Every agency proposal looks its best when you are asked to sign. The deck is polished and the logos are impressive. Yet the contract may request 12 months of budget from an untested team.
That is a real risk. You are buying a year of output after an hour of selling.
A pilot narrows the gap between pitch and proof. Give the agency 90 days, one mission, and written success metrics. Then decide from the evidence.
This playbook covers scope, metrics, contract terms, and the verdict. Our day-by-day accountability checklist explains what the agency owes during those 90 days. This post explains how to structure the pilot itself.
Why Does a 90-Day Pilot Beat a 12-Month Contract?
A 90-day pilot limits risk for both sides. You avoid funding a year of unproven work. The agency gets enough time to launch tests and produce an initial cost read. A bad fit costs one quarter, not one year.
An annual retainer makes agency revenue predictable, but it does not guarantee your results. Marketing performance does not need 12 months to become observable. It needs a clean test, enough budget, and an honest readout.
Good agencies like pilots too. Defined criteria protect them from clients who move goalposts mid-engagement. A documented win is also a stronger renewal argument than a signature from last January.
One caveat: a pilot is what you run after an agency passes basic screening. The 12 questions to ask before hiring a growth agency come first. The pilot tests the answers.
How Should You Scope an Agency Pilot?
Scope the pilot to one channel or system, never “everything.” Pick one paid platform, a lifecycle rebuild, or a creative testing program. Narrow scope makes the verdict easier to interpret. Broad scope blurs attribution.
When an agency touches five channels at once, nobody can say which effort moved the number by day 90. You end up judging vibes. A single channel against a single baseline keeps the pilot honest.
Scope also means resourcing. Agree on the media budget floor before launch. An underfunded test may lack enough data for a useful verdict, regardless of agency quality.
List client dependencies beside the agency deliverables. Access delays, slow approvals, and missing creative can consume the test window. Give each dependency an owner and due date. The final review can then separate execution failure from client-side delay.
Choosing the first lever may require strategy work. Our growth strategy service starts with that diagnostic. It finds the highest-value pilot before you commit test budget.
What Success Metrics Should You Fix in Writing Before Day 1?
Fix three checkpoints before the pilot starts. Review leading indicators by day 30, cost metrics by day 60, and pipeline by day 90. Name each metric, baseline, and owner. Without a signed baseline, the final review becomes a debate.
Here is the arc, phase by phase.
| Pilot phase | What to expect | What to measure |
|---|---|---|
| Days 1 to 30 | Access, tracking audit, signed baseline, first tests launching | Leading indicators: CTR, conversion rate, CPC vs baseline |
| Days 31 to 60 | Learning phases cleared, losers cut, weekly reporting running | Cost: CAC or cost per qualified lead trending |
| Days 61 to 90 | Winners scaled, playbook written, verdict review held | Pipeline: qualified opportunities and attributed revenue |
Day 30 is often too early for a final cost verdict. Meta’s current budget guidance recommends funding campaigns for at least seven days so delivery can learn. Many buyers use 50 weekly optimization events as a planning heuristic, not a guaranteed exit rule. Early reads should stay directional and use the agreed baseline.
Cost metrics may become useful around days 45 to 60 when conversion volume supports them. That is when CAC or cost per qualified lead becomes a fair discussion. By day 90, ask whether the work produced credible pipeline.
Want a pilot that arrives pre-scoped?
Our 90-Day Jumpstart turns this playbook into a fixed engagement. The diagnostic lands by day 30. Experiments run from day 31, followed by a 6 to 12 month roadmap. It is standalone, with no required long-term contract.
Book a Free Strategy CallWhat Contract Terms Make a Pilot Safe?
Four terms reduce pilot risk. Use a fixed 90-day term with no auto-renewal. Continue month to month with 30-day notice. Keep ownership of accounts, data, and creative. Use a flat fee that does not rise with spend.
Founders often overlook ownership. If the pilot runs in agency-owned accounts, important history may leave with the agency. You should be able to exit on day 91 with every account and asset intact.
Fee structure matters during a pilot. Percentage-of-spend pricing often runs 10% to 20% of ad budget, according to Feedbird. The agency earns more as spend rises during your efficiency test. A flat pilot fee removes that incentive.
Write each deliverable into the agreement. Require the baseline within two weeks, weekly reports, and a final playbook. You keep that playbook regardless of the verdict.
Define the data source for each metric in the same document. A CRM pipeline report and an ad-platform dashboard can tell different stories. Pick the source of truth before either side sees the result.
Include a minimum sample or spend threshold where possible. If the pilot misses it, label the verdict inconclusive. Do not turn a low-volume result into a pass or fail.
How Do You Judge the Pilot Verdict?
Judge the pilot against criteria set before day 1. Book the final review when the pilot begins. A pass beats the baseline with credible measurement. A fail misses both performance and process standards. An inconclusive result has promising evidence but too little data.
A pass does not mean the work is finished. It means the system works, so you continue month-to-month and expand scope one lever at a time.
A fail deserves an honest cause of death. Sometimes the agency underdelivered. Sometimes the tracking was broken or the budget never reached significance, and the same pilot would fail with any agency. Name the cause in the review, take the playbook, and leave cleanly. The terms above make that easy.
An inconclusive verdict can invite endless extensions. Allow one 30-day extension with a single named metric. If that metric still lacks a usable result, close the pilot.
Where Does the 90-Day Jumpstart Fit?
The 90-Day Jumpstart is our productized version of this pilot. Days 1 to 30 cover diagnosis. Experiments run during days 31 to 60. The final phase scales winners and creates a 6 to 12 month roadmap. You keep that roadmap either way.
We productized our entry engagement because this structure kept beating open-ended retainers for first engagements, ours included. Every term in this playbook is already built in: flat fee, no auto-renewal, month-to-month after, and your accounts stay yours. The day-90 review closes with a written recommendation, including when that recommendation is to take the roadmap in-house.
Run the pilot, with us or with anyone. The structure is what protects you.
Like this? Get the next one.
Short emails. New posts as they ship.
