First 90 Days With a New Paid Ads Agency

On this page

1. Days 1 to 30: discovery and stabilization
2. Days 31 to 60: hypothesis and test
3. Days 61 to 90: first scaling decisions and the 90-day report
4. The four patterns that emerge by day 90
5. The pre-kickoff prep that pays off

The first 90 days are when an agency engagement either takes hold or sets itself up to fail. Most failures are visible early. The agencies you want to keep behave one way in the first quarter; the agencies you will eventually fire behave another way. Knowing the difference is what lets you correct course in time, instead of writing it off at month six.

This is the working framework. Three phases of 30 days each, what good looks like in each phase, what the warning signs are, and how to act on what you see.

Days 1 to 30: discovery and stabilization

The agency should be reading every campaign in every account. They should stabilize what is already working, pause what is clearly broken, and resist the urge to restructure things they have not yet understood. The first month is observation, not transformation.

What good looks like:

• A kickoff meeting in the first week with the actual team you will work with, not the sales lead.
• A written audit of the current state by the end of week 2, covering account structure, performance trends, creative inventory, audiences, and reporting gaps. The audit names specific issues with specific evidence, not generic best practices.
• A baseline performance read by end of week 3: where the account is, what is working, what is at risk, what could be optimized first.
• A revised forecast for the next quarter by end of week 4. Not "we will hit your goal." A specific scenario tree: best case, expected case, and downside case.

Warning signs:

• The agency immediately starts restructuring campaigns in the first two weeks. Either they were not paying attention to the audit they wrote, or they are showing off. Both end badly.
• No written audit by week 2. The audit is the operational equivalent of "did you actually read the account?" Skipping it means they did not.
• The kickoff meeting reveals the team is different from the team you were sold. Bait-and-switch is the most common early problem and you should escalate immediately.

What to do in this phase:

Ask for the audit in writing by a specific date. Review it line by line. Push back on anything generic. The first conversation about a specific account observation is the conversation that sets the tone for the rest of the engagement.

Keep your in-house team running their current cadence. Do not let the agency take over decisions yet. The handoff is gradual, not immediate.

Days 31 to 60: hypothesis and test

Now the agency moves from understanding the account to forming testable hypotheses. The hypotheses should be informed by what they learned in the first 30 days, not generic. Each test should have a specific setup, a clear success metric, a defined time window, and a documented kill switch.

What good looks like:

• Two or three specific hypotheses, each tied to something the audit surfaced. The hypotheses are written down, and the writing makes the bet explicit.
• A test plan for each: what is being changed, what is being held constant, what the success metric is, what time window the test runs, and what the agency will do if the test fails.
• Weekly written updates that report progress on each test. Not just numbers. Interpretation.
• An honest mid-month assessment: are the tests working? Should we kill any of them early?

Warning signs:

• The "tests" are generic things every agency runs. New audiences, new creative, new bidding strategy, all at once, with no structure. This is busywork dressed as strategy.
• A test runs for 90 days because nobody set an end date. Real tests have end dates.
• The agency conflates "we tried something" with "we tested something." The two are different. Tests have controls.
• Numbers come without interpretation. Reports that are just dashboards screenshots are dashboards screenshots, not analysis.

What to do in this phase:

Read every weekly update on the day it lands. Reply to each one with at least one question. The agencies that get used to a high-engagement client produce better work than the agencies that get used to an absent one.

If by day 50 the hypotheses are not clearly producing useful information (positive or negative), have a structured conversation about whether the right tests are being run.

Days 61 to 90: first scaling decisions and the 90-day report

By the start of the third month, the agency should have answers from the tests. Some will have worked, some will not have. The scaling decisions in this phase are informed by those answers.

What good looks like:

• A clear narrative about what the tests revealed. Not just "ROAS went up." A causal story: what was changed, what moved, why we think it moved, and what we plan to do with that knowledge.
• A scaling plan for the next quarter that is grounded in what worked, not in what the agency wants to do.
• A written 90-day report by end of week 12 that you could send to a board. The report covers: where the account was at day 0, what we did, what we learned, where it is now, where we go next, and what the risks are.
• A first conversation about the agency's recommended scope for the next quarter, including any retainer or scope adjustments.

Warning signs:

• No 90-day report. Or a 90-day report that is mostly graphs with no narrative.
• The scaling plan does not connect to what the tests revealed. The agency is doing what it usually does, not what your account specifically needs.
• The agency suggests changes to the contract terms (rate increase, scope increase) without first delivering the 90-day report. Sequence matters.

What to do in this phase:

The 90-day report is the artifact you keep regardless of how the engagement goes. If the agency is great, the report is the baseline for the next quarter. If the agency is not great, the report is the document you take with you when you find a replacement.

If by day 90 you do not have a written report you would be comfortable showing to a board, the agency has not earned the next 90 days yet. Have the conversation.

The four patterns that emerge by day 90

Across enough engagements, the first 90 days produce a few recognizable patterns. Knowing them helps you act faster.

The strong start. Audit was sharp. Hypotheses were specific. Tests produced information. The agency is asking smart questions and surfacing things you did not know. Keep going, and increase the scope thoughtfully.

The slow start that turns into a strong run. The first month was uneventful. The agency listened more than they pitched. By day 60 they were producing real insight. By day 90 you trust them. This is the most underrated pattern. Resist the urge to fire them at day 45.

The strong start that fades. The kickoff was great. The audit was sharp. Then quietly, the senior team rotated off. By day 75 you are mostly hearing from a junior. By day 90 the cadence is sliding. Address it now. The pattern will not self-correct.

The misfit. The audit was generic. The hypotheses were generic. The tests had no structure. The 90-day report is missing or weak. This is rare with agencies you vetted carefully, but it happens. End the engagement before month four. The cost of staying is higher than the cost of starting over.

The pre-kickoff prep that pays off

Before day 1, send the agency:

• A complete data export from every platform, going back 12 months.
• Your current campaign structure as a screenshot tree.
• Your unit economics one-pager (contribution margin, CAC target, LTV definition).
• A list of past hypotheses you have tested and the results, even informal.
• A list of three things you wish a prior agency had done differently.

Agencies who get this package are better positioned to write a real audit in the first two weeks. Agencies who never get it have to spend their first month assembling it themselves, which means the audit lands at day 35 instead of day 14.

The rest of the agency selection framework is in the pillar guide. Before you sign, run through the 47 interview questions. And if any of the warning signs in this piece show up early, the red flags article gives you the patterns to act on.