The short version
- A creative sprint is a fixed block of days with one goal: a stated number of finished, checked ads, each tied to a question the round should answer.
- Five days suits most small teams: questions, hooks, making, review, launch. With AI generation the same shape fits into two days.
- Size the sprint from how many ads your testing can judge, then split it between new concepts and iterations of winners.
- Someone who didn't make an ad reviews it before launch, every time.
- A sprint ends with a one-page readout, so the next sprint starts from what you learned rather than from a blank page.
What a creative sprint is
A creative sprint is a fixed block of time, usually five working days, in which a small team takes one brief all the way to a set of finished, checked ads that are ready to launch. Every ad in the sprint answers a question chosen on day one, so the round ends with something to learn, not just a folder of files. Run back to back, sprints give you a steady supply of new ads without a scramble every time a winner slows down.
Unlike "making some ads this week", a sprint has a start, an end, a target number of ads and a decision every day, so nothing drifts because nobody signed off on a headline.
Where the word comes from
Ad teams borrowed the idea from software and product design. The Scrum Guide (opens in a new tab) (November 2020 edition, checked October 2026) describes sprints as fixed-length events of one month or less, used to create consistency. GV's design sprint (opens in a new tab) squeezes a product question into five days: map the problem, sketch, decide, prototype and test with customers. GV's page (checked October 2026) says Jake Knapp started running these sprints at Google in 2010 and brought them to GV in 2012.
A creative sprint for ads keeps three things from those methods and changes one:
- It keeps the time box. Five days, or two when production is quick.
- It keeps one goal. A stated number of launch-ready ads that test stated questions.
- It keeps daily decisions. Each day ends with something settled.
- It changes the test. A design sprint ends with customer interviews. An ad sprint ends with a live test on Meta, where people answer with clicks and purchases.
When a sprint beats the weekly routine
Plenty of teams make ads on a steady weekly loop, which our guide to an ad creative workflow for small teams lays out stage by stage. A sprint is a different tool. It's a burst with one deadline and one set of questions, and it earns its place at the moments when the weekly rhythm is too slow or too scattered.
| Sprint | When to run it | Typical day-one questions | Mix of ads |
|---|---|---|---|
| Launch sprint | A new product, two or three weeks before it goes on sale, so the ads are reviewed and ready on day one | Which reason to buy should lead? Packshot or product in use? | All new concepts |
| Rescue sprint | The ad that carries your sales is selling less than it did | Is the idea tired, or only this execution of it? | Half fresh takes on the old winner, half new concepts |
| Season sprint | A sale date you can't move, such as Diwali or Black Friday | Gift angle or treat-yourself angle? Offer in the first line or the last? | Mostly new concepts, built around the offer |
| New-market sprint | Selling in a new country or language for the first time | Does copy written in the local language beat English? Do local settings beat your usual ones? | Your proven winners, rebuilt for the new market |
| Restart sprint | Testing has stopped for a month or more | What still works from the old winners? | One or two old winners as controls, the rest new |
Outside those moments, a weekly loop with a modest, steady number of new ads is usually the calmer choice. A sprint every single week wears out a small team, and the readouts start to blur into each other.
What a finished sprint hands over
Write down what "done" means before the sprint starts. If an ad isn't done by this standard on day five, it rolls into the next sprint rather than launching half-checked.
| Output | Done means |
|---|---|
| The questions | Two or three written test questions, each with the ads that answer it |
| Ad files | Every ad in each size you plan to run, usually 4:5 for feeds and 9:16 for Stories and Reels |
| Copy | Primary text, headline and description for every ad, matched to its image |
| A hypothesis per ad | One line: "We think this wins because…" |
| Names and tags | Each ad named by concept and version, with UTM tags so sales trace back to it |
| Checks | Every ad has passed the review gate further down this page |
| A readout page | An empty template, ready for results, so the learning gets written down |
The hypothesis line is the part teams skip and later miss. Without it, a winner tells you that something worked but not what to make next.
Sizing the sprint
The number of ads in a sprint should come from how many your testing can give a verdict to, not from how many you can make. Making twenty ads when your spend can only judge six leaves fourteen that get a little spend each and teach you nothing. Our guide to how many ad creatives to test and the creative test calculator turn your own numbers into a weekly count.
