Setting expectations for the first 90 days

Almost every failed engagement I have heard described was a timeline disagreement that nobody wrote down. The client expected the phone to change by month two. The agency expected to be judged at month nine. Neither said so.

Write the ninety day picture down before the work starts. It costs one meeting and it removes the argument entirely.

What actually happens in the first thirty days?

Three things, and only one of them is visible.

Access and inventory. Someone collects owner level access to your profile, analytics, Search Console, ads, and site, and lists what exists. This is unglamorous and it is where hidden problems surface, such as a domain nobody can log into or tracking that has been broken for a year.

The baseline. Current calls, forms, bookings, map impressions, and positions across a grid rather than a single number. This is the most valuable hour of the entire engagement, because everything after it becomes measurable.

Then the obvious repairs. Hours corrected, categories reviewed, missing services added, a broken form fixed, duplicate listings resolved. Category selection is the piece that pays back fastest, since your primary category does more for local ranking than almost anything else you can change. Darren Shaw's study of 1.8 million Google Business Profiles across 4,209 categories found the exact match category outranking every adjacent one, with the pattern repeating for almost every category in the data. It is a research task of an hour or two and a change that takes minutes.

If your first month feels slow and administrative, that is what it is supposed to feel like.

What should the review target for the quarter be?

What should I see by day sixty?

Production, and the first honest data.

Pages being published or rewritten on a schedule you agreed. Profile activity running. Review requests going out through whatever system your business already uses. Whatever technical items came out of the audit, being closed in a documented order.

You should also be getting your first real report against the baseline. Early numbers are noisy and should be presented as such. A report at day sixty that claims a clear result is overreading the data.

What you should feel by day sixty is the working rhythm. Whether requests get answered, whether the reports arrive on the stated date, whether the person doing the work can explain a decision plainly. That is more predictive of the next year than any metric at this stage, and it depends heavily on the contact cadence you agreed at the start.

Should anyone be A/B testing my site in the first quarter?

Almost certainly not, and the rule everyone quotes for this is worth taking apart because the real reason is better than the folklore.

You will hear that A/B testing needs ten thousand monthly visitors. That number is not a research finding. It is traceable: it appears as a citation in Kohavi, Deng, Longbotham and Xu's KDD 2014 paper on running web experiments, and the reference resolves to a blog post by Neil Patel on QuickSprout, published on 14 January 2013. The four authors, from Microsoft and LinkedIn, immediately qualify it by saying the guidance should be refined to the metrics of interest. A blog post became an industry threshold by repetition.

The real rule is that sample size depends on the variance of the metric and the size of the change you want to detect. The same paper publishes a worked table from Bing that shows how far apart the answers can be for the same product. Revenue per user, a very skewed metric, needed about 114,000 users per variant to detect a 4.4% change. The capped version of the same metric needed about 9,700 users per variant to detect a 10.5% change. Sessions per user needed about 4,700 per variant to detect a 5.4% change.

Now run it against a local business. Say your site gets 600 visits a month and converts at 3%, which is 18 leads. Split into two variants, that is 300 visits and nine leads per variant per month. To reach 4,700 visits per variant, the least demanding line in that table, would take about 16 months (derived), by which point your season, your prices and your offer have all changed.

Say the domain gap out loud, because it cuts both ways. That table is search engine data at web scale, and a local lead form is a different metric with different variance. The transferable part is the method, not the numbers: ask for the minimum detectable effect and the variance before agreeing to a test, and if nobody can supply them, you are not running an experiment, you are changing the page and hoping.

What honestly replaces it in ninety days: qualitative review, heuristic fixes, and before and after measurement over long windows, described accurately as a comparison rather than a controlled test.

What is a fair test at day ninety?

Four questions, and none of them is about a single ranking position.

Did the agreed deliverables get produced, counted against the scope. This one is factual.

Has anything moved on your easier searches. Less competitive terms, longer phrases, and positions in the parts of your service area where you were already close. Movement usually appears at the edges first.

Can you now see what happens. Calls tracked, forms attributed, the map grid comparable to the starting grid.

And do you know more about your business than you did. Which services people search for, which neighborhoods respond, what your competitors are doing that you were not.

Measure all four against your own trailing ninety days, never against a published average. There is no current benchmark for Business Profile performance worth the risk. The only large sample study of profile metrics is BrightLocal's, published in July 2019 on data from 45,264 businesses gathered between September 2017 and December 2018, and it has never been refreshed. Google has since retired several of the metrics it was built on, and views are now counted as deduplicated unique visitors rather than raw views, so its conversion figure cannot be reproduced on a modern account. Your own starting numbers are the only valid comparison you have.

A ninety day engagement that produces no ranking change but a working measurement system has still earned its money. That statement is also extremely convenient for anyone selling ninety day engagements, this one included, so here is the test that keeps it honest: the measurement setup has to be something you can log into and read without the agency present, and it has to survive their departure. If it does not, it was a report, not a system.

What you should not expect: your most contested head term in your densest market, a revenue line you can trace cleanly, or stability in AI answers. That last one confuses people, and it should not, because AI results change between runs of the same question by design. The St. Gallen measurement study found engine to engine differences that are large in themselves, with Gemini the most stable of the four tested and ChatGPT the least, and it found that ChatGPT activated web search on only 42.2% of runs. Checking daily produces anxiety, not information.

Which things move quickly and which do not?

Fast, within weeks. Profile completeness, categories, services and products, hours, photos, and review flow once the asking system is running. Technical fixes on your own site. Anything that was simply missing.

Medium, one to three months. Positions on longer, less competitive searches. Pages that answer a specific question that nobody local had answered. Conversion improvements on pages that already get traffic.

Slow, six months or longer. The most competitive term in a dense market. Authority built through mentions and links. Anything requiring a body of content rather than a page.

The uncomfortable one: some things do not move because of search work at all. If your listing looks fine and your calls go unanswered at 5pm, the constraint is the phone. A good agency will tell you that in the first month rather than sell you another quarter.

What to do this week

Before the engagement starts, ask for the ninety day picture in writing. One page, three sections, thirty sixty ninety, with what gets produced in each and what is expected to have changed by the end.

Run the review arithmetic yourself. Count your top three competitors' new reviews over the last year, divide by twelve, add one, multiply by three. That is your quarter's target and it took four minutes.

Then ask the question that should be uncomfortable for whoever is selling you the quarter, and that includes us. What specifically would count as failure at day ninety, written as a number, and what happens to the fee if it is missed? SEOD does not tie its fee to a ranking outcome, because nobody honestly can, and that means we have to answer the first half of the question in writing or admit the checkpoint is decorative. An agency that cannot imagine failure has not thought about accountability.

Then agree what you owe. Approvals within a set number of days, photos, access, and a monthly half hour. Write your side down too, because half of stalled engagements stall on the client side, and an honest look at the cost of doing it yourself versus hiring starts with the same accounting of your hours.

If you have not yet chosen, the questions to ask before hiring belong in the same conversation as the timeline.

Questions about what your first ninety days should look like?

Email me at eric@seod.com with your start date and the scope you have been given, and I will send back a checkpoint list for day thirty, day sixty, and day ninety that you can hold your agency to. It is written for you to use with them, not with me. No pitch attached and nothing sent afterward.

There is more on hiring and working with agencies alongside it.