users, or clicks — expressed as a percentage. The denominator is where most misreporting starts, because platforms and analytics tools use different ones and rarely say which. In B2B the bigger problem is small numbers: a page with 300 visits a month and six enquiries has a 2% conversion rate with a confidence interval wide enough to cover almost any decision you might take. Do not compare conversion rates across channels either. Branded search converts because the visitor had already decided; a demand-creation campaign never will, and cutting it on that comparison is how programmes lose the activity that fills the funnel.
Conversion rate optimization (CRO)
Systematically increasing the proportion of visitors who take a desired action, through research, hypothesis, controlled testing, and measurement. In B2B, CRO usually cannot be practised the way it is taught, because split testing needs traffic and conversion volumes most business sites do not have. What works instead is diagnostic: session recordings and form analytics to find where people abandon, removing fields nobody uses, making the offer specific, and rewriting pages that answer the wrong question. Where volume genuinely will not support a test, say so and change things on evidence and judgement, rather than pretending a two-week experiment with forty conversions established anything.
Cookie banner
The interface asking a visitor to accept or refuse non-essential cookies and similar technologies. In the EU and UK the requirement comes from ePrivacy Article 5(3) as implemented nationally, not from the GDPR, and it covers reading or writing information on a device — which the EDPB's Guidelines 2/2023, adopted 16 October 2024, extended explicitly to URL and pixel tracking, local processing, and unique identifiers, confirming that the information involved need not be personal data. Two practical points. Regulators expect choices to be presented equally with clear information. And a banner whose tags fire before any choice is made provides no protection whatsoever, which is the most common implementation fault.
Core Web Vitals
Google's three field-measured metrics of page experience: Largest Contentful Paint, Interaction to Next Paint, and Cumulative Layout Shift. The documented thresholds are LCP within 2.5 seconds, INP of 200 milliseconds or less, and CLS of 0.1 or less, assessed at the 75th percentile of real visits and segmented separately for mobile and desktop. First Input Delay is not one of them: INP replaced FID as a stable Core Web Vital on 12 March 2024, and FID data support ended on 9 September 2024 across Chrome tools, CrUX, and the PageSpeed Insights API. Time to First Byte, First Contentful Paint, and Total Blocking Time are supporting metrics only.
Cost per acquisition (CPA)
The average cost of one recorded conversion, calculated as spend divided by conversions. CPA is only as meaningful as the conversion behind it, and in B2B that conversion is usually a form fill several steps and several months away from revenue. A campaign with an excellent CPA on newsletter signups can be worth nothing. Two specific traps. Platform CPA is computed against conversions the platform claims, using its own attribution window and modelling, so it will not agree with your analytics; and improving CPA by counting easier conversions is the most common way a paid search report is made to look better while nothing has actually improved.
Cost per click (CPC)
The average amount paid per click, calculated as cost divided by clicks. CPC is a cost report, not a performance measure, and treating it as one causes real damage: the cheapest clicks in a B2B account are usually the least commercially relevant, so pushing CPC down often means buying worse traffic. What CPC is genuinely good for is capacity planning. At a given CPC and conversion rate, a given budget can only produce so many enquiries, and if that number falls short of what the business needs, the plan fails on arithmetic before anyone executes it. Compare CPC only within the same keyword set and geography.
Crawl budget
The amount of crawling a search engine is willing and able to do on a site, set by what the server can bear and how much the engine considers the site worth fetching. Crawl budget is the most over-diagnosed problem in technical SEO. For a corporate site of a few hundred or a few thousand URLs it is effectively never the reason pages are missing from the index; the reason is duplication, thin content, or an indexing directive nobody has read. It becomes a genuine constraint on sites with hundreds of thousands of URLs, especially where faceted navigation or session parameters generate near-infinite combinations. Diagnose it from server logs, not from a tool's score.
Crawling
The process by which a search engine's automated client requests URLs, following links and sitemaps to discover pages. Crawling is not indexing and not ranking, and the distinction matters when diagnosing a problem: a page can be crawled and not indexed, indexed and not ranked, or blocked from crawling and still listed in results because other pages link to it. Robots.txt controls crawling only. If a URL must be kept out of search results, blocking crawling is the wrong instrument, because a crawler that cannot fetch the page cannot read the directive telling it to exclude the page. Server log files are the only complete record of what was actually crawled.
Cumulative Layout Shift (CLS)
The Core Web Vital for visual stability, measuring how much visible content moves unexpectedly during the life of a page. Google's documented threshold is 0.1 or less at the 75th percentile of real visits. The usual causes are images and advertisements without reserved dimensions, web fonts swapping in after first paint, and content injected late by consent banners, chat widgets, or personalisation scripts — which is why marketing tags are so often why a corporate site fails this metric. CLS is also the vital most likely to be missed in testing, because it shows up when a real visitor scrolls into content a laboratory run never reached.
Customer acquisition cost (CAC)
Total sales and marketing cost for a period divided by the number of new customers acquired in that period. The arithmetic is trivial; the boundaries are not, which is why the number is so often meaningless. Whether you include salaries, whether the events team counts, and whether the customers in the numerator's period are the ones that spending actually produced all change the answer materially. In a business with a nine-month sales cycle, this quarter's cost and this quarter's new customers are largely unrelated populations. Compute it on cohorts where the data allows, blend it across all spend rather than by channel, and treat channel-level CAC as an attribution artefact.
Customer lifetime value (LTV)
The gross profit expected from a customer over the life of the relationship, usually calculated as average revenue per period, times gross margin, divided by churn rate. Every input is an estimate, and dividing by a churn rate measured over one year for a product bought on three-year terms produces a figure that looks precise and is not. LTV is most useful as a ceiling on acquisition cost and as a way to compare segments, rather than as a number to publish. Its main practical value in B2B is showing that a segment which converts poorly may still deserve investment because it renews and expands, and that a segment which converts easily may not.
