Starting a Corporate Search Programme: The First Ninety Days

The Order of the Work Decides What You Learn

Spend the first ninety days of a corporate search programme in this order: verify the measurement, establish where you already stand, learn the vocabulary your buyers use, repair what is technically broken, improve the pages that already rank, buy a small number of high-intent clicks, and only then write anything new. Each step produces the input the next one needs.

Companies rarely start there. They start by buying clicks, because clicks are the only part of search marketing that can be bought on a Tuesday afternoon. What the spend lacks is a tested conversion event, a page written for the query, and a baseline to judge the traffic against. So the account gets optimised toward the cheapest clicks rather than the best ones, because cost per click is the only trustworthy number in the room.

The cost of the wrong order is the wasted quarter. Ninety days in, the company has spend, an unusable data set and no diagnosis, and the next budget conversation starts with nothing to argue from.

Weeks One and Two: Prove the Measurement Works

If the analytics are wrong, everything downstream is guesswork dressed as evidence. Check what is actually recorded, on the real site, in the real browser, with the consent banner in the state a visitor would leave it.

• Submit the contact form yourself and watch the event arrive. A form that shows a thank-you message but records nothing is common, and it invalidates every number below it.
• Check the GA4 data retention setting. The default is two months, not fourteen, and fourteen months is the ceiling on a standard property. Aggregated reports are unaffected, but explorations and funnel reports are limited by it.
• Verify Search Console access at domain level, covering every protocol and subdomain variant, and export the history now.
• Do not expect modelling to fill the gaps. Google documents thresholds of at least a thousand daily events with analytics storage denied and a thousand daily consenting users. Most business sites never qualify, so they take the full loss from consent refusal with none of the modelled recovery a consumer site gets.

Two weeks is enough. Do not proceed until a lead can be traced from query to form.

Weeks Two to Four: Establish Where You Already Stand

You cannot claim an improvement without a documented starting point, and after launch it is too late to take one. The baseline is a dated artefact stored somewhere permanent, not a screenshot in a deck.

Capture five things. What the site currently ranks for and receives clicks on, split brand from non-brand, because brand queries flatter every report they appear in and respond to nothing you are about to do. The page inventory Google has actually indexed, which is routinely a different set from the one the content management system believes exists. The pages that already produce enquiries, however few. The entry points earning impressions but almost no clicks, which are the cheapest wins available to you. And the technical state: how the site renders, how it performs against Core Web Vitals in field data, and which templates are involved.

This phase reliably produces one surprise — a page nobody in marketing knew about doing more work than the homepage, or a minor product line accounting for most non-brand impressions. Those surprises are the argument for finishing the diagnosis before spending the budget.

Weeks Three to Six: Learn the Vocabulary, Not the Volume

Keyword research in a business market is a translation exercise, not a volume exercise. Your engineers, your buyers, your competitors and the analysts covering your category frequently use four different names for the same thing, and search engines index only the words people type.

Take the vocabulary from sources that predate the marketing department: recorded sales calls, the wording of inbound enquiries, support tickets, the language of RFPs you have answered, and internal search queries on your own site. Then check each term against paid and organic data to see what it returns. In a market of a few thousand qualified buyers, a term with ninety monthly searches and unmistakable commercial intent is worth more than one with ninety thousand.

Two checks separate a usable list from a wish list. Look at the results page for each term and ask what kind of page is winning — a product page, a comparison, a definition, a directory. Then ask whether your company can honestly be the best answer to that query. Where the answer is no, note it and move on.

Weeks Five to Eight: Repair, Then Improve What Already Ranks

Fix the faults that stop existing pages from performing before commissioning a single new one. The list is short and dull: pages indexed that should not be, pages that should be indexed and are not, duplicate URL variants competing with each other, redirect chains left by earlier projects, titles written for a keyword tool in a previous decade, and templates that push the substance of the page below three screens of hero imagery.

Then improve what is already close. Pages sitting on the second results page for a term with genuine commercial intent are the highest-return work available in a first quarter, because the page already carries whatever credibility got it that far. Rewrite for the query actually landing on it, answering it in the first paragraph. Add the specifics a technical buyer needs — the compatibility table, the certification, the tolerances — that a competitor's page has and yours does not.

This phase produces the fastest movement in the ninety days, and it reveals whether the underlying problem is content or something structural, which no volume of new content would have shown.

Weeks Eight to Thirteen: A Small Paid Programme, Then Content

Start paid search on a narrow set of terms where intent is unambiguous — the ones naming your product category, your competitors, or the problem a buyer would search with a purchase order already approved. A narrow start is the only way to attribute an outcome to a cause while the sample is small.

Three things must be true first. The conversion event is verified and imports into the ad platform. Each ad group points at a page written for that query, not the homepage. And the negative keyword list already exists, because it will do most of the work: Google documents that all match types are eligible to match close variants with no way to opt out, so exact match means same meaning and same search intent, not identical string. That is where money leaks in business markets, as a term like implementation consultant collects students, jobseekers and competitors. Manage the account against enquiries a salesperson accepts, and do not optimise the reported Quality Score — Google states it is a diagnostic, not an auction input.

New content makes sense only in the final weeks, because only then do you know what is missing: terms producing impressions with no relevant page, paid terms that convert with no organic equivalent, and questions from sales calls the site does not answer.

The False Start Nobody Questions: Redesigning First

The most expensive way to begin is to redesign the site before anyone has established what the current site does well. It happens because the site looks dated, and looking dated is the one problem every executive can see. So the project funded first is the one that destroys the evidence.

A redesign changes templates, URLs, internal linking, page speed, copy and measurement simultaneously. When traffic moves afterwards, in either direction, nothing can be attributed to anything. Worse, redesigns routinely delete the pages that were quietly earning enquiries, because those pages were not in the new information architecture and nobody had checked which ones mattered.

This is an argument about order, not about rebuilding. Do the measurement, the baseline and the page-level diagnosis first, so the redesign brief contains a list of pages that must survive with their URLs intact and their substance preserved. That list takes about four weeks to produce and is the most valuable document a redesign project can be handed. A rebuild that starts with it is a controlled change; one that starts without it is a coin toss with a project plan attached.

What to Promise at Thirty, Sixty and Ninety Days

The internal conversation matters as much as the work, because programmes are usually killed by an expectation nobody corrected in week two. Have it early, in writing, stating deliverables rather than results.

Say the uncomfortable part at the start: in a market with a long sales cycle, the first ninety days will not produce a closed deal attributable to search, and any programme claiming otherwise is counting something else.

If someone insists on revenue by day ninety, do not accept the target. Negotiate it to leading indicators you can stand behind, and put in writing what you expect by the end of the second quarter instead.

Frequently Asked Questions