How to choose a creative agency

Short answer: Give every agency on your shortlist the same brief and the same questions, then score their answers on one scorecard. The answers that matter most cover how they plan and read tests, what their proof actually shows, and what the contract says about who owns the work, the ad accounts and the data. Be wary of guaranteed results and case studies without dates or baselines.

Who this guide is for

This is for founders and marketing leads who have decided an agency is likely the right model and are about to speak to two or more of them. If you have not made that decision yet, start with Performance Creative Agency vs. In-House Team.

SGNL is an agency, and you may be evaluating us. I have written the questions and the scorecard as the standard I would want a client to hold us to, and you should apply them to SGNL in the same way as to anyone else.

The questions that matter

Ask these in writing before any pitch meeting, so every agency answers the same thing. A poor hire usually costs you the notice period in fees, several months of testing you have to redo, and, if the agency holds your ad accounts, your account history as well. The written answers are the cheapest point at which to find that out.

How they think about your business. These show whether the agency starts from your customer and your sales, or from its own format.

• What do you think our customer needs to believe before they buy, and where did you get that view?

• Which metric would you use to judge our creative, and why that one instead of views or clicks?

• What would you need from us in the first two weeks?

How they test. These show whether results will teach you something you can reuse.

• Show me a testing plan you have used. How do you choose what to test first?

• How many variables do you change in one test? (Meta recommends testing one variable at a time for conclusive results; see its A/B testing best practices.)

• How long do you run a test, and what decides when it is over?

• How do findings from one month change the next month’s briefs?

What they make and how fast. These turn the fee into a number of usable pieces of work.

• How many finished assets do you produce a month at the fee you are quoting, and what counts as one asset?

• What is your turnaround from raw footage to an edit ready for review?

• Who is on camera in your model. creators you source, our staff, or our founder?

How they measure and report. These show whether the agency can separate its effect from everything else that changed.

• Which data sources will you use, and will you need access to our sales or CRM data?

• How do you separate the effect of creative from changes in budget, targeting, offer or season?

• What does your monthly report contain? Ask for a redacted example.

Who does the work. These show whether the people in the pitch are the people on your account.

• Who exactly will work on our account, and how much of their time is ours?

• Who has final approval on your side, and how quickly do they review?

Terms and exit. These show what you keep if you part ways.

• Who owns the footage, edits and project files at the end?

• Whose business account holds our ad accounts and pages?

• What is the notice period, and what do we receive when we leave?

Proof standards: what counts as evidence

Most agencies will show you results. The question is what the result actually shows. A case study is useful evidence when it includes:

• A named time window with start and end dates.

• A baseline showing what the number was before the work started.

• The spend during the window, if paid media was involved.

• The data source (for example an ad platform export, an analytics tool or the client’s sales system), and whether the agency or the client produced the figure.

• What else changed in the same window: budget, offer, pricing, seasonality, press, other channels.

• Client permission to share it, or a clear statement that it is anonymized.

• A reference you can speak to, ideally a client of a similar size and category.

A result missing several of these is a claim. It may be true, but it does not show what caused it. Be especially careful with correlation presented as causation: sales rising in the same month an agency started does not show the agency’s creative caused the rise. Methods built to measure causation exist, such as holdout or lift tests, which compare people who saw the ads with a similar group who were deliberately not shown them, and an agency that cites one should be able to show you how it was set up.

Fabricated testimonials are a legal issue as well as a credibility one. The FTC’s Rule on the Use of Consumer Reviews and Testimonials, in effect since October 21, 2024, prohibits fake or false consumer and celebrity testimonials (FTC Q&A). If a testimonial on an agency’s site cannot be traced to a real client, treat it as unverified.

Vaunt example: how we label our own proof

We apply the same standard to SGNL’s work for Vaunt, a private aviation membership app and SGNL client. Every Vaunt figure in our case-study files is marked Confirmed (a primary platform document or export, cross-checked), Single-source or Unconfirmed, with its date and source. The daily new-user series comes from Vaunt’s own dashboard report, sent by David Dickson of Button Up Management, which we did not build.

Here is what that looks like for one claim. On July 9, 2026, a TikTok on Vaunt’s account reached 1.2 million views (TikTok views checked in our October 4, 2026 accuracy audit), and Vaunt recorded 1,978 new users that day, 2.94 times the prior 7-day average. We present that as timing and not as cause: new users were already rising on July 7 and 8, paid ads were running, and a re-post of the same video on July 14 (44,300 views) showed no lift.

Figures the client has not confirmed in writing stay out of anything we publish.

Ask every agency on your list to show its proof with the same labels. If it cannot say where a number came from and what else changed at the time, score it low on “Quality of proof”.

Contract considerations

Have a lawyer review the agreement. These clauses decide what you keep if the relationship ends, which is when an unclear contract costs the most.

Ownership of the work

Under US copyright law, work commissioned from an independent contractor is a “work made for hire” only if it fits one of nine statutory categories and both parties sign a written agreement saying so; otherwise the contractor owns the copyright until it is assigned (US Copyright Office, Circular 30). Ask for a written assignment of the final deliverables, and decide whether you also need raw footage and project files.

• Talent and usage rights. If the agency hires creators or actors, the contract should say which rights pass to you, for which platforms, territories and terms, and who pays for renewals.