Customer relationship management (CRM)
Software recording companies, contacts, and deals, together with the activity against them — the system of record for pipeline and revenue in most B2B organisations. For marketing measurement its role is structural: the CRM is where money is recorded, so any claim about marketing's contribution must be reconcilable to it. Two hard truths follow. CRM data quality is a marketing problem rather than an IT problem, because duplicate accounts and empty source fields destroy reporting that no analytics configuration can rescue. And CRM fields are filled in by people carrying quotas, so stage dates, deal values, and lead source reflect incentives as much as events.
D
Demand capture
Marketing that converts demand which already exists into enquiries — non-branded search advertising, category and comparison content, review site presence, and the pages that answer someone already shopping. Demand capture usually shows the best measured return in a B2B programme, which is exactly why it gets over-funded: it is efficient because the demand was already there, and its efficiency is not evidence that it created anything. Capture is also finite. Once you hold the available search demand in a category of a few thousand companies, more budget buys progressively worse traffic, and further growth requires creating demand rather than harvesting it — work that will report worse and matter more.
Demand generation
Marketing that creates awareness of a problem and of your category among organisations not currently looking to buy. It is measured over quarters, and its effect appears as growth in branded search, direct traffic, inbound enquiries, and win rates — not as form fills attributable to a campaign. The term is now widely used as a job title covering everything paid, which has drained it of meaning. The distinction worth defending is between creating demand and capturing demand that already exists. Both are necessary. Conflating them lets a programme that is harvesting existing demand report itself as generating new demand, which is how companies discover their growth ceiling by accident.
Derived demand
Demand for a business input that exists only because there is demand for something further down the chain. A supplier of industrial fasteners sells because vehicles and appliances are being built; no amount of fastener marketing creates demand for fasteners in the absence of that. Derived demand explains why B2B volumes swing harder than the consumer demand behind them, and why forecasting from your own funnel alone misleads. Two practical consequences: conditions in your customers' markets belong in your own reporting, and in a downturn the honest option is usually to defend share of a shrinking market rather than to spend into demand that is not there to be captured.
E
Engagement rate
In GA4, the proportion of sessions that were engaged — lasting longer than a set threshold, containing a key event, or including at least two page views. It replaces bounce rate as the primary attention measure, and bounce rate is now derived as its inverse. The threshold is configurable, which means the metric is only comparable within one property whose settings have not changed. Its value in B2B is modest but real: it separates traffic that did something from traffic that arrived and left, which is the quickest way to spot a paid campaign buying clicks from people who never intended to visit a supplier's website.
F
Featured snippet
A block at the top of some result pages that answers a query directly, quoted from a page which is credited and linked. You cannot mark up a page to request one; the search engine selects the passage. The mechanics that appear to help are unglamorous: a concise, directly worded passage — a definition sentence, a short list, a small table — placed under a heading matching the question. Two cautions for B2B. Holding a snippet reduces the need to click, so it trades traffic for visibility, which is worth having on definitional and brand queries and rarely worth engineering elsewhere. Snippets also appear and disappear without notice, so do not build a forecast on one.
Firmographics
Attributes of an organisation used for segmentation and targeting: industry classification, employee count, revenue, location, ownership structure, number of sites. Firmographics are the B2B equivalent of demographics and carry the same warning — they describe what a company is, not whether it needs what you sell. Two practical problems recur. Vendor-supplied industry codes are frequently wrong for companies that do more than one thing, and employee counts often describe a group parent rather than the operating unit that would buy. Where a trigger condition is available — a new facility, a hiring pattern, a compliance deadline, a leadership change — it predicts buying far better than size and sector do.
First-party data
Data collected directly from your own audience through your own properties and relationships — CRM records, form submissions, email engagement, site behaviour, purchase history. First-party data is the durable asset in a market where third-party signals are unreliable and browser behaviour keeps shifting. The phrase is also routinely used to launder data that is nothing of the kind: a list bought from a broker does not become first-party because it now sits in your database. What matters is provenance and the notice given when the data was originally collected, because that notice, not the current location of the record, determines what you may lawfully do with it.
First-touch attribution
An attribution model giving all credit for a conversion to the first recorded interaction. First touch answers a narrow and genuinely useful question — what tends to introduce us to accounts that eventually buy — and is systematically biased toward demand creation and early-stage content. It is fragile in ways teams forget: the first touch it records is the first touch it could see, which excludes everything before consent was given, before a cookie existed, on another device, or offline entirely. Used alongside last touch it brackets a plausible range. Used alone as the reporting standard it will overfund awareness work and starve the activity that closes deals.
Form abandonment
Starting a form and leaving without submitting it. Abandonment is measured with field-level tracking rather than page analytics, and it is the highest-yield diagnostic on a B2B website because it identifies the exact question people refuse to answer. The offenders are predictable: telephone number, company size, budget, timeline, and free-text boxes asking for a project description. Two things to know. The recorded abandonment point is not always the offending field, since people often leave one field after the one that alarmed them. And a mandatory field sales never reads is pure loss, so audit which captured fields actually appear in a qualification conversation.
G
GDPR
The EU General Data Protection Regulation, applicable since 2018, with an equivalent UK regime after Brexit. Two misconceptions cause most of the wasted effort in B2B marketing. First, the cookie consent requirement does not come from the GDPR at all — it derives from the ePrivacy Directive Article 5(3), implemented nationally and in the UK as PECR regulation 6, and legitimate interests is not available as a basis for setting non-essential cookies. Second, business contact data is personal data, so the GDPR does apply to B2B marketing. What varies by member state is the ePrivacy rule on unsolicited email, set under Article 13, which must be checked country by country.
