Somewhere right now a business is testing whether a green button beats an orange one.
In a few weeks it will have a result that barely matters, and it will run its Black Friday campaign with the same untested offer as last year.
Button colours and tiny headline tweaks are popular tests for one reason, which is that they are easy. They are also the least likely to change your results in a way you will ever notice.
The test worth your money is the offer: what people get, on what terms, by when. Run it before Black Friday, while there is still time to act on the answer. Pick one offer variable, a guarantee, a bonus, a deadline or a bundle. Run two versions head to head for at least a week with everything else identical. Then build your whole Black Friday campaign around whichever one clearly wins.
Why test the offer and skip the colour
A colour change asks the same person to make the same decision against a slightly different background. An offer change asks them to make a different decision.
So offer tests tend to produce big, readable differences, while cosmetic tests produce small ones that vanish into the noise. On a small budget, big differences are the only kind you can detect.
Timing matters too. Constant Contact surveyed more than 1,600 small business decision makers across Australia, Canada, the UK and the US in 2024, and 57% said they increase the frequency of their marketing in the last three months of the year. The middle of that rush is a terrible time to discover your offer was the weak one. As we argued in Q4 is won in the weeks before Black Friday, the work that decides the peak happens before it.
Test what the customer is deciding on. The decoration on the page can wait.
Four offer variables worth testing
The guarantee. Same product, same price. One version mentions no guarantee. The other puts a clear, specific promise near the top: "If it does not fit, we swap it free." "If we are late, you do not pay the call out fee." Taking risk off the buyer is one of the strongest levers you have, and it costs nothing unless you fail.
The bonus. Put a percentage discount up against a bonus of similar perceived value, such as a free add on or an extended warranty. Bonuses protect your price and often feel more generous than a small discount.
The deadline. Run an open ended offer against one with a real, specific end date. Or pit a date deadline against a quantity limit, like the last 40 units or the last six booking slots. Only test deadlines you will honour.
The bundle. Try a single product against a starter kit, or against a good, better, best choice. Bundles can lift order value, and they make it harder for a shopper to line your price up against a competitor's.
To choose which one goes first, start with the biggest objection your buyers raise. If they worry it will not work, test the guarantee. Price shoppers call for the bonus or the bundle. And when you keep hearing "I will think about it", test the deadline.
Pick one. Change two variables at once and you will never know which caused the difference. For more raw material on building offers worth testing, the Godfather offer breaks down how to stack them.
A worked example: the skincare brand that almost discounted
Say you run a small Australian skincare brand selling online, with a hero moisturiser at $58. Last year's Black Friday offer was 25% off sitewide. It sold, but it wiped out most of the margin and trained customers to wait for the sale.
This year, in late October, you test two versions of a single ad set for the moisturiser. Version A is 20% off. Version B is full price, plus a free travel size cleanser and a "love it or return it within 30 days" guarantee. Same video, same audience, same landing page layout, same daily budget, split with the platform's experiment tool.
After ten days, B has produced roughly as many purchases as A. On sales alone it is a tie. On value it is a blowout. The bonus costs you a few dollars in product, while the discount costs you $11.60 on every order. Refund requests under the guarantee are negligible.
So your Black Friday campaign leads with the bonus and the guarantee, and the percentage off stays in the drawer. These numbers are invented, but the pattern is common enough: a tie on volume can hide a clear winner on margin, and you only see it if you test before the rush.
How to run it on a small budget
You do not need a data science team. You need discipline.
- Keep everything else identical. Same creative, same audience, same landing page layout, same budget. Only the offer changes.
- Use the platform's experiment tool. Meta's A/B testing tool, for one, divides your audience into random, non-overlapping groups, so nobody sees both versions and your two offers are not muddying each other's results.
- Put the difference up front. The offer has to show in the first line of the ad and at the top of the landing page, or nobody will notice there is anything to react to.
- Measure a real action. Purchases if you have the volume. Otherwise, use the closest action to a sale that you get enough of, such as add to cart or booking requests.
- Set the rules before you start. Fix the end date and what counts as a winner before the first dollar goes out, so you are not tempted to call it early.