Then split that count between new ideas and follow-ups:
- No winner yet: almost all creative concepts, three or four distinct ideas with two or three executions each.
- One or two winners running: roughly half iterations of the winners and half new concepts.
- A winner slowing down: mostly new concepts, plus one or two fresh takes on the winner. The creative refresh schedule covers the timing.
Production capacity is the other limit. If your team makes static ads by hand, without AI generation, these are the starting points we suggest for a five-day sprint:
| Team | Static ads a sprint | Videos a sprint | The usual bottleneck |
|---|---|---|---|
| Solo founder | 4 to 6 | 0 to 1 | Time away from running the business |
| Two people | 8 to 12 | 1 to 2 | Review, because nobody has fresh eyes |
| Four people | 15 to 25 | 3 to 5 | Agreeing on the questions on day one |
These are our planning ranges, not industry benchmarks. Track your own output for three sprints and replace them with your real numbers.
The five days
Day 1: choose the questions
Start from the last sprint's readout, not from a blank page. What won, what lost, what stayed unclear? From that, pick two or three questions this round will answer. Good questions are specific: "Does a price-per-wear angle beat a comfort angle for the three-pack?" Weak ones are wishes: "Can we find a better ad?"
Write the brief before the day ends (the creative brief template has the fields), including the proof points you may use. A claim you can't back up on day four costs a day.
Day 2: hooks and angles
Write far more opening lines than you need, then cut. Pull from customer reviews, support emails, the hook bank and the patterns rivals keep paying for (BulkAds' competitor ad research sorts their ads by how long they have run). Pick an ad angle per concept and two or three hooks per angle.
By evening each planned ad has a row in a sheet: concept, angle, hook, layout, size and the question it answers. A row that answers no day-one question gets dropped.
Day 3: make
This is the only day spent mostly on files. Work from the sheet, not from inspiration. Reuse layouts from the ad templates so the idea changes between ads while the craft stays steady. Make the 4:5 version first, then adapt it to 9:16 and keep text inside the safe zone (the safe zone checker shows where it is).
Day 4: review and fix
Someone who didn't make the ad reviews it against the gate below. Fix what fails and cut anything that can't be fixed today. Expect to lose one or two ads here. That's the gate doing its job.
Day 5: launch-ready, then launch
Finish the copy, name and tag every ad, and load the round into your ad account paused for one last look. Then leave time for Meta. Its Advertising Standards page (opens in a new tab) in the Transparency Center says review can look at an ad's images, video, text and targeting as well as its landing page, and is typically completed within 24 hours, though some reviews take longer (as of October 2026). If you need ads delivering on Monday, submit them on Friday.
Who does what
Roles matter more than headcount. One person can hold several, but every role needs a name next to it.
| Role | Owns | Team of one | Team of four |
|---|---|---|---|
| Decider | The questions and the final cuts | The founder | Founder or marketing lead |
| Strategist | Brief, hooks, hypotheses | The founder | A marketer |
| Maker | Files and copy | The founder, or AI generation | A designer and a copywriter |
| Reviewer | The review gate | A friend or freelancer, for one hour | Anyone who didn't make the ad |
| Launcher | Names, tags, upload, readout | The founder | The marketer |
The reviewer is the role solo founders drop first and should drop last. You can't see the typo in a headline you wrote yesterday.
A worked example: a sprint for a sock brand
Tarnwick (a fictional brand) sells merino wool socks: a single pair for $8.25 and a three-pack for $21. Two people run it. Their last readout showed one winner, a flat lay of the three-pack with the line "Three pairs. Thirty days. Zero holes." Ads about comfort had lost twice.
Their day-one questions:
- Does an outdoor angle (long walks, dry feet) beat the durability angle that's winning now?
- Does a phone-camera, UGC-style shot beat the studio flat lay?
- Does price per wear persuade better than the plain three-pack price? The three-pack works out at about $6.85 a pair, or about $0.05 a wear if a pair lasts 100 wears.
