Paid pilot or free test with an agency: which one really decides?

In an agency-client relationship, a paid pilot is a short, invoiced engagement of two to four weeks, designed to produce a measurable decision, whereas a free test delivers only an isolated sample. The pilot includes a volume of creatives, a media budget and written success criteria; the free test includes none of the three. Rule of thumb: accept a free test to judge a style, demand a paid pilot to judge performance, and get the pilot credited against the first month. This article is about the format of the first engagement, not about choosing the agency: the shortlist is assumed to be done. According to the SRI's 35th Observatoire de l'e-pub, the French digital advertising market reached €12.4 billion in 2025 (+11%), growth driven in particular by Social and Video.

Paid pilot, free test, speculative pitch: what exactly do these three terms mean?

Three different things, often confused in the same prospecting email, which commit neither the same resources nor the same level of proof. Before negotiating anything, you need to know which of the three you are discussing, because the price shown says nothing about the content.

• The paid pilot: a deliberately reduced scope, invoiced, time-limited, with success criteria written before the start. A batch of creatives is produced, run, measured, and a verdict is delivered. What you are buying is not the video, it is the decision.
• The free test: a free deliverable, usually a video, sometimes an edit of rushes you already own. No media, no iteration, no commitment on turnaround. It is a sales argument disguised as proof.
• The speculative pitch (spec work): unpaid idea production to win a budget. Strategic recommendation, concepts, sometimes storyboards and mock-ups. It is the large-account version of the free test, and it has been documented in the advertising industry for fifteen years. According to the EACA's Cost of Pitching 2025 study, relayed by Media Marketing, only 4 in 10 winning creative ideas are actually implemented by clients.

The defining difference comes down to one sentence: the paid pilot is instrumented, the free test is not. Instrumented means that, before you begin, you know which metric will decide, over what volume, over what period and with what delivery budget. A deliverable that is not instrumented can be gorgeous and prove nothing.

We are regularly asked about a fourth category, the “free audit”. That one is honest, provided it remains an audit: a read of your ad account, an analysis of your existing creatives, a written recommendation. It does not produce a video, so it does not claim to prove production capability. That is its limit and also its virtue.

Not to be confused with an airline pilot

The word “pilot” also refers to the civil aviation profession and to ENAC training. If that is what you were looking for, the FAQ at the bottom of the page answers in a few lines. Here, “pilot” means a commercial test project between an advertiser and an agency.

Why can't a free test prove that an agency is good?

Because it lacks the four conditions without which no creative measurement is possible: volume, media, iteration and duration. This is not a question of the agency's good faith. It is an experimental design problem: a free video cannot answer the question “will these people bring my acquisition cost down?”

Volume. A single creative is not a sample. On our French e-commerce accounts in 2025, the observed ratio is around two to three concepts genuinely usable in media out of every ten shot; the range varies with product category and brief quality, and this is an internal observation, not a published statistic. In other words, a single video is statistically more likely to prove nothing than to prove something. If it flops, you do not know whether it is the agency, the angle or the product. If it works, you do not know whether it is repeatable. The minimum volume threshold is the same one that tips a budget from a freelancer to an agency structure: the 8-video threshold also applies to sizing a pilot.

Media. A video that is not run generates no audience data. It generates opinions. We have seen management committees decide a free test on an argument like “the music sounds cheap”, while the same creative, pushed with an €800 budget, delivered the best 3-second retention rate in the account. Internal opinion and media data do not say the same thing, and it is always the latter that pays the bills.

Iteration. What distinguishes a video agency from a shoot provider is the second round: the ability to rework a hook that does not hold, to re-run an angle that shows a signal, to produce six variants of an opening in 48 hours. A free test stops at the first deliverable, by definition. So you are testing the one skill that predicts nothing: making a nice standalone video.

Duration. Seven days is not enough to get an ad set out of its learning phase, let alone produce a verdict. The time needed depends directly on the number of creatives you can send out each week: time to results is a function of creative throughput, not of the calendar.

A free test measures an aesthetic. A paid pilot measures a production system. They are two different purchases, and only the second translates into media budget committed behind it.

A caveat, and it works against our commercial interest: the free test has a legitimate use case. When a brand has never shot in vertical and needs to convince a committee or a sceptical founder, a visual deliverable unlocks an internal decision that ten slides will not. In that specific case, ask for it and own it for what it is: an internal sales tool, not a performance test.

How much does a “free” test really cost, and who pays for it?

A free test costs the agency between 1.5 and 2.5 person-days, and that cost is always recovered somewhere: in the seniority assigned, in recycled assets, in your first month or in the supplier's financial health. There is no such thing as free in a business where the product is the work itself.

The breakdown, as we saw it when we still did them: half a day of scoping and writing, half a day of creator casting or product preparation, half a day to a full day of shooting, half a day of editing, then the quarter-day of back-and-forth that nobody ever counts. Add the creator's fee, product costs, transport. An “offered” test cost us between €800 and €2,000 in full cost depending on the format, an internal 2025 range, not a market price.

