The First 90 Days of an Agency Account, in the Order They Actually Matter

The first ninety days of an agency account are worth more than the next ninety put together. Not because more work happens in them — often less does — but because the decisions taken in those ninety days determine whether the account will ever produce clean data, whether the strategy has a baseline to be measured against, and whether the relationship is one where the agency’s word is trusted or perpetually second-guessed. Get the sequence right and month four is a strategic conversation. Get it wrong and month four is a re-do of month one.

This post is that sequence, in the order that actually matters, from the perspective of the person running the account rather than the person selling the retainer. It is short by the standards of this site because the point is the order and not any single item in it.

The four things, in strict sequence

The order is: access, tracking, baseline, spend. Every account I have watched succeed did these in this sequence. Every account I have watched struggle skipped at least one of the first three and tried to make it up later.

The sequence is not a checklist that you can rearrange to fit the client’s preferred timeline. It is a dependency chain. You cannot verify tracking without access. You cannot get a baseline without tracking. You cannot judge spend without a baseline. If a client wants you to skip a step to “start seeing results faster”, what they are actually asking for is to make month four unmeasurable — and month four unmeasurable is what makes a retainer eventually feel like it isn’t working, whether it is or not.

Days 1-14: Access

Access is the boring step, and it is the step that decides everything else. The rule is: on day one, request every single credential you will need, in one email, and do not do any billable work until you have all of them.

The credentials for a paid-media retainer usually include: ad accounts (Google Ads at account and MCC level, Meta Business Manager as admin, LinkedIn if applicable), analytics (GA4 Editor at minimum, Google Tag Manager Publish), the DNS or CMS for pixel/tag placement, the CRM’s admin-side for offline conversion imports if that will run, and the payment/e-commerce backend for validating actual revenue against reported revenue.

The mistake here is not asking; it is being polite about receiving. You are told “we will send them over by end of day” and by end of week you have three of the seven, with the other four pending “we need to loop in IT” or “the person who has that has gone on leave until Monday”. Every day the missing access persists is a day the tracking check you are supposed to do in step two cannot happen, and the baseline in step three has less data.

The specific move that avoids this: name a hard deadline for full access, in the same email as the initial request, and say what happens after it. “If we do not have all seven items by end of this week, I will not be able to give you a strategy by day 30. Please push whichever ones need internal approval today.” That sentence is uncomfortable and it is the difference between month one being productive and month one being a series of Slack pings.

The second common failure at this stage is accepting “user” access when you need “admin”. User access to Google Ads lets you view. It does not let you install a new conversion tag, change attribution settings, or link accounts. You will discover this halfway through step two, at which point you have to re-request, which pushes the whole sequence out by another week. Read the specific permission requirement on the same day you request access, not the day you try to change something.

Days 14-30: Tracking

Tracking is not a step you complete; it is a step you verify. The distinction matters because “tracking is set up” is the default assumption every client has about their account, and it is wrong roughly 90% of the time. You are not adding tracking; you are auditing tracking and fixing whatever the audit finds.

The verification, in order:

1. Fire the primary conversion yourself and confirm it lands. Not “the client told us the checkout tracking works” — go through the checkout with a real card, on the primary browser your target audience uses, and confirm the conversion shows up in the ad account, in analytics, and in the CRM within the expected timing. This single step catches roughly half of all silent tracking failures. Google’s own guidance is unambiguous — Set up conversion tracking says the tracking must be created and verified BEFORE campaigns launch, precisely because corrupted conversion data teaches Smart Bidding to waste budget from the first day.

2. Reconcile ad-platform revenue against back-office revenue for the last 30 days. If Google Ads’ converted revenue is 47% of what the CRM has, one of them is wrong. Usually it is the ad platform (view-through attribution, deduplication, cookie loss), and the exact discrepancy is what your reporting has to explain going forward. If you never do this reconciliation, your monthly reports will show numbers the client’s finance team cannot see in their bank account, and the credibility gap is the credibility gap for the whole engagement.

3. Check the conversion definitions match the client’s definition of a conversion. “Purchase” in the ad account is often “add to cart plus 30 days” or “landing page view after ad click plus 24 hours” — an event that is far softer than the client believes they are being reported on. Change the definitions to match the business, not the platform’s default.

4. Instrument the transitions, not just the pages. A tag that fires on the thank-you page tells you nothing about whether the checkout button worked. A heartbeat on the transition — cart submitted, payment gateway called, thank-you page reached — tells you where the funnel actually breaks. The transitions are almost always the invisible failure mode.

The failure of the tracking step compounds. Every account that skipped a real tracking audit spent the first month running campaigns whose reported conversion count could not be reconciled to real revenue, then spent the second month re-running that first month because the strategy had been built on unreliable numbers. This is what “lost the first month re-running it” means in the sequence: not that the month was wasted, but that its data was, and everything downstream then rests on nothing.

