Most Marketers Ignore This Half of Their Data

Every marketing measurement result has two parts: the metric you care about and how sure you are about it. Most teams only look at the first one, and it can be a massive mistake. Let's say you run a media mix model or a geo experiment and the result comes back: TikTok prospecting has a 5.2x ROI. That single number is what statisticians call a point estimate. It's the midpoint of the analysis. It's the best guess given the data you have. Most people stop right there. They put the number in a deck, they celebrate the result, or they worry about it. Or they just make a budget decision and move on. But good marketing measurement methods don't just give you a single number. They give you a range of estimates that reflect measurement uncertainty, the full range of estimates that are compatible with the data you observed. That range is your uncertainty interval. It says things like, "The true CPA for this channel is probably somewhere between $35 and $72." The point estimate is usually in the middle of that range. The interval is how uncertain you should feel about it. The challenge here is that ranges are uncomfortable for many people and they feel less precise. A 5.2x ROI sounds clean and confident. TikTok has an ROI somewhere between 2.8x and 8.1x. Sounds like your team doesn't know what they're doing. So most teams just drop the interval. They only report the point estimate. It's easy to do this even though it can be a critical mistake for your business. A tight uncertainty interval, say 4.9x to 5.5x, means your estimate is precise and you can really plan around that number. But a wide interval, say 0.8x to 9.6x, it means something very different. If you invest in the channel and the true performance is at the bottom of the range, then you lose money. Performance is at the top of the range, you'll be extremely profitable. And the data just can't tell you which one is closer to reality. Both of those analyses could have the exact same point estimate, 5.2x ROI, but one is a confident bet and the other is practically a coin flip, or worse. The uncertainty interval is what tells you the difference. So when someone hands you a measurement result, don't just ask, "What's the number?" Ask, "What's the range?" If it's a massive interval or one that spans negative to positive, well, that's hardly an answer. It's closer to a guess with a decimal point. And you shouldn't be moving your hard-won marketing budget just based on a guess.