- Write down your prediction. Note which version you expect to win and why. It keeps you honest when the result surprises you.
With no ad budget to split, run the test through email. Send version A to one random half of your list and version B to the other, with the same subject line, and compare purchases. It is rougher, and it still beats guessing.
How to read the result honestly
Small budgets trip people up at this step.
Do not call it after a day. Early results swing wildly. Run for at least a full week so both versions see weekdays and a weekend.
Treat a small gap as a tie. With modest numbers, a slight difference between two versions is very often luck. You are looking for an obvious gap that holds as the numbers grow.
Judge value over volume. A discount might win on sales and lose on margin. A guarantee might bring fewer leads and better ones. Look at what each version is worth after costs.
A tie is still an answer. If a bonus performs the same as a discount, choose the bonus. Your price stays intact.
Confirm it once. If there is time before the peak, run the winner against a fresh challenger. A result that wins twice is one you can lean on.
Mistakes that ruin an offer test
- Testing two things at once. A new offer and a new video in the same variant tells you nothing about either.
- Burying the offer. If the bonus only appears on the checkout page, most people never see it, and the test measures nothing.
- Fake scarcity. A deadline that quietly extends, or "limited stock" that never runs out, damages trust. It is also a legal risk. The ACCC's guidance is blunt: any statement that creates a false impression about goods and services can be breaking the law.
- Stopping when you like the answer. Decide the end date up front and stick to it, even if one version looks ahead on day three.
- Ignoring fulfilment. A guarantee or bonus your team cannot deliver at Black Friday volume turns a winning test into a losing season.
Your timeline from now to the peak
Black Friday falls on 27 November this year. Working back from that date, the schedule is tight but doable.
- This week. Pick the variable, based on your biggest customer objection. Write both versions of the offer and check you can fulfil either.
- By the end of October. Build the two variants and launch the split test.
- First week of November. Let it run. Do not touch it.
- Second week of November. Read the result on value as well as volume. If there is time, confirm the winner against one fresh challenger.
- Mid November. Brief the creative and rebuild the landing page around the winning offer. Warm up your audience so the offer lands on people who already know you. Our guide to building your audience before Q4 costs climb covers that part.
The Black Friday decision
Once you have a winner, build the whole campaign around it. Every ad and every email leads with the offer that won, even if a different one sounded cleverer in a meeting.
Carry it through every touchpoint, including the ones after the click. If the ad promises a free travel size cleanser, the product page and the cart should say so in the same words. An offer that disappears between the ad and the checkout feels like a bait and switch, even when it was an honest slip.
Keep the loser on file. An offer that lost in October might still be the right follow up for people who did not buy the first time, or the right challenger for next season.
Get the test live before the end of October. That leaves time to read it and brief the creative before the busiest fortnight of the year.
Pick one variable tonight and write two versions of your offer. Everything else about your Black Friday campaign can wait until you know which one your customers choose. If you want a second pair of hands designing the test and building the campaign around the winner, that is exactly the kind of work our advertising team does before every peak.
Frequently asked questions
What should I A/B test before Black Friday?
Test the offer and leave cosmetic details like button colours alone. The four most useful variables are the guarantee, a bonus versus a discount, the deadline, and bundles. Choose the one that answers your buyers' biggest objection, change only that, and keep creative, audience, landing page and budget identical so any difference comes from the offer.
How long should an offer test run?
At least a full week, so both versions see weekdays and a weekend. Longer is better on small budgets, because early results swing wildly. Set the duration and your definition of a winner before launch, then stick to it. If there is time before the peak, confirm the winner once more against a fresh challenger.
Is a free bonus better than a discount?
It is worth testing, because a bonus protects your price and can feel more generous than a modest discount. If both versions sell about the same, the bonus usually wins on margin, since a free add on costs you its product cost while a discount comes off every order. Run the test before you assume either way.
Can I test offers without a big ad budget?
Yes. Use the ad platform's built in experiment tool with whatever budget you have, and measure the closest action to a sale you get enough of, such as add to cart or booking requests. If you have no ad budget, split your email list randomly in half and send each half a different offer.