Their sheet for the round:
| Ad | Concept | Execution | Hook | Answers |
|---|---|---|---|---|
| 1 | Trail day | Socks on a rock beside muddy boots | "Twelve miles. Dry feet at the top." | Q1 |
| 2 | Trail day | A hand pulling socks from a backpack, UGC-style | "What I pack first for a long walk" | Q1, Q2 |
| 3 | Trail day | Bold line on forest green | "Wool that walks as far as you do" | Q1 |
| 4 | Durability | The winner rebuilt as a phone shot on an unmade bed | "Three pairs. Thirty days. Zero holes." | Q2 |
| 5 | Durability | Phone close-up of a heel after a month of wear, with a notes-app caption | "Things I've stopped rebuying every month: socks" | Q2 |
| 6 | Price per wear | Receipt-style graphic, three-pack | "About $0.05 a wear" | Q3 |
| 7 | Price per wear | The same graphic with the single-pair price | "$8.25 for a pair that lasts" | Q3 |
| 8 | Price per wear | UGC-style phone shot of a tidy sock drawer holding one three-pack | "The last socks I bought this year" | Q2, Q3 |
What the notes on that sheet said:
- Ads 4 and 5 keep the winning durability idea and change only the look, so they answer the UGC question without testing a new angle at the same time. The current winner runs beside them as the control.
- Ad 8 is the one ad that answers two questions. It was kept because its job is to show whether price per wear still works when the price isn't printed on the image.
- Ads 6 and 7 differ only in the price shown, so whichever wins says something clean about how to present price.
- "Zero holes", "a pair that lasts" and the 100 wears all need proof. On day one the decider confirmed Tarnwick has a written durability guarantee to point to. Without it, those ads would have been cut at the gate.
- Two videos of ads 1 and 4 moved to the next sprint, to be made only if those stills win. A video of a losing still is money spent twice.
Day four cut ad 3, which said nothing the other trail ads didn't. Seven ads loaded paused on Friday and went live on Monday once review had cleared.
The review gate
Run every ad through the same checks before it can launch:
| Check | Why it matters |
|---|---|
| The product matches the real product | An ad showing a different colour or label tests a product you don't sell |
| Every word is spelled right, in every size | Crops and resizes for 9:16 are where typos hide |
| Each claim is backed by a proof point you can show | Unsupported claims get ads rejected and customers annoyed |
| One idea per ad | Two messages in one frame make the result impossible to read |
| Text sits inside the safe zone in 9:16 | Stories and Reels overlays cover the edges of the frame |
| The landing page shows the same product, price and offer | Meta's review can include the landing page, and a mismatch loses the sale even when the ad is approved |
| The ad answers a day-one question | Otherwise it's a file, not a test |
| Name and tags follow your system | See the naming conventions and UTM parameters guides |
The readout
The sprint isn't over when the ads launch. It's over when the readout is written, usually a week later, once each ad has a verdict. Keep it to one page:
- The questions and their answers. "Q2: the UGC-style shots beat the studio flat lay in two of three pairs." Write "unclear" when it's unclear.
- A verdict for every ad. Won, lost, unclear, or barely spent. An ad that barely spent counts as untested, not beaten.
- What we think caused it. Check each winner against its hypothesis line.
- Where the next sprint starts. Which winners get iterations, which angles retire, which questions carry over.
To show the shape, here is a made-up result for Tarnwick's round. The trail ads all trailed the durability control, so Q1 gets a no and the outdoor angle retires for now. Ad 4, the phone-shot rebuild of the winner, outsold the studio original while ad 5 barely spent, so Q2 reads "phone look wins once, ad 5 untested". Ad 6 beat ad 7, so Q3 says price per wear beats the plain pair price.
That readout writes the next day one almost by itself: put the price-per-wear line on ad 4's phone shot, make the video of ad 4 (and drop the video of ad 1, because its still lost), run ad 5 again so it gets a real verdict, and open one new concept. Our guide to iterating on a winning ad covers what to make from the winners a readout names.
Faster sprints with AI generation
When making files stops being the slow part, the sprint shrinks. With BulkAds the same shape fits into two days:
- Day 1, morning: questions and brief. The brand kit already holds your products, colours, fonts, voice and the proof points you're allowed to use, so the brief only needs the questions and the sheet.
- Day 1, afternoon: generate. In the AI ad generator, run one batch per concept, up to 12 ads each, from your real product photo. Choose the "from your instructions" route and describe the concept's rows in a sentence or two, for example "trail day, socks beside muddy boots, hooks about dry feet on long walks". Each batch is made in one size (1:1, 4:5 or 9:16), so a round that runs in feeds and Stories needs a batch for each size.
