If a finished plan lands on your desk in week one, they wrote it before they met you.
That is worth saying first because speed is the thing buyers reward, and it is the clearest early signal that you have hired a template. Nobody can understand a business in four days. What they can do is produce a document that looks like understanding, assembled from the last three clients in your industry, and most people receiving it feel reassured rather than suspicious.
So here is the claim I would defend. A good first ninety days produces almost no new marketing. It produces a small number of reversible fixes, one properly argued decision, and a great deal of finding out. If the campaigns are live in week two, the diagnosis was skipped.
That framing decides what you should be looking for at each stage.
Days 1 to 14: access and evidence
The first fortnight is admin, and the quality of it predicts everything.
They should be asking for read access to your ad accounts, analytics, Search Console, CRM and email platform, in your ownership rather than theirs. They should want your last twelve months of enquiries and, importantly, which of those closed. They should want thirty minutes with whoever answers the phone.
The tell is what they ask about outside marketing. Gross margin per customer. Which service line actually makes money. What your capacity looks like if this works. A consultant who never asks about the money is going to optimize channels rather than the business, and you will get a better cost per lead attached to a worse quarter.
The other tell is conversations. If nobody talks to your salespeople or your customers in the first two weeks, every recommendation afterwards is built on dashboards, and dashboards do not contain the reason people chose you.
What you should have by day fourteen: access confirmed, a list of what data does not exist, and no recommendations whatsoever.
Days 15 to 30: a finding, not a plan
At the end of the first month you want one page that says what is actually wrong.
A real finding is uncomfortable and specific. Your best channel is subsidized by branded search that would have converted anyway. Half your paid budget targets a city you cannot serve. Sales contacts inbound leads four days late, so the lead quality argument is misdirected. Your best margin service has no page on your site.
A weak finding is a list of everything. If the document says you should improve SEO, refresh creative, do more email, add video and fix your website, that is a description of marketing, not a diagnosis of your marketing.
And a genuinely good consultant will sometimes tell you the finding is that you should not spend more. That the constraint is fulfillment, or price, or that nobody answers the phone. I went through why the same lead can be both fine and junk depending on who is measuring separately, and that argument surfaces in a lot of these first-month findings.
Days 30 to 60: small reversible fixes, and one bet
Now things start moving, and the shape matters more than the volume.
The right pattern is a handful of cheap, reversible changes plus exactly one substantial bet.
Reversible means settings, tracking, exclusions, a rewritten page, a follow-up process. Things you can undo on a Friday afternoon if they are wrong. These should be shipping continuously and quietly.
The one bet is the thing that costs real money or real time. A new channel, a rebuilt landing page, a repositioning of the offer. One, because with five simultaneous changes you will never know which one moved the number, and in ninety days you do not have the volume to run five clean tests anyway.
If your consultant is proposing five bets, they are hedging against being wrong at your expense.
Days 60 to 90: the first measurable change
By day ninety you should be able to point at one number that moved and explain why.
Realistically that number is a conversion rate, a cost per qualified lead, a response time, or a close rate on a specific segment. Something with enough weekly volume to show a difference inside a month.
Be careful about what counts as measurable here. Paid search bidding needs roughly thirty conversions in a thirty day window before it performs reliably, so if you generate eight a month, a change to the account cannot be evaluated in ninety days no matter how good it was. The honest report says that out loud rather than showing you a chart with an encouraging slope.
The report itself should be readable in four minutes and should contain at least one thing that did not work. A ninety day review where everything went well is a document about the consultant, not about your business. The four numbers worth reporting is the version of this argument I keep coming back to.
What will not have moved, and you should not accept a story about it
Say this to each other on day one, in writing.
Organic search. The 2026 consensus across the tool vendors and agencies publishing timelines is three to six months for early measurable movement and six to twelve for anything compounding. Local and technical wins can appear inside ninety days. Competitive national rankings will not.
Brand. Nobody moves brand perception in a quarter, and anyone claiming to has redefined brand as impressions.
Anything with a sales cycle longer than the engagement. If your deals take five months, ninety days of work produces pipeline, not revenue. Judging it on closed revenue is a guaranteed way to kill work that was going fine.
Volume-starved channels. See above.
Agreeing this at the start protects both sides. It stops you cancelling something that was working and it stops them promising something that cannot happen.
Three failure modes worth naming
Activity in place of findings. Weekly updates full of tasks completed and no statement of what is wrong. Ask, once, what the single biggest problem is. If there is not an immediate specific answer at week six, there is not a diagnosis.
The plan that never narrows. Month three should have fewer priorities than month one, not more. Widening scope usually means nothing has been confidently ruled out.
Never saying no. Somebody who agrees with every idea you raise is managing the relationship rather than the work. The most useful thing an outside pair of eyes provides is a defensible refusal, and if you never get one you are paying for agreement.
What you owe them
Fair is fair, and half the engagements I have watched go badly failed on this side.
Access on day one, not day twenty. One decision maker, not a committee that reverses things after the call. An honest account of what you have already tried and what it cost, including the parts that were embarrassing. And a real answer about your margins, because the recommendation changes completely at 20% versus 60%.
Withhold those and you will get generic work, and it will be your doing rather than theirs.
The short version
Ninety days should buy you a diagnosis you did not have, a handful of fixes you can keep, one honest bet, and a clear-eyed account of what is still unknown.
If what you have on day ninety is a lot of activity and a chart that goes up, ask which decision it changed. The answer to that question is the whole test.
My notes on how I work cover the thinking in more detail. If you’d like to talk any of this through, drop me a line on email, WhatsApp or LinkedIn and we can have a quick chat. I’m contracted full time so this isn’t a pitch.