When to fire your marketing agency — and when the problem is not the agency

Fired too late, fired too early: both are expensive

An agency relationship that should end but does not costs you compounding mediocrity — 10% inefficiency on a ₹20 lakh monthly budget is ₹24 lakh a year, quietly. But the opposite mistake is just as common: firing a competent agency over one bad month, resetting platform learning, losing a year of creative insight, and starting the same cycle with a new logo on the deck. We are an agency; we have been on both ends of this. What follows is the calibration we wish every client had — including the section on when the problem is not the agency at all.

The discipline is to fire on structural signals — how the agency reports, tests, produces and behaves — rather than on outcome noise, because outcomes in any single month are mostly variance, seasonality and platform weather. Structure predicts; noise misleads. The six flags below are structural, they are checkable from your side of the table this week, and none of them requires you to understand a single campaign setting.

Red flag one: vanity reporting

The single most reliable predictor. Warning signs: the monthly report leads with platform ROAS and never shows MER or contribution after ad spend; retargeting and branded-search performance are blended into headline numbers to flatter them; metrics change month to month so nothing is comparable; screenshots replace spreadsheets; and delivered-order revenue never appears in a COD-heavy business. None of this is accidental. Reporting structure is a choice, and an agency that chooses opacity is managing your perception instead of your account.

The test is simple: ask for one page showing total spend, total delivered revenue, MER, and contribution after ad spend, monthly, for the whole engagement. A good agency produces it in a day and probably already sends it. Weeks of delay, or a lecture about why blended numbers are misleading, is your answer. And note what vanity reporting costs beyond deception: every quarter it survives, your own team’s intuitions are being trained on fiction — the damage outlasts the agency that caused it.

Red flag two: no incrementality agenda

An agency paid on or judged by attributed ROAS has a financial interest in never testing whether the attribution is real — which is why the absence of an incrementality agenda after six months is a structural signal, not an oversight. You are looking for: a standing test calendar; at least one holdout or geo test run without you asking; retargeting and branded search sized on tested lift rather than dashboard credit; and honest labelling of what cannot be measured.

The self-serving version of this flag matters most: an agency that volunteers tests likely to shrink its own attributed numbers — branded search holdouts, retargeting holdouts — is showing you its incentives are aligned with your P&L. An agency that resists them is telling you where its numbers come from. If you have never seen the case for these tests, our note on measuring the halo effect explains what dashboard attribution systematically misses.

Red flag three: creative volume theatre

Scaled accounts live and die on creative production, so agencies have learned to perform it. Theatre looks like: twenty new ads a month that are one concept in five aspect ratios and four headline swaps; no articulated hypothesis behind any test; winners flogged until they die with nothing behind them; no learning log, so failed ideas return quarterly in new fonts. Real production looks like: genuinely distinct concepts shipped on a contracted monthly quota, each test tied to a hypothesis, and a written record of what the account has learned about hooks, formats and angles — the discipline described in our creative testing framework.

The test: ask what the account learned from creative in the last quarter, and what next month’s concepts are testing as a consequence. Fluent answers come from a system. Improvised answers come from a render queue.

Red flags four and five: access and the vanishing seniors

Access: you must own the ad accounts, pixels, data and creative files, with the agency working inside accounts you control. An agency that runs your media from accounts it owns, resists admin access, or goes vague about data portability has built its retention strategy out of switching costs — which tells you what it thinks of its service as a retention strategy. This alone justifies an exit, on principle and on risk.

The vanish: the senior team that pitched you dissolves within a quarter into a rotating cast of juniors; response times stretch; the strategist appears only at renewal. Some delegation is normal and priced in — juniors execute everywhere, including here. The flag is the gap between the seniority you were sold and the seniority you receive, and whether anyone senior still knows your business without a briefing document. Pair this with rotating blame — every miss attributed to iOS, CPMs, seasonality or your website, with no miss ever owned — and you have an agency spending its senior hours managing the relationship instead of the account.

A good agency tells you bad news before you find it. The month you start discovering problems in the data before your agency mentions them is the month the relationship started ending.

Red flag six: strategy stagnation

The quietest flag, and the one retained agencies earn most often: the account still looks exactly like it did nine months ago. Same campaign structure, same three audiences, same channels at the same split, no test that risked anything, no proposal that would change next quarter. Healthy accounts leave a visible trail of decisions — structures retired, channels trialled, budget migrated toward what the tests said. A stagnant account is not stable; it is depreciating politely, because markets, CPMs and creative tastes moved while the media plan did not.

Distinguish this from disciplined consistency: an agency holding a winning structure steady while shipping fresh creative through it and testing around its edges is doing the job. The tell is the answer to one question: what did we try in the last quarter that did not work? A confident, specific answer means the account is alive. A pause means the retainer is.

What good looks like, so you can calibrate

Before acting on any of the above, check the other column. A good agency: delivers bad news early and unprompted, with a diagnosis and a plan; reports MER and contribution beside platform numbers without being asked; runs a standing test calendar including tests that could shrink its own attributed credit; ships distinct creative on a quota with a visible learning log; keeps everything in accounts you own; escalates real decisions to people who know your business; and pushes back on your bad ideas in writing. If your agency does most of this and last quarter was still poor, the honest next step is a diagnostic conversation, not a termination letter — by these standards the problem may be the offer, the site, the margin structure or the market, and a new agency inherits all of it. Agencies that meet this bar are rarer than they should be; how to find them is its own guide — see how to choose a D2C agency.

How to run a clean transition

Decided? Then sequence it like an operator, not like a breakup. The goal of a transition is that the account never notices it happened: spend steady, learnings preserved, platforms undisturbed. Every step below exists to protect one of those three.

• Audit access before anything. Quietly confirm admin ownership of ad accounts, Business Manager, pixel, product feeds, analytics, landing pages and creative files. Fix gaps while the relationship is still warm — this is leverage you only have once.
• Export the history: full reporting, creative library with performance data, test results, audience structures. The learnings are yours; a professional agency hands them over without friction.
• Select the replacement before giving notice — run the process properly with an evaluation checklist — and negotiate a two-to-four-week overlap where the incoming team shadows the account live. The overlap is awkward and worth it.
• Do not pause spend to punish anyone. A dark account resets platform learning and hands your shelf to competitors; you will pay re-entry costs for months. Keep spend at maintenance level through the handover.
• Honour the notice period and part professionally. The Indian D2C world is small, your next agency will hear how you exited, and the outgoing team’s cooperation in week two of transition is worth more than a point of principle.