Generative engine optimization (GEO)
A recently coined label for trying to influence how AI-generated answers describe your company, whether in AI Overviews, AI Mode, or a standalone assistant. Treat the term cautiously: it is largely vendor language wrapped around practices that are mostly ordinary SEO and public relations, and the causal claims made for specific techniques are unproven. Google documents that no new files, markup, or formatting are required to appear in its AI features, and that structured data is not required for generative AI search. What is defensible is dull: be described accurately on the pages and third-party sources these systems retrieve from, and measure AI referrals in your analytics rather than buying a visibility score.
Global Privacy Control (GPC)
A browser-level signal a visitor can enable to tell every site they visit that they opt out of the sale and sharing of their personal information. Honouring it is not discretionary where it applies. California's regulations require a business that sells or shares personal information to treat a conforming opt-out preference signal as a valid opt-out request, and GPC is the only mechanism on Colorado's recognised universal opt-out list, which Colorado controllers have had to honour since 1 July 2024. California AB 566 further requires browsers operating in the state to offer such a signal from January 2027. Failure to honour GPC recurs repeatedly in Californian enforcement actions.
Google Analytics 4 (GA4)
Google's current analytics product and the default on most corporate websites. Standard Universal Analytics properties stopped processing hits on 1 July 2023, UA 360 received a one-time extension to 1 July 2024, and from the week of 1 July 2024 the historical data was deleted and interface and API access removed — so any pre-migration comparison now depends on whatever was exported before then. GA4 is event-based rather than session-based, uses key events in place of goals, and applies retention limits that affect explorations and funnels. It is free, it belongs to Google, and it is not a system of record for revenue. Reconcile to the CRM.
H
H1
The primary heading of a page in HTML. Its SEO significance is routinely overstated — a perfectly written H1 will not move a page that has nothing worth ranking on it — and its accessibility significance is routinely understated. Heading elements are how a screen reader user navigates, so headings should describe sections in a logical order rather than being chosen for visual size. Two errors are common: using the H1 for a logo or a company strapline on every page, so the heading says nothing about the page; and skipping levels because a smaller style looked better, which breaks the document outline that assistive technology depends on.
Hreflang
Annotations telling search engines that a page has equivalents in other languages or for other regions, so the right version can be served. Google treats the three implementation methods — HTML link elements, an HTTP Link header, and XML sitemap entries — as equivalent. Language uses ISO 639-1 codes and region uses ISO 3166-1 Alpha 2; a region code cannot be given on its own, x-default is the no-match fallback, and Google is explicit that if two pages do not both point to each other the tags are ignored. Note that EU, UN, and UK are invalid region values — the United Kingdom is GB, which is the most common error in international B2B implementations.
I
Ideal customer profile (ICP)
A written description of the type of organisation you can serve profitably and win consistently — sector, size, geography, business model, technology already in place, and the trigger conditions that make them buy. An ICP describes a company; a persona describes a person inside it. The two are not interchangeable, and most teams write the persona and skip the company. A useful ICP names its exclusions: the segments you will not pursue, and the reason. The test is operational. If the profile cannot be turned into a finite list of named accounts, or into a filter in a data source someone can actually query, it is a positioning statement rather than an ICP.
Incrementality
The additional result caused by a marketing activity over and above what would have happened anyway. Incrementality is the question attribution models pretend to answer and cannot. A branded search campaign converting beautifully may simply be intercepting people who would have arrived regardless, in which case its measured return is a transfer rather than a gain. The only honest way to establish it is experimental: withhold the activity in a matched region or from a matched set of accounts and compare outcomes over a realistic sales cycle. B2B teams resist this because it means forgoing conversions to learn something. Given how much spend rests on unverified attribution, the trade is usually worth making.
Indexing
Storing and organising a page's content so it can be returned for queries. Indexing is a decision rather than an entitlement — Google does not undertake to index every page it discovers — and on large corporate sites a substantial share of thin, duplicate, or near-identical pages will be crawled and deliberately excluded. Search Console's page indexing report gives a reason per URL, which is where diagnosis should start. The most common cause of accidental exclusion on B2B sites is not a technical block but redundancy: twelve pages saying roughly the same thing, of which the engine keeps one and reports the rest as duplicates without a user-selected canonical.
Information architecture
How content is organised, labelled, and connected so people can find it — the structure underneath the navigation, not the navigation itself. On corporate sites the architecture is usually a picture of the organisation chart: sections matching departments and product families, labelled with internal names buyers have never heard. Fixing it means grouping content by the problems buyers arrive with and naming those groups in the words buyers use. Two consequences reach beyond usability. Architecture determines internal linking, which shapes which pages search engines treat as important; and restructuring is a URL change, so it requires a redirect map or it becomes a traffic loss.
Intent data
Signals suggesting an organisation is researching a topic — content consumption across third-party publisher networks, review site activity, or behaviour on your own properties. First-party intent, meaning what accounts do on your site, is observed directly and is reliable. Third-party intent is inferred, usually at company level from IP-based attribution of anonymous readers, and no vendor publishes an independently audited accuracy figure for it. Used well, intent data changes prioritisation: which twenty accounts a small sales team calls this week. Used badly it becomes a reason to phone a company because somebody there read an article. Never hand a third-party intent surge to a salesperson as a fact about a person.
Interaction to Next Paint (INP)
The Core Web Vital for responsiveness: the time from a user interaction — a click, tap, or key press — to the next frame the browser paints, reported near the worst interaction in a visit. Google's documented threshold is 200 milliseconds or less at the 75th percentile. INP became a stable Core Web Vital on 12 March 2024, replacing First Input Delay, whose data support ended in September 2024. One point matters more than the rest for anyone reviewing a developer's report: INP cannot be measured by Lighthouse, so a perfect Lighthouse score does not mean a site passes Core Web Vitals. Lab measurement is not a substitute for field measurement.