There are four recovery mechanisms, and no fifth one:

• Assigning a junior. Unbilled work never goes to the best editor in the company. So you judge an agency on its lowest level, then sign hoping for its highest.
• Recycling assets. Rushes from another client, editing templates already amortised, a reused voice-over. The video is clean because it has already been produced, for someone else.
• Folding it back into month 1. The most common, and the quietest. The test is free, but the first month is often billed above the rate card, a premium that, from what we observe, we estimate at a few tens of percent. You have paid, with a delay.
• Adverse selection. The agencies that systematically offer free work are the ones with spare capacity. Sometimes because they are starting out, which is honourable. Sometimes because they are losing clients, which is less so.

The advertising industry has quantified this problem at its own scale. According to the study The Cost of the Pitch by the 4A's, relayed by LBBonline, a pitch costs the participating agency around $204,461 on average, and a consultation on the advertiser side approaches $408,500 on average when no incumbent agency is involved. The 2025 State of the Pitch Report indicates that 66% of pitches required speculative work and that 77% did not use an online tendering system, in a climate marked by longer timelines, unpaid work and frustrated agencies. In 2025 the EACA published a Cost of Pitching study whose conclusions point the same way at European level.

These figures concern seven-figure budgets, not your free video. The mechanism, however, is identical at every scale: unpaid work that is billed to someone, later, elsewhere. Transposed to an e-commerce SME, it comes out as a video offered by a junior, a slightly inflated month 1, and a relationship that starts on a misunderstanding.

The comparison with SaaS often comes up in these discussions. It does not hold. In software, the free trial works because the product already exists: the marginal cost of one more user is close to zero. In SaaS, free-trial conversion rates vary widely depending on how the trial is designed, with no single benchmark that everyone agrees on: the gaps between free trials can be large depending on their design, with no published figure serving as a reference. In video production, nothing exists before it is produced. The marginal cost is the full cost. That is precisely why B2B pricing analyses place consulting, agency work and complex implementations outside the freemium model, and recommend a short paid pilot, typically two to four weeks.

What does a decidable paid pilot look like in vertical video?

A pilot is decidable when its six parameters are written down before the first shoot: number of angles, number of variations per angle, duration, media floor, a single decision metric, reporting format. If even one of the six is missing, you are buying videos, not a decision.

1. The number of angles

Three to four distinct angles, never just one. An angle is a different promise addressed to a different problem: time saved, comparison with a competing product, social proof, price per litre. Testing a single angle amounts to betting your agency decision on a creative strategist's hunch on a Tuesday morning.

2. Variations per angle

Two to three different hooks per angle, same video body. This is the cheapest and most informative part of the pilot: it tells you whether the agency knows how to isolate a variable. A team that delivers ten videos all different on ten dimensions will never be able to explain to you why one won.

3. Duration

Three weeks minimum, four for a wide catalogue. Below that, delivery does not have time to stabilise. Beyond six weeks, it is no longer a pilot, it is an open-ended engagement that you are paying for at the price of a test.

4. The media floor

This is the parameter that brands systematically want to trim, and it is the one that decides whether the pilot produces data or conversation. You need enough budget for each variant to exit the learning phase, otherwise you are comparing noise. The media floor rule applies to the pilot exactly as it does to cruising speed: no floor, no verdict. For its Search Ads campaigns, the TikTok for Business help centre (updated April 2026) recommends aiming for more than 20 conversions in the first week to get through the learning phase.

5. A single decision metric

Just one, chosen in advance, written into the quote. Not a twelve-column dashboard that will be interpreted in a meeting depending on the mood. For most merchants, it is the cost per winning creative, not the number of videos delivered. For a brand at launch with no history, it will rather be the 3-second retention rate and the cost per outbound click, because there will not be enough conversions to decide otherwise. Say so beforehand, not afterwards.

6. The reporting format

A written document, not a video call. Which angles performed, which hooks were killed, which hypotheses are still open, what the agency would do in month 2 with budget X. This deliverable is often worth more than the videos: it remains your property even if you stop there, and it serves as a brief for the next agency. It is the hardest test for a supplier, because it forces them to write down in black and white what did not work.

How much should a pilot cost, and how far is it reasonable to pay?

A pilot should never exceed the equivalent of one month of the target contract, media excluded. Beyond that, the agency is no longer selling you a test, it is selling you a short engagement while letting you believe it is a trial period. This cap rule settles 80% of negotiations.

Three formats cover nearly every situation. The ranges below are our 2026 orders of magnitude for the French vertical video market, production only, excluding media budget. They shift depending on product category, the number of SKUs to film and whether paid creators are used.

The discovery pilot is the only one worth weighing against a free test: same order of magnitude of work, but you pay to get a written scope and a senior on the project. The decision pilot is the one we recommend in 7 cases out of 10. The pre-scale pilot is only justified if you arrive with a five-figure monthly media budget and the real question is “will this team keep up the throughput”, not “do they know how to film”.

A point of caution on media: it should not pass through the agency during a pilot, except in special cases. You pay for delivery from your own account, you keep the data, you keep the pixel. An agency that insists on carrying the media budget for a three-week test is in reality asking you to entrust it with the very measuring instrument that is meant to judge it.

Which clauses make a paid pilot acceptable, and which make it toxic?

