Most Meta advertisers don't have a testing problem. They have a what to test problem.
They've built a campaign structure that works. They have a creative cadence in place. They know how to read the data and optimise to the winners. But the moment they introduce something genuinely new into the business, a new price point, a new product, a new offer, they don't have a clean way to find out whether the new thing actually works before committing real budget behind it.
That's what a smoke test is for.
A smoke test is an A/B test built to answer one specific question before a full campaign rolls out behind it. It isn't a long-term structure. It isn't part of your normal creative testing cadence. It's a short, narrow diagnostic, one variable, two versions, run long enough to produce a clear answer, then shut down.
Used well, it's the difference between scaling into a viable offer and scaling into a wall. Used badly, or skipped entirely, it's how brands end up running £20,000 behind a campaign that was never going to work because the price was wrong, the offer was off, or the product wasn't the right one to lead with.
This post covers when to use a smoke test, how to set one up cleanly, how to read the results, and the mistakes that turn the test into a waste of budget.
What A Smoke Test Actually Is
A smoke test answers a single, structurally important question: will the market accept this?
It's not "which hook converts better." It's not "which creative format outperforms." Those are creative tests, and they belong inside your normal testing cadence, the two-week test, two-week analyse, two-week create rhythm covered elsewhere on this blog.
A smoke test is upstream of all of that. It tests something more fundamental, the variable that, if wrong, makes the rest of the testing meaningless.
There are three classic smoke test scenarios:
Price point. Will the market pay £200 for this? Or £150? Or £80? You don't know until you put it in front of them.
Product or service. If your business has multiple offerings, which one should you actually lead with on paid? A lead-gen business with five services rarely advertises all five, one will produce cheaper leads, higher-quality leads, or higher-converting leads than the others. The same logic applies to ecom brands with multiple hero product candidates.
Offer. Which packaging of the same product wins? Free gift versus discount. Bundle versus single-unit pricing. Subscription versus one-time. Different structures appeal to different audiences and produce different unit economics.
If you're uncertain about any of these three variables, a smoke test gives you data instead of a guess.
When To Run One
Smoke tests are most obviously useful for new businesses and startups, but the technique isn't limited to first-timers.
For new businesses, the answer is almost always yes. You don't have a proven price point, a proven hero product, or a proven offer. Anything you assume about these is a guess, and building a full Meta strategy on top of guesses is how startups burn through their first £10,000 of ad spend with nothing to show for it. The smoke test produces real data on real variables before you commit to the long-term setup.
For established advertisers, smoke tests come into their own at specific moments:
- Launching a new product. You have a proven brand and a proven price tier, but the new SKU is an unknown. The market may or may not respond to it the way the existing line does.
- Changing the price. You're considering raising your hero product from £49 to £79. The volume and conversion impact at the new price is unknown. A smoke test gives you a directional answer before you commit.
- Testing a new offer structure. You've been running "20% off your first order" forever and you're considering moving to a bundle-based intro offer. The two will produce different AOVs, different conversion rates, and different unit economics. A smoke test tells you which is better for the business, not just which is better for top-line conversions.
- Entering a new vertical. You've expanded into a new product category and you're not sure which of three candidate products to lead with.
The common thread: anywhere in the business where you've introduced a genuine unknown that paid spend depends on, the smoke test is the cheapest way to get a real answer.
What A Smoke Test Is Not
Before the setup, a few things smoke tests are not useful for:
- Creative testing. Different hooks, formats, or angles, that's what your normal two-week creative cadence is for. Smoke tests aren't built to compare creative variations.
- Audience testing. Targeting audiences isn't really a meaningful variable inside Advantage+ Shopping or modern broad-targeting setups. Don't smoke test audiences.
- Bid strategy testing. Bid caps versus cost caps versus highest volume, these are operational decisions, not market-validation questions.
- Anything where you already have data. If you already know your price point works and your offer converts, you don't need a smoke test. Run your normal cadence.
The point of a smoke test is to answer a question you can't answer any other way. If the question can be answered with existing data, common sense, or a back-of-envelope calculation, the smoke test is overkill.
How To Set One Up
The structure of a smoke test is deliberately simple. Most of the discipline is in keeping it that way.
Step 1: Pick the variable. Decide what you're testing. One variable per test. Not two. Not three. If you try to test price and offer in the same test, you can't isolate the effect of either one, and you end up with an inconclusive result you can't act on. If you have multiple variables to test, run them as separate sequential smoke tests, not as one combined test.
Step 2: Build two versions. Build two campaigns, identical in every respect except the variable you're testing.
If you're testing price: same product, same creative, same landing page structure, same offer, only the price differs.
If you're testing product: same creative format, same brand voice, same offer structure, only the product being marketed changes.
If you're testing offer: same product, same creative, same landing page, only the offer construction (bundle vs discount, free gift vs percentage off) differs.
The creative itself can be near-identical between the two versions, with the only changes being the parts that have to change to reflect the variable. If you're testing price, the only edit to the creative is the price callout. If you're testing offer, the only edit is the offer mention. Everything else stays constant.
This is harder than it sounds. The temptation is always to "improve" the second version slightly, better hook, cleaner thumbnail, sharper headline. Resist it. Every improvement contaminates the test. The point is to isolate the variable, not to build the better ad.
Step 3: Use Meta's built-in A/B test tool. Meta's A/B test feature in Ads Manager is the right tool for the job. It allows you to set up two campaigns side by side, run them in parallel, and use Meta's own split methodology to ensure they're being compared fairly on the same audience.