Internal linking
The links between pages on your own site. Internal linking is the most under-used lever in corporate SEO because it costs nothing and requires nobody else's cooperation: it distributes ranking signals, indicates which page is authoritative for a topic, and gives readers a path from a general page to a specific one. Two failures are typical. Everything links to the homepage and the contact form while nothing links between related deep pages, leaving those pages isolated. And the navigation menu is treated as the whole strategy, when it is contextual links inside body copy that carry the topical signal and that readers actually follow.
K
Key event
In GA4, an event marked as important — the replacement for the Universal Analytics goal, and what GA4 counts where older reports counted conversions. Two constraints shape how you use them. A standard GA4 property allows up to 30 key events, so the list must be curated rather than accumulated; and they are counted per event, so a form firing on both submission and thank-you page view will double count. Naming matters more than teams expect, because key event names propagate into Google Ads and into audiences. Agree them once, write down what each means in business terms, and stop creating new ones opportunistically.
Keyword cannibalisation
Two or more pages on the same site competing for the same query, so search engines alternate between them and neither accumulates the signals it needs. It is endemic on corporate sites that have published for a decade: a service page, three blog posts, and a case study all addressing one topic in slightly different words. The fix is editorial rather than technical. Decide which URL should own the query, fold in what the others contained, redirect the weaker pages to it, and update internal links to match. Note that overlapping topics are not automatically cannibalisation — the diagnosis requires seeing several of your URLs actually ranking for the same queries.
Keyword difficulty
A proprietary score offered by SEO tools estimating how hard it would be to rank on the first page for a term. No search engine publishes such a metric. Each vendor calculates it differently, mostly from the link profiles of the current top results, so scores are not comparable between tools or across time. Difficulty scores also miss what usually decides B2B outcomes: whether the ranking pages actually answer a commercial query, how much of the result page is taken by advertising and AI features, and whether you have anything better to say. Use it to sort a long list quickly, never as evidence in a plan.
Keyword research
Finding the words people actually use when they search for what you sell, and estimating how much of that demand exists. It is not a list-building exercise. The output that matters is a mapping of query groups to pages, with one page owning each intent. In B2B two problems dominate. Internal product vocabulary is often not what buyers type, so research has to start from the problem rather than the product name; and volumes are small enough that a tool's estimate for a specialist term can be a rounding error either way. Search Console query data for your own site beats any third-party estimate you can buy.
L
Landing page
A page built for a single purpose, usually converting traffic from a specific campaign, with navigation and distractions reduced. The label is used two ways and the ambiguity causes arguments: in advertising it means a purpose-built page behind an ad; in analytics it means whichever page a session started on, which is frequently the homepage or an old blog post. In B2B the highest-value landing pages are often not campaign pages at all but the evergreen pages answering a category question, because those receive traffic for years. Stripping navigation helps a paid campaign page. Stripping it from an organic entry page removes the buyer's route to the rest of your evidence.
Largest Contentful Paint (LCP)
The time from when a page begins loading to when its largest visible content element — usually a hero image, a video poster frame, or a large block of text — has rendered. Google's documented threshold is 2.5 seconds or less at the 75th percentile of real visits. LCP is the Core Web Vital most often improved by removing something rather than adding technique: an oversized hero image, a font that blocks rendering, a tag manager loading marketing scripts ahead of content. Note that laboratory tools measure LCP under simulated conditions while the assessment that counts uses field data from real visitors, and the two frequently disagree.
Last-touch attribution
An attribution model giving all credit to the final recorded interaction before a conversion. It remains the most common default and it systematically flatters demand capture, especially branded search, which mostly harvests demand created elsewhere and will therefore look like your best channel indefinitely. Last touch is not useless: it is stable, easy to explain, and adequate for spotting operational faults. But an organisation that allocates budget on last touch will progressively cut everything that creates awareness, then watch its branded search volume decline with no explanation available inside the model. Report it beside first touch and treat the gap between them as honest uncertainty.
Lead
An individual who has identified themselves to you, usually by completing a form, and about whom you now hold contact details. A lead is a person, not a company, and not an intention to buy; someone who downloaded a report to cite in a university assignment is a lead in exactly the same technical sense as a procurement director shortlisting vendors. In B2B the unit that buys is an account containing several people, so lead counts systematically misrepresent progress. Three leads from one target company are one buying process, and one lead from a company that will never qualify is nothing at all. Count accounts engaged alongside leads captured.
Lead generation
Getting identified contact details from people who might buy, usually by offering something in exchange for a form. Lead generation is a tactic inside demand generation, not a synonym for it, and the two are swapped constantly. Its central failure mode is optimising for volume of forms rather than quality of buyers, because form volume is easy to measure and cheap to improve — remove fields, weaken the offer, buy cheaper traffic. In a market of a few thousand realistic accounts, a programme producing 400 leads and four qualified opportunities is worse than one producing 40 leads and twelve, though it will look better in every report that stops at the form.
Lead scoring
Assigning numeric values to lead attributes and behaviours so leads can be ranked and routed. Scoring formalises judgement rather than producing truth, and every model reflects the assumptions of whoever built it — usually invented in a workshop rather than derived from which leads actually became customers. Two practices make it worth having. Score fit and behaviour separately, because a perfect-fit company with no activity and an unqualified company reading everything are opposite problems requiring opposite responses. And validate the model against closed-won deals at least annually. An unvalidated score is a routing rule wearing statistical costume, and sales will work out which it is.
