The First Days Decide

A churn analysis on one B2B company’s daily utilization data asked a plain question: how soon can you tell whether a customer is going to make it? The answer came back at ten days.

Customers who hit a minimum usage threshold inside their first 10 days churned 4% over the first year. Customers who got there later churned 40%. Four years on, 87% of the fast starters were still customers, against 6% of the slow ones.

Stretching the window did nothing. Moving the threshold from 10 days to 30, 60, or 90 made less than 1% difference. The customers who didn’t start fast did not start late. They mostly never started at all. Ten days isn’t a deadline you set. It’s the length of the customer’s willingness.

That’s the reason speed matters here and not much elsewhere. A customer’s motivation to change how they work fades on its own, and other problems arrive to take the attention. You get one short stretch where the account is willing to work differently, and you either spend it or lose it.

Customers who started inside 10 days churned 4% in their first year. Customers who started later churned 40%. Four years on, 87% of the fast starters were still there — and 6% of the slow ones. Giving them 30, 60, or 90 days instead of 10 changed almost nothing.

I wrote down why that window carries so much weight nine years ago, before I had the data to prove it. Early churn is the signal that the threshold of behavior change required to get results is too high — the product is asking too much, too fast. And small changes in retention early produce disproportionately large overall impacts, because there are increasing returns to improvements in initial retention. Churn is infant mortality: most of it happens early, and the causes of it happen earlier still.

Your technology does not produce results. Behavior change produces results, and technology only makes that change possible and scalable. Your software will not amount to anything unless customers change how they work. It’s that simple. Business results are process change multiplied by technology, so zero change in the business process means zero results no matter how good the technology is.

So look at what your onboarding actually delivers. If it is mostly training — a tour of the product, a configuration checklist, a certification — it is transferring product knowledge, and product knowledge is not the thing that produces the result. Adoption is not success. When customer success is mostly training, that’s a warning sign. Whether a customer succeeds depends primarily on the degree to which they change how they work.

Your technology does not produce results. Behavior change produces results. So onboarding that teaches the customer your product and stops there has skipped the only step that ever produces a result.

Customers change what they do under four conditions, and all four have to hold: they can change, they’re motivated to change, they know what to change, and they know how to change. Most failures are the fourth one. Customers imagine what they’ll have to do only vaguely, and that’s the essence of the problem — many don’t realize they have to change anything at all. A product tour doesn’t touch any of that.

The practical move is smaller than you’d think.

Stop treating the start as an implementation. There’s no such thing as implementation. There’s just individual results, and some results build on others. Once you see it that way, the work is to find the earliest result that can stand on its own, strip everything that isn’t a prerequisite for it, and get the customer there. Only a few behaviors matter for results, and the ones that matter drive first results. Most customers will never work out which ones those are on their own.

Aim lower than you think you need to. Customers fail earlier and lower than you think, they value much less improvement than you assume, and they aren’t failing to use your tool — they’re failing to do their work. They don’t doubt your tool, they doubt themselves. Most churn comes from customers who achieved no results at all, which makes helping those customers get their first small result the largest opportunity in the business.

Then take work off the customer entirely where you can. Bake the best practices into the default installation so the product ships configured and ready to produce results — every configuration step you remove is a behavior change the customer no longer has to complete before value. And change what onboarding is measured on. Its job is behavior change and materialized results, not training and adoption, which makes time to first results the number it answers for.

Move the checkpoint up while you’re at it. Instead of judging an account at three months, measure at 10 days: if there’s no attributed result, put the account in the risk group and identify the failed step.

Support tickets during onboarding look like trouble. They’re the opposite. In a client study, customers who submitted tickets during onboarding renewed at double the rate of customers who didn’t, and the effect showed up immediately after they started, so it isn’t long-tenured customers accumulating problems. A ticket means the customer is trying to use the thing. Trying is what produces results, and results are what keeps them. The customers most likely to leave are the ones who never ran into a problem, because they were never engaged enough to hit one.

It goes further than that. In an early segmentation I ran, the customers who hit a software bug and got a good, proactive service experience were the most loyal group in the base at 80% — more loyal than the customers who never hit a bug at all. What made that service good was not the fix. It was engaging, understanding, listening, learning, and building solutions.

Customers who submitted support tickets during onboarding renewed at double the rate of customers who didn’t. The riskiest customer in your base is not the one with problems. It’s the quiet one.

You can check all of this against your own numbers this week. Take your last eight quarters of new customers, define the first real result on your product in one sentence, and mark the date each customer reached it. Split the cohort at 10 days and plot the two survival curves. Then look at how many customers in the slow group never reached the result at all, and how many of them never opened a ticket either.

Whatever those two curves do, they were drawn in the first two weeks. The renewal conversation a year later is a formality on a decision your onboarding already made.