A restaurant creative test worksheet

Why write the brief before the ad exists?

The short rib photo collects the most reactions, and the team agrees it won. A month later someone asks what exactly was being tested, and nobody can say. It was never a test. It was two ads that happened to run at the same time, and anything looks like a win once you know the result.

What tends to happen is that results arrive with no decision waiting for them. The numbers are interesting, the meeting moves on, and the next shoot starts from the same assumptions as the last one. The fix is to write down, before anything is produced, what you want to learn and what you will do differently depending on the answer.

What goes in the test brief?

Nine questions, answered in this order.

1. What business outcome are we trying to improve? Pick one: introduce a new restaurant, fill an early seating, bring guests back.
2. Which location, service or occasion are we promoting? Name one of each. A test spread across everything hides where it worked.
3. What is our hypothesis? Write a sentence that can turn out wrong.
4. What is the one creative element we will change? The image, the headline or the format. Not two of them.
5. What will stay the same? The offer, the audience, the spend and the timing.
6. What is our primary success measure? Choose it now, not after the results are in. Google's guidance on ad experiments says to pick one or two measures of success before you start (Test with confidence with the Experiments page).
7. What supporting signals will we review? Engagement and bookings are useful context, but they are not the verdict.
8. What else could affect the result? A holiday, the weather, a menu change, another campaign, a full room.
9. What result would change our next creative brief? Say it in advance, so the answer can't be explained away later.

What does a filled-in brief look like?

Here is a made-up one. Say a restaurant runs two ads for two weeks, $300 each, to introduce its fall dinner menu.

• Outcome and occasion: more paid dinner visits at one location, Thursday to Sunday.
• Hypothesis: showing the room full will produce more paid dinner visits than a close-up of the plate.
• One change: the image. Ad one is the short rib close-up. Ad two is the dining room on a busy evening.
• Held constant: the same offer, the same audience, the same dinner-hours window, $300 each.
• Primary measure: paid checks traced to the ad. Supporting: bookings and reactions.
• Other factors: no holiday in the window, and one Saturday at full capacity noted.
• Result that changes the brief: if the room clearly outsells the plate, the next shoot starts with the room.

Suppose the short rib gets 2,300 reactions, 14 bookings and 8 paid checks worth $640, about 2.1x. The room gets 520 reactions, 22 bookings and 17 paid checks worth $1,445, about 4.8x. The feed would have picked the plate. The brief had already named the register as the judge.

When do you call the test?

Not after one good day. Run it long enough, and with enough activity, to tell the two versions apart. If there isn't enough to separate them, record that honestly and plan a better test with a bigger difference or a longer window.

Watch for anything that moves the result for reasons that have nothing to do with the creative. A rainy Saturday or a competing promotion can swing a small test, and a room that is already full can't show any lift at all because it had no seats left to fill.

How do you read the result?

Read it in three layers. Engagement shows attention, a booking shows intent, and a paid check shows value. Use the measure you chose in advance as the verdict and the others as context. Where a guest's path can be connected to a paid check, look at sales. Where it can't, say so and don't present a booking as a meal. We explain why in how to tell which restaurant ad actually sells meals.

What do you do with the answer?

Write two sentences: What did we observe? What will we produce or run differently next time? In the example: "For this offer and audience, the busy room outsold the plate on paid dinner visits. The next brief leads with the room at its best." Keep that note with the creative brief, so the next shoot starts from what the restaurant learned, as in our photo shoot brief.

Where restauWant comes in

The restauWant app's advertising view reads which ads carried a campaign and which only collected clicks, down to the individual creative, with estimates kept apart from the checks traced to the ad. When a campaign doesn't have enough matches yet, it says so instead of ranking on a guess.

Next step: Fill in the nine questions for your next campaign and share them with whoever is producing the creative before anything goes live.

Before and after a creative test

1. Write the business outcome and the one location or occasion before any creative is made.
2. State a hypothesis that could turn out wrong.
3. Change one thing only: the image, the headline or the format.
4. Hold the offer, audience, spend and timing the same for both versions.
5. Choose the primary measure now, and no more than one or two supporting signals.
6. List what else could affect the result: holidays, weather, capacity, menu changes, other campaigns.
7. Decide in advance what result would change the next brief.
8. Don't call the test early. If the activity is too thin to separate the versions, record that and plan a better test.
9. Judge by the measure you chose, and keep bookings and reactions as context.
10. Write the two-sentence decision and keep it with the creative brief.

How long should a creative test run?

Long enough, and with enough activity, to tell the two versions apart. There is no single number, because it depends on your spend and how busy the location is. If the activity is too thin to separate them, record that and design a bigger test.

Can I test more than one thing at once?

You can, but you won't know which change did the work. Changing one element, such as the image, keeps the answer readable. Run the next test on the next element.

Should I judge by likes or by bookings?

Neither is the verdict by itself. Likes show attention, bookings show intent, and a paid check shows value. Choose the measure that matches your business outcome before the test starts.

What if both ads perform about the same?

That is a result. It means the difference you tested doesn't matter much for this offer and audience. Test something bigger next time, such as a different subject rather than a different crop.

What if a full room hides the result?

If a location is already at capacity, extra demand has nowhere to go and the test can't show a difference. Pick a service period with room to grow, or note the full nights and set them aside.