Link building
Deliberate work to earn links from other websites. The honest version overlaps almost entirely with public relations and product work: publish something worth citing, take part in the trade press, contribute to industry bodies and standards, and build the tools, data, and documentation that people reference. The industry version is buying placements, sold on volume because volume is what can be promised. For most B2B companies a handful of links from genuinely relevant industry sources will outperform a hundred paid placements and will not require a clean-up exercise two years later. Judge any proposal by a single test: would a real buyer ever read that page.
Long-tail keyword
A longer, more specific query with lower individual search volume, typically describing a situation rather than a category. The term originally described the shape of the demand distribution rather than the length of a phrase, and that is the more useful reading: most of a category's total search demand sits in many rare queries rather than a few common ones. In B2B the tail is where most qualified traffic lives, because specificity usually indicates a real problem. The mistake is building a thin page for each phrase. The tail is better served by substantial pages that answer a range of related questions in one place.
M
Marketing automation
Software that executes marketing programmes from rules and data — email sequences, form handling, lead scoring, segmentation, campaign tracking, and CRM synchronisation. Marketing automation does not create demand. It removes manual work from programmes that already work, and installing it in front of a programme that does not work is the most reliable way to spend a large licence fee for nothing. Two dependencies decide whether an implementation succeeds: the CRM integration has to be reliable in both directions, and a named person has to own data hygiene. The default failure mode is an expensive platform used as an email tool because nobody had time to build the rest.
Marketing funnel
A model describing buyers moving from awareness through consideration to purchase, used to organise campaigns and reporting. It is a useful reporting abstraction and a poor description of behaviour. Gartner characterises B2B buying as non-linear, with buying groups looping back through six buying jobs — problem identification, solution exploration, requirements building, supplier selection, validation, and consensus creation — rather than descending tidily through stages. Treat the funnel as a way to group activity and measure conversion between recorded states, not as a claim about what buyers are doing. Any statistic asserting that buyers are a fixed percentage through their process before contacting sales should be discarded outright; that family of figures is untraceable to any retrievable methodology.
Marketing qualified lead (MQL)
A lead marketing considers worth passing to sales, based on who they appear to be and what they have done — job title, company fit, pages viewed, assets downloaded, forms completed. MQL is an internal agreement, not an industry standard, and its definition belongs to whoever wrote the scoring rules. That is why MQL counts are not comparable between companies and are trivially easy to inflate: loosen one threshold and the number rises without a single additional buyer existing anywhere. An MQL is only meaningful as part of a chain. The figures worth reporting are how many MQLs sales accepts, how many become opportunities, and what those opportunities are worth.
Match type
The setting governing how loosely a keyword may be matched to a query. Google Ads has three: broad, phrase, and exact. Broad match modifier was absorbed into phrase match, and since July 2021 new plus-sign keyword entries cannot be created, with existing ones behaving as phrase match. None of the three is literal, since all are eligible for close variants with no opt-out. B2B accounts consequently run tighter structures than consumer ones, because loose matching in a small market spends the budget on the wrong industry. Do not assume Microsoft Advertising behaves identically; it applies close-variant matching too, but check its current documentation rather than transferring Google's definitions.
Meta description
An HTML element giving a short summary of a page, used by search engines as one candidate for the snippet beneath the title in results. It is not a ranking factor, and Google frequently rewrites or ignores it in favour of page text that better matches the query — normal behaviour rather than a fault to fix. What the element genuinely does is affect click-through on the queries where it is used, so a specific description naming the audience, the format, and what the reader gets will outperform a keyword-stuffed sentence. Write them for the pages that matter, leave the rest to be generated, and note there is no penalty for a missing one.
Minimum detectable effect (MDE)
The smallest difference between variants an experiment could reliably detect, given its sample size and baseline conversion rate. MDE decides whether a test is worth running at all, and it is the calculation nobody performs. Small B2B samples produce large minimum detectable effects — an experiment may only be capable of detecting a 40% improvement, which no headline rewrite will deliver — so a flat result means the test was incapable, not that the change had no effect. Work out the MDE from your actual traffic before designing anything. If it comes back implausibly large, spend the effort on a bigger change or on qualitative research.
Multi-touch attribution
A family of models distributing credit for a conversion across several recorded interactions — evenly, weighted toward first and last, or by algorithm. Multi-touch is intuitively fairer than single-touch and no more causal: it still sees only what was tracked, and it converts unverifiable assumptions into decimal places that look like measurement. In B2B the obstacles are severe. Deal counts are small, cycles cross quarters, consent gaps remove touches, and much of what moves a buying committee never produces a click. Build it when volumes are large enough for the output to be stable. Below that, incrementality testing and self-reported attribution answer the same question more honestly.
N
Negative keyword
A term you exclude, preventing ads from showing for queries containing it. Negative keywords are the primary cost control in B2B paid search, because the platform controls matching and close variants cannot be switched off. The work is continuous rather than a one-off: read the search terms report weekly and exclude by pattern rather than one query at a time — job-seeking language, free and template and tutorial modifiers, competitor brands you cannot displace, and the consumer sense of an industrial term. Two mechanics people forget: negatives have their own match types, and a negative phrase set too broadly will quietly suppress the queries that were converting.
Nofollow
A link attribute telling search engines not to pass ranking credit, introduced originally to combat comment spam. Google announced in 2019 that it would treat nofollow, along with the more specific sponsored and ugc values, as hints rather than strict directives for ranking and crawling purposes. Two practical implications. Paid placements and advertorials should be marked sponsored, because failing to disclose a paid link is the actual policy problem. And a nofollowed link is still worth having when real buyers click it: referral traffic from a relevant trade publication does not care about the attribute, and in a market of a few thousand companies that readership is the whole point of the placement.