The specific case for affiliate accounts, where the postback failure modes are subtler than the general marketing case, is covered in how affiliate tracking breaks quietly — same discipline, sharper edges.

Days 30-60: Baseline

A baseline is the number you compare all subsequent numbers against. Without one, every result you produce for the next quarter is a floating claim that has to be justified in isolation, and it is exhausting to defend a floating claim every month.

The baseline needs to answer three specific questions, in writing, before you touch spend:

What was the account doing before we arrived? Not “the account was underperforming” — a real four-week averaged number, per channel, for the specific KPIs you will report on. If the client says “our CAC was around ₹1,200”, you write down whether it was ₹1,200 across all channels blended (which is a lower-quality signal), ₹1,200 on prospecting alone (which is the useful one), or ₹1,200 on retargeting-heavy blended (which flatters the number). The baseline is only useful if it is like-for-like against what you will report next.

What are the confounds we know about? If the last four weeks included a promotion, that is a confound. If a competitor’s launch spiked the auction, that is a confound. If tracking was measurably broken (see step two), that is a MASSIVE confound and the baseline may need to be reconstructed from the CRM rather than the ad platform. Write the confounds into the baseline document so that six weeks later, when the client’s boss asks “why did CAC go up in June?”, you can answer without guessing.

What is the expected trajectory in the next 60 days at zero intervention? If you left the account running exactly as it was, where would it be? Some accounts drift up in cost even with no change — the market moves, creative fatigues, seasonality bites. The baseline should include a “do nothing” projection, so that when you present month-three results, you are comparing to what would have happened without you, not just to what was happening at the start.

That third question is the one most agencies skip, and it is the one that decides whether your work is legible as work. A CAC that went from ₹1,200 to ₹1,100 looks like a 9% improvement — but if the do-nothing projection was ₹1,400, the real improvement is 27%, and being able to say that in a way the client can verify is what buys you the next quarter.

Days 60-90: Spend

Only now do you turn up the spend dial. Everything before this is the setup that makes spending money legible; without the setup, spending money is a bet on a scoreboard you cannot read.

Three rules for the spend phase:

Change one thing per week, not five. The temptation in month two is to overhaul everything — new audiences, new creative, new landing pages, new bidding. That produces a beautiful new month-three number that cannot be attributed to anything in particular. Change one thing per week, so that when a number moves, you know which lever moved it. This is slower than the client wants and it is faster than what most agencies actually deliver, which is “five things at once and a hopeful monthly report”.

Set the payback rule before the spend, not after the result. Decide before you increase spend what your payback threshold is — the payback window inside which a new customer’s revenue has to cover their acquisition cost. The wrong time to decide the threshold is when a channel just missed it; the right time is now, in writing, so that “keep spending” and “cut” are decisions with a rule behind them rather than negotiations. The full arithmetic is in how a profitable retainer quietly becomes an unprofitable one for the retainer economics and performance marketing the practical playbook I actually use for the channel-level detail.

Report at 60 and 90, not weekly. Weekly reports produce noise. Monthly ones let a change accumulate signal. Two written check-ins in the first 90 days — one at day 60 (what did we do, what changed, what is the running theory) and one at day 90 (three-month reconciliation against baseline, next-quarter proposal) — are the reporting cadence that respects both the client’s need to see progress and the account’s need for real signal. Anything more granular is data theatre.

The day-30 conversation nobody wants to have

At day 30, one specific conversation almost always needs to happen and almost never does: “here is what we found in the tracking audit that changes what we thought we were doing.”

It is uncomfortable because it usually reveals that the previous agency, or the previous in-house team, was reporting numbers that were not what the client thought they were being told. Nobody comes out looking clean. The client feels foolish for not having caught it; the previous agency is being implicitly criticised; you are the messenger, which is a role with well-known career risks.

Have the conversation anyway, on day 30, with the specific findings written down. Two reasons.

The first is that if you do not have this conversation, you will spend the next quarter reporting numbers the client’s boss can compare unfavourably to the numbers the previous agency was reporting, and every month you have to explain why yours are lower. That explanation eventually costs you the account.

The second is that the client’s trust in you is highest in month one, before you have shipped anything they can second-guess. That is the month to say the difficult things, because they land as evidence of competence rather than as excuse-making. Month four is too late.

The specific shape of the conversation is: “During the tracking audit we found that [specific thing]. This means the number you thought was X was actually Y. Going forward we will report on Y; the baseline in the report will be re-stated so we can compare like-for-like. Here is the document.” Do not soften it, do not explain it away, and do not skip the “here is the document” — the document is what makes it a transition rather than an accusation.