Six clauses make a pilot healthy, three make it toxic, and none costs a penny more to an agency that plays fair. These are clauses specific to the first engagement: the general terms of a long-running production contract are another subject.

The six clauses to demand

• Pilot credited against the first month. All or part of the pilot amount is deducted from the first invoice if you continue. 50% is a reasonable standard, 100% can be negotiated on discovery pilots. An agency that refuses any credit is telling you it considers the pilot a sale, not a step.
• Ownership of rushes and usage rights even without a follow-up. You paid for the production, so you leave with the source files and the agreed usage rights, whether or not you sign afterwards. Creator rights are a separate line, with explicit duration and territory.
• Read access to the ad account for the agency. Without it, it optimises blind and can always say the verdict was biased.
• No tacit renewal. The pilot stops on its end date. What follows is signed affirmatively, in a new document.
• No exclusivity during the pilot. You keep the right to carry on producing in-house or with a freelancer. Exclusivity during a test period is a lock-in clause in disguise.
• Written debrief of learnings. Delivered no later than seven days after the end of delivery, whether or not there is a follow-up.

The three clauses that should make you back away

• A paid pilot tied to a 6- or 12-month commitment signed in advance. That is not a pilot, it is a deposit. The test loses all meaning if it has no exit door.
• A pilot invoiced without a usable deliverable: concepts, a moodboard, a recommendation. You are paying for a pitch that others give away.
• Transfer of rights conditional on signing the contract. An elegant variant of hostage-taking: your videos exist, they are paid for, but you cannot run them if you leave.

These six clauses are set at the quote stage, not when the annual contract is signed. They complement the list of questions to ask a video agency before signing, several of which are asked even before the pilot.

Paid pilot or free test with an agency: how to decide in ten minutes?

The answer depends on a single variable: do you have, yes or no, a media budget available to run during the test. If the answer is no, the paid pilot is pointless, you would be paying for an impossible measurement. If the answer is yes, the free test becomes a waste of time.

The tree comes down to four questions. No media budget you can mobilise in the next three weeks: accept a free test if one is offered, otherwise postpone. Media budget available and the first video agency of your history: decision pilot, three to four weeks. Media budget available and replacing an incumbent agency: decision pilot too, but with an imposed benchmark, your three best current creatives keep running against it. Wide catalogue and approaching seasonality: pre-scale pilot, because the real question is throughput, not unit quality.

What we cannot measure, and we may as well say so: the comparative follow-through rate after a free test and after a paid pilot. We stopped offering free tests at the end of 2024, and the earlier sample is too small and too heterogeneous to draw an honest rule from. What we do see, however, is the length of relationships: accounts that came in through a properly scoped pilot more often go past the sixth month, because the success criteria had been written by both parties before starting.

One last case observed, and it comes up often. A food supplements merchant consulted us in spring 2025, ended up going for a free test at a competitor, received a decent video, never ran it for lack of an identified media budget, then came back in October with exactly the same question as in April. The free test cost him nothing in euros. It cost him five months of the season and a Q4 prepared in a rush. It is the only bill you do not see on a quote.

So ask the question differently. Rather than “will you agree to do a free test”, ask: “what is your paid pilot format, how many angles, what media floor, what output metric, and how much of it is credited against my first month”. An agency that answers in three minutes with figures has already proved more than a free video ever will. The one that dodges has answered you too.

Frequently asked questions

Does a serious agency agree to make a free video?

Some do, especially at launch stage or when they have spare capacity. It is not a bad sign in itself, but the deliverable will be produced by a junior and without media delivery. Accept it to judge a style or unlock an internal decision, never to decide on a team's performance.

Is the paid pilot deducted from the contract if I continue?

It should be, at least partly. The standard we practise: 50% of the pilot amount credited against the first monthly invoice, up to 100% on a discovery pilot. Have the clause written into the pilot quote, not into the following contract, otherwise it disappears at the final negotiation.

Who owns the videos if I do not continue after the pilot?

You do, if you paid for the production and the quote says so. Demand three things: delivery of the source rushes, advertising usage rights for the agreed duration and territory, and the written debrief of learnings. Transfer of rights conditional on signing a follow-up contract is a clause to refuse.

How long should a pilot with a video agency last?

Two to four weeks in the vast majority of cases, six at most on a wide catalogue. Below two weeks, delivery does not leave its learning phase and the verdict is a matter of opinion. Beyond six weeks, it is no longer a test but an engagement billed at the rate of a test.

What is the minimum media budget for a pilot to make sense?

Enough for each variant to accumulate usable impressions over the duration of the test. The calculation works backwards: number of variants multiplied by the volume needed per variant, over three weeks. Without that floor, the pilot produces opinions and not data, and you might as well keep the money for the following month.

Were you looking for “airline pilot”: ENAC, free training, salaries?

A different subject. The ENAC EPL pathway is a state-run airline pilot training programme, free, demanding and highly selective. Becoming a pilot without paying goes through this public route, through the air force, or through an airline funding the qualification. The highest salaries are at the large long-haul carriers. Details on enac.fr. This page deals with the paid pilot in the commercial sense.