O
Organic search
Traffic from unpaid search engine results. Organic is often called free, which is only true in the sense that no media is bought — content, technical work, and the effort to earn references all cost money, and the cost is incurred months before any traffic arrives. The more damaging error is treating organic search as one thing in reporting. Branded organic traffic, from people searching your company name, mostly measures the effect of everything else you do, including your advertising and your sales team. Non-branded organic traffic measures whether people who do not know you can find you. Report the two separately or the number tells you nothing you can act on.
P
Paid search
Advertising on search engine result pages, bought at auction against keywords and matched to queries by the platform. What you are buying is presence at a moment of stated intent, which makes paid search the strongest demand-capture channel in most B2B programmes and a weak demand-creation one. The structural feature to understand is that you bid on keywords but are matched to queries, and the platform controls the matching — Google documents that all match types are eligible for close variants and that there is no way to opt out. Consequently the work that decides whether a B2B account performs is query-level: reading search terms, adding negatives, and keeping match types tight.
Pay-per-click (PPC)
An advertising model in which the advertiser pays when someone clicks, not when the ad is shown. Pay-per-click is a pricing model rather than a channel, though it is used loosely as a synonym for Google Ads and paid search generally. Paying only for clicks sounds like protection against waste. It is not: the click is the platform's revenue event, and it will cheerfully send clicks from queries that cannot possibly convert. In a B2B account the money is nearly always lost to loose matching against irrelevant intent rather than to the price per click, which is why search term reports and negative keyword lists matter more than bid adjustments.
People Also Ask (PAA)
The expanding list of related questions shown on many result pages, each opening to a short answer drawn from a page. PAA is best used as free query research rather than as a feature to chase. The questions come from real searches, so they show how buyers phrase a problem — including the phrasing your internal product vocabulary conceals. The block is dynamic, changing by query, session, and over time, so a screenshot is a snapshot rather than a target list. Answering those questions plainly on a page is defensible work whether or not the block ever cites you, because the questions are what buyers actually asked.
Pipeline
The total value of open opportunities, usually reported by stage and expected close date. Pipeline is a sales forecasting instrument that marketing has adopted as a scoreboard, and the switch introduces two problems. Pipeline is created by salespeople at their discretion, so it can grow without any new demand existing; and pipeline created this quarter may close two or three quarters later, which means marketing spend and pipeline value in the same monthly report describe different periods and should not be divided by one another. Pipeline coverage — open pipeline against the target for the period — is more informative than the raw figure, and only if stage definitions have been stable.
Progressive profiling
Collecting information across several interactions instead of demanding everything at once, so a returning visitor is asked new questions rather than ones already answered. It is the sensible resolution of the tension between short forms and sales qualification requirements, and it depends on recognising the returning person — so it works for known contacts and not for first-time anonymous visitors. Two caveats worth stating plainly. The technique only pays off if the additional fields are genuinely used by someone downstream. And accumulating data across visits must stay within the privacy notice given when the data was first collected, because this is data accumulation whatever it is called internally.
Q
Quality Score
A 1 to 10 diagnostic in Google Ads estimating the quality of a keyword's ads and landing pages relative to other advertisers, built from expected click-through rate, ad relevance, and landing page experience. A dash means there was not enough exactly matching search volume to score it, which is common in specialist B2B accounts. The widely repeated claim that raising Quality Score lowers your cost per click misstates the documentation. Google states that Quality Score is not an input in the ad auction, that it is a diagnostic tool, and that it is not a key performance indicator and should not be optimised. Fix the relevance problems it points at instead.
R
Reverse-IP lookup
Mapping a visitor's public IP address to an organisation, using regional internet registry allocation records, ASN and routing data, and reverse DNS, augmented by vendor crawling and inference. What it identifies is the organisation an IP range is assigned to — which may be an internet service provider, a VPN operator, a datacentre, or a building landlord rather than the visitor's employer. The degradations are uncontroversial and well known: residential broadband from remote workers, mobile networks and carrier-grade NAT, VPNs and proxies, serviced offices, universities and hospitals, stale registry records after ranges change hands, IPv6, and bot traffic. No vendor publishes an independently audited accuracy figure.
Rich result
A search result presented with additional features drawn from structured data: review stars, event details, product prices, breadcrumbs. Rich results are eligibility-based — valid markup makes a result eligible, Google decides whether to show it, and the set of available types changes without notice. Several types marketers still plan around no longer exist. FAQ rich results stopped appearing in Google Search on 7 May 2026, with reporting and testing support withdrawn in the following months; HowTo rich results were removed and their documentation deleted on 14 September 2023. Google's structured-data investment since 2025 has gone overwhelmingly into merchant and commerce properties, which is where most B2B sites have nothing to gain.
Robots.txt
A plain text file at the root of a host telling compliant crawlers which URL paths they may request. Robots.txt governs crawling only, and the most consequential misunderstanding in technical SEO follows directly from that. A disallowed URL can still appear in search results if other pages link to it, because the crawler is not permitted to fetch the page and therefore cannot read the noindex directive that would exclude it. To keep a page out of results, allow crawling and serve noindex, or require authentication. Robots.txt is also a public file, which makes it a poor place to enumerate the paths you would rather nobody found.
S
Sales cycle
The elapsed time from a defined starting point to a closed deal. The definition of the starting point does most of the work and is rarely stated: first touch, first meeting, opportunity creation, and proposal issued give wildly different numbers for the same business. Published average B2B sales-cycle lengths are not comparable across companies, because they come from individual CRM datasets with different stage definitions and different contract values; treat any single industry figure as decoration. Measure your own, segment it by deal size, and use the median rather than the mean, because one eighteen-month enterprise deal will distort an average built on thirty transactions.
