138 studies, 19,951 people, one finding: being asked gets goals done.

There’s one number underneath my whole product, and it isn’t mine.

In 2016, Harkin and colleagues published a meta-analysis in Psychological Bulletin. They pooled 138 studies covering 19,951 people and asked a plain question. Does monitoring progress toward a goal help people actually reach it? It does. Reliably. Across every type of goal they examined. And the effect grew stronger when progress was reported to someone else.

Read that last clause again. Reported to someone else. Another person, expecting an answer. That clause is the product.

A meta-analysis is the boring end of science. No single dramatic result, just 138 experiments stacked on top of each other until the noise cancels and the signal stands. That’s why I trust it more than any one flashy study. One study can be a fluke. 138 pointing the same way is a fact about people.

Here’s the problem coaches live with. Coaching works in the session. Then the client goes home and the week happens. Argent, Daly and Caulfield looked at home exercise programs in 2018 and found only around 35 percent of patients stick to the plan between appointments. In rehab, non-adherence runs as high as half. The client left the session with a plan and plenty of intent. What they don’t have is anyone who will ask about it before next week. A coach can’t ring every client every day. There aren’t enough hours in coaching.

There’s a third strand. Wood and colleagues showed in 2016 that simply being asked about a behaviour moves the behaviour. In one part of that work, self-reported exercise rose from 14 percent to 26 percent over two months. The question itself is an intervention. Asking is doing.

I see it in myself. I keep every promise I make to other people and about half the ones I make to myself. My own promises matter just as much. Nobody checks on them, that’s all. When someone is going to ask me on Friday whether I did the thing, the thing gets done. When nobody asks, Friday quietly stops existing.

So that’s ShoUp. An AI agent rings a coach’s clients on the schedule the coach sets, in the coach’s own voice, and asks how it went. It chases when nobody answers. The coach sees everything. The research says being asked gets goals done, so we built the asking. The rest is engineering.

Now the story I actually want to tell you. The statistic we killed.

Early on, our notes carried a beautiful number. Telephone check-ins lift adherence to 86 percent. For a company that makes accountability phone calls, that statistic is nearly perfect. Which was the tell. Before letting it anywhere near the website, we went hunting for the primary source. Six separate search angles. Databases, citation trails, the papers cited by the papers. There is no primary source. Nobody ever ran that study. The number is a rumour wearing a percentage, passed from slide deck to slide deck until it sounded like science.

We retired it. It sits on a banned list in our marketing notes now, with an instruction beside it: never reintroduce. A companion claim went out with it, the one about the phone call being the single most effective adherence intervention. Same disease. Nobody can say who measured it.

It has plenty of company. The accountability statistic this industry loves most, the famous one about your odds of success once you commit to another person, traces to nothing as well. We looked. No study exists. It survives by recursive citation, everyone citing someone who cited someone who cited no one.

Why be this fussy? Because my product makes exactly one claim: being asked works. The whole company is a bet on a finding. If a coach checks my evidence page and one number crumbles when she pushes on it, she’ll discount every number on the page, including the true ones. She’d be right to. A padded evidence page is a confession that you don’t think the honest version is enough.

So the rule at ShoUp is short. Every claim carries a named source, or it doesn’t ship. Retiring the 86 percent cost us a lovely headline. It bought us a page where everything survives checking. Evidence discipline is a product feature here, the same way uptime is. A coach stakes her reputation on us every time she puts a client on the platform. The least we can do is stake ours on numbers that exist.

And the real finding never needed the fake ones propping it up. That’s what I keep coming back to with Harkin’s paper. Across 138 studies and nearly twenty thousand people, the thing that moved the needle was the oldest mechanism we have. You told someone what you’d do. They’re going to ask.

The phone rings. Someone asks. The thing gets done.