• Ad accounts, pages and pixels. Keep them in your own business account and give the agency partner access. Meta supports this directly (give a partner access to your business assets).

• Data access and confidentiality. Define what data the agency can see, how it is stored, and what happens to it when you part ways.

• Scope and deliverables. Monthly volume, the definition of one asset, revision rounds, turnaround times and the rate for out-of-scope work.

• Approvals. Who approves, by when, and what happens to the schedule if an approval is late.

• Media spend. Whether ad spend is billed to you directly by the platform or passes through the agency, and whether any markup applies.

• Compliance. Who writes and checks disclosures and claims. The FTC’s Endorsement Guides say advertisers should guide and monitor their endorsers, so you cannot assign that responsibility away entirely.

• Term, notice and exit. Minimum term, notice period, and a list of what is handed over on exit: files, access, reports and any creator contracts.

Warning signs

• Guaranteed results. No agency controls the auction, the algorithm or your customers. Google makes the same point about search in its guidance on hiring an SEO: “No one can guarantee a #1 ranking on Google” (Google Search Central). The same caution applies to promised return on ad spend.

• Ad accounts in the agency’s name. If the agency wants to run your ads from its own account, you may lose your history and data when you leave.

• Case studies with no dates, no baselines or no source.

• No written testing plan, or a plan that changes many variables at once.

• Testing by switching ads on and off. Meta advises against informal testing of this kind because it can produce overlapping audiences and unreliable results (About A/B testing).

• A senior team in the pitch and a different team on the account. The people who win the work are not the people who do it.

• Vague scope. “Unlimited content” or “as needed” without a volume or a definition of an asset.

• Reports built only on views and likes when you sell a product or service.

The scorecard

Score each agency from 1 to 5 on each criterion, multiply by the weight, and divide by 5. A perfect score is 100. The weights are my suggestion; change them to match what matters most to you before you see any pitches, so the weights do not bend toward a favourite.

A downloadable version of this scorecard is being prepared.

How to run the selection

• Write a one-page brief with your goal, metric, budget range, platforms and constraints.

• Shortlist three agencies. In my judgment more than that makes comparison harder without adding much.

• Send the same brief and the questions above in writing, with a deadline.

• Score the written answers before any pitch meeting.

• Meet the people who would work on your account, not only the sales lead.

• Check proof and call references. Ask the reference what went wrong and how the agency handled it.

• Consider a paid pilot with a defined scope and end date before a long commitment.

• Negotiate the contract using the clauses above, then re-score if terms change.

Tradeoffs and limitations

• A scorecard reduces bias but does not remove it. Interviews still favour agencies that present well.

• New agencies have less proof. A young agency may score low on proof and still be a good fit; a paid pilot is a fairer test for them.

• Some good proof cannot be shared. Clients often do not allow agencies to publish their numbers; SGNL’s own Vaunt figures stay internal where they are not confirmed. Ask for a reference call or anonymized data rather than treating a lack of named cases as a failure.

• This is not legal advice. Use the contract section as a checklist for your lawyer.

• When to skip the agency route. If you have a strong creative lead in-house and only need production capacity, a freelance editor or a production company may be enough.

Next step

Fill in the weights, send the questions to your shortlist, and score the written answers before you take any meetings. If SGNL is on your list, ask us the same questions; see SGNL’s services. Before you sign anyone, make sure your team can run the creative testing plan the agency proposes.

Sources

All accessed October 9, 2026.

• Federal Trade Commission, Guides Concerning the Use of Endorsements and Testimonials in Advertising, 16 CFR Part 255 (revised 2023). https://www.ecfr.gov/current/title-16/chapter-I/subchapter-B/part-255

• Federal Trade Commission, Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465, effective October 21, 2024. https://www.ecfr.gov/current/title-16/chapter-I/subchapter-D/part-465

• Federal Trade Commission, “The Consumer Reviews and Testimonials Rule. Questions and Answers.” https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers

• US Copyright Office, Circular 30, “Works Made for Hire.” https://copyright.gov/circs/circ30.pdf

• Meta Business Help Center, “Best practices for A/B testing.” https://www.facebook.com/business/help/290009911394576

• Meta Business Help Center, “About A/B testing.” https://www.facebook.com/business/help/1738164643098669

• Meta Business Help Center, “Give a partner access to business assets in your business portfolio.” https://www.facebook.com/business/help/1717412048538897

• Google Search Central, “Do you need an SEO?” https://developers.google.com/search/docs/fundamentals/do-i-need-seo

• SGNL Media internal data for Vaunt (not public). daily new-user reports from Vaunt’s Looker/Data Studio dashboard, prepared by David Dickson of Button Up Management (January 14 to October 4, 2026), stitched into a daily series in SGNL’s case-study files (October 4, 2026).

• SGNL Media, flyvaunt content audit (logged-out capture, September 30, 2026) and content-library accuracy audit (October 4, 2026). Not public.

About the author

Andrew Moullin is co-founder and Creative Director of SGNL Media, where he leads creative and strategy for a four-person team that builds media systems for founders, experts and brands. Before marketing, he played professional soccer for San Antonio FC and Atlanta United 2, then coached.