An agency brief works as a filter rather than a wish list. One page carrying the number you are trying to move, your real figures, the constraint and a decision date does more work than a nine-page RFP, because it lets the wrong agencies withdraw early.
- Score five criteria with weights: diagnosis 30%, named team and time 20%, evidence at your size 20%, operating cadence 15%, commercial terms 15%.
- A weak diagnosis disqualifies an agency regardless of the total, and so does refusing to name the people who will do the work.
- Ask for a client at your size who left in the last year, then call them. That conversation beats the three references they volunteered.
- Replace the pitch with a paid ninety-minute working session on your own data, same problem for both finalists, their named leads only.
Most agency briefs I get asked to review are shopping lists. Two pages of deliverables, a budget band, a start date, and a line at the bottom asking for case studies. Send that out and six agencies send back six decks that look like each other, because you asked for a proposal instead of an opinion.
The brief is the one part of this process you fully control. It decides who bothers to respond, what they respond with, and whether you can compare the responses side by side four weeks later. A good one is shorter than the bad one and does more work.
I sit on the brand's side of this table. I do not run creative, buy media, or write code, so I have no retainer to steer you toward. What I do is help operators work out which lever matters next and who should pull it, which means I read agency responses with the client and sit in the calls next to them rather than across from them. A three-person team I worked with went from $8M to $42M in annual revenue over 28 months, and a real share of that came from bringing in two strategic partners who were better than the ones already on the account. Good agencies compound growth. This process is built to find those and let the rest withdraw early.
If you are still deciding whether an agency is even the right shape of help, start with who does what across experts, agencies and consultants and the fractional advisor versus agency versus hire math. Everything below assumes you have already made that call and are choosing between agencies.
The brief exists to
help the wrong agencies
say no.
A nine-page RFP full of deliverable checklists reads like procurement, and it gets procurement answers. What you want instead is a page that gives a capable agency enough real information to form a view, and gives an incapable one a reason to pass.
Seven blocks, one page. Write it once and send the identical document to everybody, so that when the responses come back the differences are theirs and not yours.
| Block | What you write | What it filters |
|---|---|---|
The number | The single metric this engagement is judged on, with today's value | Agencies who cannot move that metric withdraw |
Current state | Revenue band, AOV, contribution margin, blended CAC, channel split, platform | Separates the ones who read numbers from the ones who read briefs |
Already decided | What is not up for debate: platform, brand system, roles you keep in-house | Kills the pitch that quietly rescopes into a rebrand |
The constraint | What you tried, what happened, what you think is in the way | Strong agencies argue with it, weak ones ignore it |
Money | A real range, with media and fees split apart | Ends the two-week dance about budget |
Your side | Who they work with weekly, and how many hours that person actually has | Shows whether you have the capacity to be a good client |
Dates | Response deadline, working session week, decision date, start date | Tests whether they can hit a deadline before you pay for one |
Two things people get wrong here. The first is hiding the budget, which produces responses priced to win rather than priced to work, and costs you a round of negotiation you did not need. The second is asking for free work. Spec creative and sample audits get handed to whoever on their bench has capacity, which is rarely the person you would be buying, and the good shops decline the request anyway.
Eleven questions,
and what a strong
answer sounds like.
Ask them in writing first, then again out loud with the named leads in the room. The written answer shows you how the agency thinks. The spoken one shows you who in the building is doing the thinking, which is usually a different question.
