In this note
- What a prediction is
- Why being wrong is the point
- The four parts of a good prediction
- Common ways to dodge a prediction
- What it does to a team
- Do this this week
- Questions people ask
Here's a pattern almost every marketing team will recognise. A new home page launches. A month later, enquiries are up a little. Everyone agrees the new page worked. Six months later, enquiries are down a little, and everyone agrees that's the market.
Nobody lied. But nobody learned anything either, because nobody said in advance what the page was supposed to do. Without a prediction, any result can be explained after the fact, and an explanation that fits every outcome isn't knowledge. It's a story.
That's why every change we make, for clients and for ourselves, starts with a written prediction.
What a prediction is
A prediction is a short, specific statement made before the change goes live. It says what will move, by how much, and by when. It's written down, dated and shared, so nobody can quietly revise it later.
"The new page will improve engagement" isn't a prediction. It's a hope. "Adding price ranges to the services page will reduce 'how much does it cost?' questions on first calls, and lift the Answered score for the money buyer, within one month" is a prediction. It can be checked, and it can be wrong.
If a change can't be wrong, it can't teach you anything.
Why being wrong is the point
It feels strange to welcome misses, but they're the most useful thing a prediction gives you. A hit tells you your understanding of the buyer was right. A miss tells you something you didn't know about them, and that's exactly the information you were missing.
Say we predicted that named results under the hero would lift trust, and they didn't. That's not a failure to hide. It might mean buyers don't trust the results because the clients aren't ones they recognise. It might mean the sceptic was never reading the hero at all. Either way, the next change starts from a better guess.
This is also why we publish our misses. When we run the method on ourselves in Generation One, the predictions about Evolut go up before the results land, and they stay up whether they hit or miss. A method you can't check is just an opinion.
The four parts of a good prediction
The change
Exactly what's being changed, in one sentence. One change per prediction where possible. If you change five things at once, you won't know which one moved the result.
The moment
Which buyer, at which stage, this is aimed at. "The sceptic at Believed" or "the champion at Found". Naming the moment forces clarity about who the change is for, and it tells you where to look for the effect.
The expected movement
What should change, and roughly by how much. It can be a score, a count, or a behaviour you can observe, like the kind of questions people ask on sales calls. Be honest about what you can actually measure.
The time frame
When you'll check. Most of our predictions are checked within one Generation, which is usually a month. Some effects take longer, and that's fine, as long as the date is set before you start.
Common ways to dodge a prediction
Teams that are new to this find creative ways to avoid being pinned down. Watch for these.
Vague verbs. "Improve", "boost" and "strengthen" can't be checked. Replace them with what you'd actually see.
No baseline. If you don't know where a number starts, you can't say how far it moved. Measure before the change, not after.
Moving the goalposts. When a prediction misses, it's tempting to find another metric that went up and call that the real goal. Write down the miss first. Then, by all means, note the surprise.
Predicting everything. A prediction that lists ten outcomes will always hit one of them. Pick the one that matters most.
What it does to a team
The practical benefit is learning. The cultural benefit is just as big. Predictions move conversations from opinion to evidence. Instead of the most senior person's favourite design winning, the team agrees in advance what success looks like, then lets the buyers decide.
It also makes it easier to try bolder things. When every change is framed as a test with a written expectation, a miss isn't a career risk. It's a finding. Teams that are allowed to be wrong in public tend to be more creative, because the cost of a bold idea that doesn't work is a paragraph in a report, not a reputation.
Do this this week
- Find your next change. Something already planned: a page update, a new email, a revised proposal template.
- Write the four parts. The change, the moment, the expected movement and the time frame. One sentence each.
- Record the baseline today. Whatever you said you'd measure, write down where it stands now, with the date.
- Put a date in the calendar. Book fifteen minutes on the check date to compare the result with the prediction. Write down hit or miss, and one thing you learned either way.
Questions people ask
What is a marketing prediction?
A short, written statement made before a change goes live, saying what will move, for which buyer and stage, by roughly how much and by when. It's dated and shared so it can be checked honestly afterwards.
What happens if a prediction is wrong?
You learn something you didn't know about your buyers. A miss is recorded alongside the hits, the team looks at why, and the next change starts from a better guess. Publishing misses is what makes the hits believable.
How is this different from A/B testing?
A/B testing usually compares small variations of one element. Predictions apply to any change, including the bigger things buyers decide on, such as the story, the proof and the offer. They can sit alongside A/B tests, and they also work where a split test isn't practical.