- Day 2, morning: review. Every ad arrives with a pre-flight report that scores product match, on-image spelling, brand fit and visual quality, and flags likely policy risks. Edit the copy by hand or rewrite it with a direction such as "shorter", and re-render an image whose product drifted. A person still runs the gate.
- Day 2, afternoon: launch. The bulk ad launcher creates the campaign, one ad set and one ad per creative in your own account, all paused, and tags each ad's link with its creative ID.
The cost is known before you start: an image ad uses 2 credits and a video 20, so Tarnwick's seven stills plus its two later videos would use 54 credits. See pricing for what each plan includes.
Two honest limits. AI speeds up making, not deciding, and a sprint without good questions is still a pile of ads. And BulkAds' video ads add gentle motion to a product scene. For a creator talking to camera, book the creator a sprint ahead and run that video from Meta Ads Manager, with BulkAds making the stills and AI video variations around it.
If you'd rather not run the test by hand, Autopilot can take approved ads into its own tests. It calls a winner once an ad in one of those tests has three or more sales at or under your target cost per purchase after 48 hours, and drafts four fresh variants of each winner, ready to become iterations in your next sprint.
Sprint mistakes to avoid
- Starting on day three. Teams that skip the questions make ads they can't learn from.
- Moving the deadline. A sprint that slips once slips every week. Cut ads instead.
- Changing the questions midweek. Save the new idea for the next sprint.
- Mixing tests in with your steady winners. Keep new ads apart from what already works, as the 3-2-2 method guide explains.
- Skipping the readout. Without it, sprint five quietly re-tests sprint two.
Questions people ask
What is a creative sprint in advertising?
A fixed block of time, usually five working days, in which a team takes one brief to a set of finished, reviewed ads ready to launch. Each ad is tied to a question the round should answer, so the sprint ends with something learned as well as something launched.
How long should a creative sprint be?
Five working days suits most small teams that make ads by hand. When AI generation does the file work, two days is often enough. Much longer than two weeks and it stops behaving like a sprint, because the deadline no longer forces choices.
How many ads should one sprint produce?
As many as your testing can give a verdict to, and no more. Work that number out from your spend first, then check it against what your team can make and review in the time. Extra ads that barely spend teach you nothing.
Who needs to be in a creative sprint?
Five roles: a decider who owns the questions, a strategist who writes the brief and hooks, a maker, a reviewer who didn't make the ads, and a launcher who names, tags and uploads them. One person can hold several roles, except the reviewer, who should always be someone other than the maker.
How is a creative sprint different from a design sprint?
Both use a fixed time box and daily decisions. A design sprint, as GV describes it, ends with a prototype tested in customer interviews. A creative sprint ends with finished ads launched to a live test, so the answer comes from real clicks and purchases.
What happens after the sprint ends?
The ads run until each has a verdict, then someone writes a one-page readout: the questions and answers, a verdict per ad, the likely causes and where the next sprint starts. That readout is the first input to the next sprint's day one.
Keep going
- Features
AI ad generator
On-brand image ads in batches of up to 12
- Features
Bulk ad launcher
A whole batch live on Meta in one go
- Free tools
Creative test calculator
How many new ads you can test, and how long until every one has a verdict
- Guides
Creative brief template for ads, with three filled-in examples
A ten-field creative brief template for image and video ads, three filled-in examples for a static, a creator video and an AI batch, policy notes and a send checklist.
- Guides
How many ads should you make per week? A pace your team can keep
The two limits that decide how many new ads to make each week, what counts as one ad, starting numbers by team size and a worked production week for a small brand.
- Glossary
Creative velocity
Creative velocity is the rate at which a brand makes and launches genuinely new ads, usually counted as new creatives per week. It measures the production side of testing: how quickly fresh ideas reach the ad account to replace the ones that have stopped working.
Written by the BulkAds team. Published , last checked . Meta changes its tools often — if something here no longer matches Ads Manager, tell us at contact@bulkadsai.com.
BulkAds AI is an independent product and is not affiliated with, endorsed or sponsored by Meta Platforms, Inc. Facebook and Instagram are trademarks of Meta Platforms, Inc.