| Question | Weak answer | Strong answer |
|---|---|---|
Reading our numbers, what is our real constraint? | "Lots of opportunity here. We would start with a full audit." | Names one constraint, shows the arithmetic behind it, and tells you which of your stated goals they would drop |
Who does the work, and what share of their week is ours? | "You get a dedicated team." | Four names, their other accounts, hours per week, and who covers when one of them is out |
Show me a client at our size where results went flat. | "That has not really happened to us." | Tells the story with numbers, says what they changed, and offers the contact |
What do you need from us every week? | "Just approvals and assets." | A specific list with hours attached, plus what stalls when it slips |
How do you charge, and what happens if spend drops? | A percentage of spend, no discussion of the floor | Raises the incentive conflict before you do and proposes a structure that survives a bad quarter |
What would make you fire us as a client? | "Nothing, we love a challenge." | Names the conditions, usually slow approvals or a founder who reverses decisions |
Five more, less quotable but worth the time. How do they define incremental, and do they look at anything beyond platform-reported numbers when they say a channel worked. What would they stop doing in the first month. Who owns the ad accounts, the pixel, the creative files and the raw footage when the relationship ends. What their first 60 days looks like, dated. And what happens internally when the number goes sideways for three straight weeks, because that is the month you will actually be buying.
"The most useful question in the whole process is what they would tell you to stop doing. An agency that wants the retainer badly enough will not answer it."
You will notice none of these ask about process, tools, or awards. Process decks are the cheapest thing an agency produces. A named person willing to disagree with your brief in writing is the expensive thing, and it is what you are paying for.
A scorecard you can
fill in while the call
is still live.
Five criteria, weighted, each scored 1 to 5. It fits on one screen and it takes about a minute per agency once the responses are in front of you. The weights matter more than the precision of any single score, because they force you to say out loud that a sharp diagnosis beats a cheaper fee.
| Criterion | Weight | A 5 looks like |
|---|---|---|
Diagnosis | 30% | Names your constraint, shows the math, and disagrees with something in your brief |
Named team and time | 20% | Real people, real hours, and the seniority that showed up to the pitch stays on the account |
Evidence at your size | 20% | Two references in your revenue band with a similar margin structure and channel mix |
Operating cadence | 15% | A weekly rhythm you recognize, a report you would actually read, escalation defined in advance |
Commercial terms | 15% | Fees whose incentive points at contribution, 30-day notice after a pilot, you own the accounts and the files |
Two disqualifiers sit outside the arithmetic. A 1 or 2 on diagnosis takes an agency out regardless of the total, because everything else you are buying rests on their read of the problem. So does refusing to name the people who will do the work, which is almost always a staffing plan they do not want to show you.
Set a floor and hold it. Below roughly 3.5 weighted, run the search again rather than signing the least bad option in the pile. Score independently before anybody talks, then compare, because the spread between two scorers on the same response is the most interesting number the process produces. Three scorers is plenty, and a fourth will cost you a week.
Run the same five criteria against the alternatives while you are at it. An in-house hire and a fractional operator both score on diagnosis, time, evidence, cadence and terms, and sometimes one of them wins. That comparison is the whole subject of the first operator hire and of whether a growth consultant is worth it at your size.
Replace the pitch
with ninety minutes
of real work.
Give the two finalists the same anonymized dataset and one decision you genuinely have not made yet. Ninety minutes, their named leads only, no deck. You take notes and try not to fill the silences.
What you learn in that room does not show up in any written response. Whether the lead reaches for the data or for a framework. Whether they can hold a position when you push, and change it when the number says they should. Whether the senior person talks over the person who would actually run your account, which tells you exactly how month four will feel.
A hardware brand I worked with went from $6M to $28M in GMV over 18 months, and getting there meant upgrading development, creative and paid partners at roughly the same time. The working session is how we chose all three. The development shop we picked spent most of its ninety minutes cutting the roadmap we had sent them, which was annoying in the moment and correct by the second sprint.
Pay for the session if you are asking for real analysis. A few thousand dollars buys you their attention and removes the awkwardness of asking a business to work for free, and it puts the two finalists on identical footing.
Ask for the client
who left.
Every agency has three references ready. Ask for a fourth: a client at your size who stopped working with them in the last year. The reaction to the request tells you as much as the call does, and the agencies worth hiring usually hand one over with a shrug and a story.
On the calls, skip the general questions. Ask what changed in month four. Ask who was still attending the weekly after month two, and whether it was the person who pitched. Ask what work the client ended up doing themselves that they thought they had bought. Ask whether they would hire the same lead again if that lead moved to a different agency, which is the question that separates a good shop from one good employee.