Sales qualified lead (SQL)
A lead sales has examined and judged to be a real opportunity worth working — usually meaning the company fits, a need has been articulated, and there is some evidence of budget and timing. The definition varies more than the acronym suggests. Some organisations use sales qualified lead for anything sales agreed to accept; others reserve it for a lead that has survived a discovery call. Write your own definition down, date it, and treat changes to it as a change to your reporting baseline. The MQL-to-SQL conversion rate is the most useful single number in a B2B funnel, because it measures whether marketing and sales agree on what a good lead looks like.
Sales-accepted lead (SAL)
A lead sales has formally acknowledged receiving and agreed to work, sitting between marketing qualified and sales qualified. The stage exists to settle an argument. Without it, marketing counts leads that were passed and sales counts leads that were worked, and nobody can tell whether a poor conversion rate reflects bad leads or leads nobody ever called. Making the stage real requires a service-level agreement: a response window, a minimum number of contact attempts, and a rejection reason picklist. The rejection reasons are the valuable output, because they tell you which part of your targeting is wrong — wrong company, wrong role, wrong timing, or no need at all.
Search engine marketing (SEM)
Marketing on search engines. The term has two live uses. Historically, and in the title of this glossary, it covers everything done to gain visibility in search results, paid and organic; in Google's own vocabulary and much current usage it means paid search specifically. Because the ambiguity is genuine, state which you mean rather than assuming. The distinction that matters operationally is not the label but the buying mechanism. Organic visibility is earned, cannot be switched on for a launch date, and does not disappear when a budget is paused. Paid visibility is rented, is available immediately, and stops the day the card is declined. Most B2B programmes need both for different reasons.
Search engine optimization (SEO)
Improving how well a website is understood, indexed, and ranked by search engines, and how often searchers choose it. It covers three distinct kinds of work bundled under one label: technical work to make pages crawlable, renderable, and indexable; content work to answer queries better than the pages currently ranking; and earning references from other sites. In B2B the constraint is rarely traffic volume — non-branded demand in a specialist category may be a few hundred searches a month — so SEO here is judged on whether it reaches the right several hundred companies, not on sessions. Programmes that chase volume in a narrow market usually end up ranking for the wrong audience.
Search engine results page (SERP)
The page a search engine returns for a query. Describing it as ten blue links has been wrong for years: a commercial B2B query now routinely returns advertisements, an AI Overview, People Also Ask questions, video blocks, and local or shopping modules before the first traditional organic result. Two implications for planning. Position one no longer means top of page, so rank reports that do not record what else appeared overstate visibility, sometimes badly. And page composition varies by query, device, and location, which is why you look at the actual result pages for the twenty queries that matter to your business rather than reasoning from averages.
Search intent
What the searcher is trying to accomplish, commonly grouped as informational, navigational, commercial investigation, and transactional. Intent is not a property of the words alone — the same phrase carries different intent in different markets — and the practical way to determine it is to look at what the search engine already rewards for that query. If the first page is entirely definitions and guides, a product page will not rank there whatever is done to it. The most common B2B error is answering a definitional query with a sales page and then concluding that SEO does not work, when the mismatch was between the page's purpose and the query's.
Search volume
An estimate of how many times a term is searched in a period, usually a monthly average for a named country. Every published volume figure is modelled and rounded, differs between tools, and averages away seasonality — a term reported at 200 a month may be 600 in September and 40 in July. In B2B, low volume is not low value: a phrase searched thirty times a month by plant managers in one industry can be worth more than one searched thirty thousand times by students. Judge terms by who searches them and what the search implies about their situation, then use volume only for capacity planning.
Self-reported attribution
Asking buyers directly how they heard about you, usually as a question on a form or in the qualification call. It is imprecise, subject to recall bias, and typically the most useful attribution data a B2B company holds, because it captures what tracking cannot see: a recommendation from a peer, a conference conversation, a podcast, a colleague's forwarded email. Two implementation notes decide whether it works. Use a free text field rather than a picklist, because a picklist records what people recognise rather than what happened; and read the answers rather than only counting them, since the phrasing shows which of your messages actually landed.
Session
A group of interactions by one user inside a time window. In GA4 sessions are derived from events — a session start event carrying a session identifier — and they time out after 30 minutes of inactivity by default, so one person researching across a working day can generate several. Session counts commonly differ between analytics tools, and differed between Universal Analytics and GA4, so never compare across systems. The B2B limitation is more fundamental: a session is a poor unit for a buying process involving several people at one company over months, which is why account-level and user-level views tell you more than session counts do.
Statistical significance
The probability that a measured difference between variants is unlikely to have arisen from chance alone, conventionally reported as a p-value against a 95% threshold. Significance says nothing about whether the difference is large enough to matter commercially, and it is not permission to stop a test early. Two abuses are routine in marketing testing: checking daily and stopping the moment a tool first displays 95%, which badly inflates false positives; and reporting significance for a test that never reached its planned sample. In B2B, where conversion counts are small, a great many published test wins are noise that somebody declared a result.
Structured data
Machine-readable markup added to a page, almost always JSON-LD using schema.org vocabulary, describing what the page is about in a form a search engine can parse without interpreting prose. Its main documented purpose is eligibility for rich results, and eligibility is not entitlement. Two things worth knowing. Google states that Article and Organization types have no required properties — you add the ones that apply to your content — and that Organization markup belongs on the homepage or a single about page rather than on every page. And Google's own guidance for its AI features says structured data is not required for generative AI search and that no special markup exists for it.