Setup notes:
- Keep the campaign objective the same as your normal campaigns (Sales for ecom, Leads for lead gen). The smoke test should mimic your real-world conditions, not a test environment.
- Run a single consolidated ad set per campaign. No need for complex structure here. Broad targeting, default placements. The simpler the structure, the cleaner the test.
- Set the test duration. 14 or 28 days, depending on your budget. The lower the budget, the longer the test needs to run to produce a meaningful sample. Lower budgets and shorter tests produce inconclusive results that you can't act on.
- Set the budget at a level that produces enough conversions to be statistically meaningful. A useful guide: aim for at least 50 conversions per variant by the end of the test. If your CPA is £40, that's £2,000 of spend per variant, or £4,000 across the test. If you can't budget that, the test will likely be inconclusive, and an inconclusive test is worse than no test at all, because it gives you false confidence in a non-answer.
Step 4: Let it run. This is the part most advertisers fail at. Once the test is live, you do not touch it. You do not adjust budgets mid-flight. You do not turn one variant off because it "looks like it's losing" on day three. You do not change creative. You do not optimise.
The smoke test is a fixed-duration experiment. Interfering with it corrupts the result and forces you to start over. The discipline is to set it up, let it run for the full window, and judge it only on the complete data at the end.
The one exception is the same as in your normal testing cadence: if a variant is genuinely catastrophic, spending real budget at 3x or worse your breakeven CPA, turn it off. Otherwise, leave it alone.
How To Read The Results
This is where smoke tests differ most from creative tests, and where most advertisers misread the data.
A creative test gets judged on cost per result. The cheapest CPA wins.
A smoke test gets judged on business outcomes, not just CPA.
Here's why that matters. Imagine you're smoke-testing two prices: £200 and £150 for the same product. At the end of the test, Meta's data might say the £150 variant won, lower CPA, more conversions, better ROAS. From inside the Ads Manager view, it's the obvious winner.
But step outside the platform for a moment. The £200 variant produced fewer sales but at higher margin per sale. The £150 variant produced more sales but at lower margin per sale. Which one made more profit? Which one produces better LTV (since price often correlates with customer quality)? Which one positions the brand better for the next year?
The answers to those questions don't show up in the Ads Manager column you sorted by. They show up in your business P&L.
So the rule for reading a smoke test: the winner is whichever variant produces the best outcome for the business, not whichever produces the best in-platform metrics.
For most price tests, that means calculating gross profit per variant, not just looking at CPA. (Total conversions × per-unit gross margin = gross profit. Subtract ad spend. Compare.)
For product tests, that means looking at downstream behaviour, does one product produce more repeat purchases, higher LTV, better customer reviews? The cheapest CPA isn't the only metric that matters.
For offer tests, that means modelling out the unit economics of each offer at scale, including AOV impacts, refund rates, and subscription take-up rates.
The smoke test gives you the raw data. The decision is yours, based on what the data means for the business, not just what Meta says converted.
What To Do With The Winner
Once the smoke test is over and you have a clear answer, the winner becomes the input to your long-term setup.
You don't keep the smoke test campaigns running. They're not built for scale. They were a diagnostic, and the diagnostic is done.
You take the winning variable, the price, the product, the offer, and you build it into your normal campaign structure. Advantage+ Shopping or your equivalent consolidated setup, your normal creative testing cadence, your usual scaling approach.
The smoke test answered the upstream question. Now everything downstream, research, creative production, testing cadence, scaling, has a stable foundation to run on top of.
Common Mistakes That Waste The Test
A few patterns that turn smoke tests into wasted spend:
Testing too many variables. Trying to test price and offer simultaneously means you can't isolate either effect. Run them as separate sequential tests.
Budget too low. Tests that don't reach statistical significance just produce inconclusive results. If you can't budget enough spend to generate ~50 conversions per variant, the test isn't worth running yet.
Test duration too short. Two weeks is the minimum. Four weeks is better for lower-volume accounts. Three-day tests are useless, same diagnostic logic as the normal creative cadence.
Touching the test mid-flight. Adjusting budgets, pausing variants early, or "tweaking" creative during the test invalidates the results.
Judging on the wrong metric. Reading the smoke test off CPA alone, rather than on business outcomes, leads to wrong decisions. The winner in-platform isn't always the winner in the P&L.
Skipping it when there's a real unknown. The most expensive mistake of all. If you genuinely don't know whether a price point, product, or offer will work, the £2,000–£4,000 cost of a smoke test is dramatically cheaper than the £20,000–£50,000 cost of scaling behind something the market wasn't going to accept.
Why The Discipline Is Worth Building
Smoke tests aren't glamorous. They're not the part of media buying that gets case-studied or screenshotted on LinkedIn. They produce small, narrow answers to specific questions, not headline results.
But they're one of the highest-leverage habits a serious advertiser can build.
The brands that scale efficiently are the ones that don't put real money behind unknowns. They test first. They get the answer. Then they build the engine.
The brands that struggle are the ones that skip this layer entirely, that commit to a price, a product, or an offer based on assumption, then spend three months wondering why the account isn't working. By the time the data is clear, they've already lost the budget they could have spent on the right thing.
A smoke test costs you a few thousand pounds and two to four weeks. Skipping a smoke test on a variable that turns out to be wrong costs you exponentially more. The decision is structural.