What the sequence looks like when it goes wrong

For contrast, here is the failure pattern I have seen most often, from account intakes I have taken over from other agencies:

• Days 1-14: access is partial. Ad accounts granted, analytics not, CRM promised.
• Days 14-30: the agency launches “quick-win” campaigns to show early progress, because access is still incomplete and tracking cannot be verified without full access, and the pressure to show something is real.
• Days 30-60: quick-win campaigns produce numbers the client cannot reconcile. A month is spent explaining the discrepancy rather than fixing it. Full access finally arrives. Tracking audit reveals the reported numbers from month one are wrong.
• Days 60-90: baseline is finally established from a real audit. The strategy pitched in month one is now known to have been based on the wrong numbers and needs to be reworked. The client asks the reasonable question “what were we doing in months one and two?” and there is no good answer.
• Month four: the account is either fired or renewed at a discount that reflects the client’s belief that the first 90 days were wasted. The agency reports “we are seeing early momentum” and the client hears “you have not paid me enough”.

The reason this happens is not laziness. It is the same commercial pressure everywhere: month one is expected to show progress, so the agency skips the setup that would make progress measurable, and by the time the measurement is real the account has already lost its trust reserve.

The way out is to name the pattern in the sales conversation, not the onboarding one. “The first 30 days of the account will feel slow because they are — we are getting access, verifying tracking, and establishing a baseline. If the reporting in month one shows numbers that look like results, we are lying to you. In exchange, month four will be a real strategic conversation and not a re-do of month one.” Clients who do not accept this framing are clients who are going to be difficult to serve honestly, and it is better to know that before you sign than after.

Where this fits with the rest of the work

The 90-day sequence is one input into the larger question of whether a retainer is priced correctly for the delivery model. If access, tracking and baseline are done properly, month four’s re-scope conversation has real evidence behind it — and that conversation is in how a profitable retainer quietly becomes an unprofitable one.

If your problem in the first 90 days is not the sequence but the specific tracking failure modes — postbacks, discrepancies, attribution drift on the affiliate side — the sharper version is in how affiliate tracking breaks quietly and how to catch it before payout day. The mechanics are different from paid media; the discipline is the same.

And if the paid-media strategy itself is the question — what to actually spend on once the sequence completes — the playbook is in performance marketing: the practical playbook I actually use.

FAQ

How long should day-one access take before it becomes a real problem?

If you do not have every credential you need by end of week one, the 90-day sequence has already slipped. Name the deadline in the same email as the initial request and say what happens if it is missed, otherwise “we are chasing IT” quietly consumes the entire access window.

What if the client refuses admin access and only offers user-level?

Push back before you accept the engagement, not after. User access to Google Ads cannot install a conversion tag or change attribution settings, per Google’s own permission levels, and discovering this in week three costs you the tracking step. A client who will not grant admin to their agency is a client who is not going to trust the agency’s numbers either.

Should I launch quick-win campaigns while I wait on tracking, or hold spend entirely?

Hold spend. Quick-win campaigns launched before tracking is verified generate numbers the client cannot reconcile in month three, which is the exact failure pattern the sequence exists to prevent. The commercial pressure to show early motion is real, and giving in to it is what makes month four a re-do of month one.

How much reconciliation gap between the ad platform and the CRM is normal?

Some gap is expected — view-through attribution, cookie loss and deduplication all pull the ad platform’s number away from the back-office number. What matters is not the size of the gap but whether you can explain it; a 30-40% delta that you can account for is fine, an 8% delta you cannot explain is not.

Do I need to redo the baseline if we find tracking was broken?

Yes, and reconstruct it from the CRM rather than the ad platform where you can. A baseline built on numbers you have already established are wrong is worse than no baseline, because every subsequent report inherits the error and the credibility gap widens each month.

What if the client asks for weekly reports instead of the 60 and 90 cadence?

Offer a weekly one-line status update by email — what shipped, what is next — and hold the written reports to day 60 and day 90. Weekly written reports produce noise the client then asks you to explain, which turns the reporting cadence into the work rather than a reflection of it.

How do I have the day-30 conversation without making the client feel foolish?

Frame it as a tracking finding, not a previous-agency failure, and hand over a written document at the same time so it reads as a transition rather than an accusation. The document is what turns the conversation into evidence of your competence instead of an implicit criticism of anyone who was there before.

The one-sentence version

Do the first 90 days in strict order — access, then tracking, then baseline, then spend — refuse to skip a step even under pressure to show early progress, and have the day-30 conversation about what the tracking audit found while your credibility is still highest. Every account I have seen that did this properly had a strategic month four; every account I have seen that skipped a step used month four to re-do month one.