T
Tag manager
A container script that lets marketing teams deploy and configure tracking and advertising tags without a code release. Tag managers solve a genuine bottleneck and create three predictable problems: tags accumulate because nobody owns removal, marketing scripts loading early damage Core Web Vitals, and the container becomes an undocumented dependency nobody can audit. It is also usually where consent enforcement lives, which makes it a compliance artefact rather than a convenience. Keep an inventory of what each tag does and what data it sends, review it at least twice a year, and apply the change control you would apply to any other production code.
Technical SEO
The work of making a site crawlable, renderable, indexable, and quick enough — server responses, status codes, canonicalisation, link structure, JavaScript rendering, structured data, and page performance. Technical SEO is necessary and not sufficient: it removes obstacles to ranking, it does not create reasons to rank. That asymmetry explains a familiar pattern, where a corporate site passes every audit tool and still receives no non-branded traffic, because nothing on it answers a question anybody searches. Invest here when the site has genuine defects — pages not indexed, a migration that lost URLs, content that only exists after a script runs — and stop when the remaining findings are cosmetic.
Technographics
Data about the technology an organisation uses — platforms, frameworks, analytics tools, and third-party services detectable on its websites and named in its job postings. Technographic data is genuinely useful when your product integrates with, replaces, or depends on a specific system, because presence of that system is a real qualification criterion rather than a proxy. Its limits follow from how it is collected. Detection is mostly from public web pages, so it sees the marketing site rather than the internal estate, records can be years stale, and absence of a signal is not evidence of absence. Use it to prioritise outreach, never as a fact to assert in a sales call.
Third-party cookie
A cookie set by a domain other than the one in the address bar, historically the mechanism behind cross-site retargeting and view-through measurement. The state of play is widely misreported, including in current marketing content. Chrome is not deprecating third-party cookies: on 22 April 2025 Google said it will not roll out a new standalone prompt and will maintain its existing approach to third-party cookie choice, reaffirming that position in October 2025. Safari has blocked cross-site cookies by default since Safari 13.1 and iOS 13.4 in March 2020, and Firefox partitions them per site under Total Cookie Protection rather than blocking them outright. Most Privacy Sandbox advertising APIs are being retired.
Title element
The HTML title element, which supplies the clickable headline in search results and the label on the browser tab. It remains one of the more influential on-page signals, and like the meta description it can be rewritten by Google — most often when the title is boilerplate, over-long, or stuffed with repeated terms. Practical points for corporate sites: put the distinguishing words first and the company name last, keep the important part short enough to survive truncation, and make every title unique. Templated titles differing only by a category name are the most common cause of a site's own pages competing with each other for the same query.
Total addressable market (TAM)
The total annual revenue available if every organisation that could buy your category bought it from someone. TAM is a sizing exercise for investors and for deciding whether a market is worth entering. It is not a marketing target, and using it as one produces plans with no arithmetic behind them. The number that governs day-to-day decisions is much smaller: the count of organisations that match your ideal customer profile, can be reached, and could plausibly buy inside your planning horizon. In many B2B categories that list runs to a few thousand companies, sometimes a few hundred. A bottom-up count built from real company lists is far more defensible than a percentage of an analyst's market figure.
U
UTM parameter
Tags appended to a URL — utm_source, utm_medium, utm_campaign and others — which analytics tools read to classify incoming traffic. UTMs are the only channel data an organisation fully controls, and they degrade for one predictable reason: nobody enforces a naming convention, so email and Email and e-mail become three channels and a year of reporting becomes unrecoverable. Publish a convention, use lower case, and build links from one shared template. Two things never to do: do not put UTMs on internal links, because that restarts the session and destroys the original source attribution; and do not put anything identifying a person into a parameter, because those values end up in analytics and in referrer logs.
V
Visitor identification
Software that attempts to tell you which companies, or in some products which individuals, visited your website. Two quite different products share the label. Account-level identification uses reverse-IP mapping and infers an organisation. Person-level de-anonymisation depends on a third-party identity graph already holding a persistent identifier for that browser, performs better on desktop Chrome than on Safari or iOS, and is generally sold for US traffic only. Every published match rate is vendor-defined and unaudited. In the EU and UK, reading identifiers from a device engages ePrivacy Article 5(3), and obtaining personal data about someone from a third party triggers GDPR Article 14 duties, including disclosing the source.
W
WCAG
The Web Content Accessibility Guidelines, published by the World Wide Web Consortium, organised around four principles — perceivable, operable, understandable, robust — with success criteria at three conformance levels: A, AA, and AAA. AA is the level named in most procurement requirements and public sector obligations. Two practical points for a marketing team. Conformance is claimed per page and per page state, so a site is not conformant merely because one template was fixed. And a conformance claim in a tender is a representation about the site as it currently stands, which makes an untested claim a commercial risk as well as an ethical one. Test with real assistive technology.
Web analytics
The measurement of website behaviour — who arrived, from where, what they did, and what they completed. Web analytics reports events on your own property, and it cannot see the things that decide B2B outcomes: the buying committee's internal discussion, a colleague forwarding a PDF, the conversation on a stand at a trade show. Two structural limits belong beside any analytics figure you present. Consent requirements mean a proportion of EU and UK visits are never measured; and the same visit is counted differently by the analytics tool, the advertising platform, and the CRM, none of which is wrong — they answer different questions under different rules.
X
XML sitemap
A file listing URLs you want a search engine to know about, with optional last-modified dates. A sitemap helps with discovery on large sites, on sites with weak internal linking, and for newly published pages; it does not make a page rank, and Google treats sitemap inclusion as a weak canonicalisation signal. Its most valuable property is diagnostic. If the sitemap is generated from the CMS and contains only canonical, indexable, 200-status URLs, then the gap between submitted and indexed counts in Search Console becomes a real measurement. A sitemap padded with redirects, noindex pages, and 404s makes that comparison worthless and hides the problem it should reveal.