Check tenure on the named lead too. Agency reputations lag their current staffing, and the case study you liked was often produced by three people who have all since left.
The scorecard should
survive into the
contract.
Whatever earned the fives needs to appear in writing. If the diagnosis was the reason you picked them, the first 30 days should produce a dated version of it. If named people were the reason, name them in the agreement along with what happens if they roll off. If terms were the reason, keep the notice period short enough that the second quarter has to be earned.
Start with a paid pilot wherever the work allows it, usually 60 to 90 days with two concrete outputs and a review date on the calendar before anybody signs. Migrations and replatforms are the exception, since nobody can sensibly half-build a checkout.
Then be a client worth having. The agency relationships that produce the numbers at the top of this post had an owner on the brand side with real hours in their week, weekly decisions made on time, and a fee structure that pointed at contribution rather than at activity. If you want the arithmetic on that last part, it runs through contribution margin and your max allowable CAC, and measuring the return on outside help covers how to judge it after the fact. Discipline-specific reality checks live in the Google Shopping agency cost math and the honest take on Shopify SEO agencies.
The process side of this, how to run all of it inside four weeks instead of losing a quarter to discovery calls, is in how to run an agency search without losing a quarter.
Questions that come
up on every
search.
Q: How many agencies should I put in an RFP?
Five to seven for the written brief, two for a working session, one for a paid pilot. Fewer than five and you are comparing a small sample against your own assumptions. More than seven and the scoring turns into admin, responses start arriving after your decision date, and you end up giving the slowest agency an extension because you have lost track of who is still in. The written round is cheap for both sides, so spend your generosity there and be strict about who gets your ninety minutes.
Q: Should I share real numbers with agencies before I hire one?
Yes, under a mutual NDA, and it is the single biggest upgrade most brands can make to their brief. Revenue band, AOV, contribution margin, blended CAC and channel split are enough for a capable agency to form a view, and without them every response comes back generic because it has to. Hold back customer-level data, supplier pricing and anything that would matter in a competitor's hands. If an agency cannot be trusted with a margin figure under NDA, that is useful to learn during the search rather than in month three.
Q: Is a paid pilot better than a three-month minimum?
Usually, and the difference is what the pilot has to produce. Give it 60 to 90 days, two dated outputs, a review meeting booked before signature, and a fee that reflects real senior time rather than a discount you will pay for later in staffing. Pilots work badly for anything you cannot half-build, so a replatform, a checkout project or a migration should be scoped as a project with milestones instead. For creative, paid media and retention work, a pilot tells you more in eight weeks than a reference call ever will.
Q: How do I score agencies that do completely different things?
Use the same five criteria and change what a 5 means inside each one. A creative shop earns a 5 on evidence by showing work that sold at your price point, while a development partner earns it by showing a build at your traffic level that did not need rebuilding a year later. Compare within a discipline and never across, because the weighted totals are not commensurable and a paid media agency will always look more measurable than a brand studio. Run separate scorecards, decide separately, and sequence the two searches so you are not onboarding both in the same month.
Q: What if the best-scoring agency is the most expensive?
Commercial terms are 15% of the scorecard on purpose. Do the arithmetic before you flinch: a fee difference of a few thousand a month is small against the contribution a better diagnosis produces, and it is tiny against a year spent on the wrong constraint. Where price should decide is when two agencies score within half a point of each other, which happens more often than people expect. If the expensive one wins on diagnosis and named team, negotiate the shape of the deal, a shorter pilot or a lower fee against a performance component, rather than shopping down to the cheaper score.
Choosing between agencies, or wondering if you need one.
I am not an agency, so I have no reason to sell you a retainer. I help operators decide what to do next and who should do it, including which agencies to hire and how to hold them to the terms. If you are mid-search and the responses all look the same, email hello@taylorsicard.com.
Start a conversation Run the search in four weeks →