“Buy X, get Y free” promotions reliably increase sales conversions.

However, you’ll drive even more conversions by framing a customer’s target product as the free gift, instead of the other way around.

For example, if a customer searches for “fitness tracker” on your website, show them the offer “Buy a weighing scale and get a fitness tracker free,” rather than “Buy a fitness tracker and get a weighing scale free.”

Why it works:

• People don’t expect to see a product they actually want as the free gift.
• This makes them feel lucky—the promotion seems more attractive.
• And the novelty of the offer makes them more likely to buy.

Steps to implement:

1. Choose a target product (the free gift). This can be something your customers commonly search for or a known best-seller.
2. Choose a secondary item (the main product) of equal or similar value. If one product is significantly more expensive than the other (e.g., buy a printer, get a free laptop), people will either question the quality or think it’s a scam and refuse the offer.
3. Enhance the effect with messaging that makes customers feel lucky (e.g., “It’s your lucky day!”)

When you’re deciding which TikTok creators to work with, we suggest narrowing your search to those who make quality content in a similar industry to yours—instead of basing your decision on the size of their following.

Why focus on lesser-known influencers? They give brands the most bang for their buck.

Influencers with large followings generally charge more—and engagement rates are often lower than those of smaller creators. But more crucially, TikTok users see nano- and micro-influencers as “people like me,” making viewers more likely to trust them and take action based on their recommendations.

With smaller influencers, you aren't riding on anyone's celebrity, only the quality of their performance.

Another benefit of lesser-known influencers: You can commission a higher volume of creatives to test in your campaigns. Since many of them are either early in their careers or striving to work as creators full time, they’ll work harder to deliver a result you’re happy with.

While other brands try to land deals with TikTok celebrities, focus on finding expert craftspeople who know how to engage an audience.

For more on TikTok, we’ve got a fresh, definitive playbook for you: How to Acquire Customers with TikTok Ads. In it, we cover how to make A+ ad creatives, source content creators, structure your ad account, and launch your first campaign.

We spent months interviewing top TikTok ads experts to make sure it’s the most credible, in-depth resource on TikTok acquisition. Dive in here.

There are a few “rules” to running experiments. Scientists should follow them, but marketers can break them sometimes.

Rule 1: Your hypothesis should test one discrete variable.

It should have one cause and one effect.

• Example of a one-variable hypothesis: If we build tailored landing pages for our audience segments, our unsubscribe rate will decrease.
• Example of a multi-variable hypothesis: If we build tailored landing pages for our audience segments and add personalization tags to our emails, our unsubscribe rate will decrease.

When to break it: Testing one variable, then another, then another isn’t always feasible in a fast-paced startup environment. You might need to break this rule to have fast impact instead of exact insights.

Do you want precise learnings? Follow the rule. Is it more important for you to move quickly and gauge cumulative impact? Break it.

Rule 2: Don’t peek at A/B test results early.

Early in an experiment, the likelihood of a false positive is high. If you peek early and see the result you want, you might be tempted to call the experiment too soon.

When to break it: Looking at your test results early gives you a chance to catch anything that’s critically broken. If you don’t, you could end up running a test for weeks, only to discover a bug that not only invalidated the test but even hurt company performance.

Plus, it’s extremely difficult to guess what your goal for a test’s outcome should be. For example, if your goal is a 5% conversion change but your test ends up producing a 15% effect, you’ll waste time if you let it run without looking at it.

So, although it’s not good science, we recommend breaking this rule.

First, do a spot-check one or two days after launching a test. Look for any critical issues, bugs, etc.

Then, if you’re running a longer test, look at it again at the halfway mark. But with one important rule in mind: If you’ve set a confidence level of 90%, then when you do your midway peek, only call the test if there’s a statistically significant result with a 98% confidence level or higher.

There’s still a chance that the effect you’ll see at the halfway peek is a false positive. But we’ve found that this method is more practical than the “don’t look at a test at all until it’s reached its target sample size” method that no one actually follows, and it provides some added protection against false positives.