Scoring a launch against the prediction you wrote first

One launch, invented for this page, scored against the prediction its team wrote six weeks earlier. Every number here is made up. What matters is the method: three lines written before anything was built, and the two moments where they stopped the team from concluding something false.

What to take away

The venture

Invented case. Two people, a service that reconciles invoices for small construction contractors. Delivered manually. Priced monthly. No product yet beyond a spreadsheet and a phone. What they did not have was anyone who had paid.

Week minus six: what the prediction contained

Write the prediction before anything is built, and date it. It needs a number for each step of the launch and a line that would count as a negative result. Here the steps ran from 200 people reached to 2 paying.

The negative-result line states in advance which count would mean the test failed, so a result near it cannot be talked into a win afterwards. Without it, two sales could be called a good start or a poor one, depending on the mood of the week.

They also wrote the sentence they expected to be tempted by: "people liked it, we just need more traffic." Writing that in advance is a small thing, and it did most of the work later.

Week minus four: two conditions that moved the date

The truth list had eleven conditions, and nine were met early. Two were not, and the list found both before any customer could. One was a registration wait. The other was a claim nobody could source.

The payment path needed a business bank account, which needed completed registration. Requirements vary by state and structure. The SBA describes registering your business name with state and local governments as what most small businesses need, and some need nothing. Here it took nine days.

A launch planning guide turns a wait like that into a dated line instead of a surprise. A truth list of this kind is the "what must be true" calendar described in the launch planning framework.

The landing page claimed the service "cuts reconciliation time in half." Nobody could source it. It had two friendly trials behind it and no measurement. Advertisers are expected to have a reasonable basis for advertising claims before they run, so the team cut it and described what the service does.

The date moved by a week. This is the first place the plan earned its keep: both problems were found by a checklist rather than by a customer.

Week one: scoring the result line by line

Score the result against the prediction before anyone offers an explanation. Two paying customers landed exactly on the negative-result line, which is the worst possible outcome: every reading is available and none is forced.

Week two: the argument, settled by channel

The temptation sentence arrived on schedule. One of them said it almost verbatim: people liked it, we need more traffic. Because it had been written six weeks earlier, they had to deal with it rather than agree with it.

They checked where the nine starts came from. Seven came from direct outreach to people the founders already knew by name. The newsletter, the only channel that could scale, produced 140 of the reach and two of the starts.

That reframes the result. The launch tested no scalable channel. It tested whether people who already trusted them would pay, and the narrow answer was a qualified yes.

Warm and cold response differ, which is why market validation says the two columns must never be averaged together.

What the result allowed them to decide

Not to quit, and not to build software. They treated the newsletter as an unproven channel, not a failed one: 140 people seeing a headline is not a test of anything.

They also rewrote the offer sheet to describe the first two weeks after payment in plain steps.

Then they wrote a new prediction for the second attempt, splitting warm and cold reach into separate lines. That change came straight from the week two argument, and a plan can only make that kind of correction if it was written before the result.

What was still unknown at the end

Quite a lot, and naming it was part of the debrief.

Whether the service works at ten customers instead of two, since the delivery was manual and the founders were the delivery. Whether the two customers stay past the first month, which no launch activity can answer because it needs time to pass. And whether the price is right, since two acceptances at one number is not a shape.

Staying past the first month is the territory of an invented first hundred days. The pillar on launch planning sets out the sequencing that lets one launch produce two separate readings.

Common questions

Is this a real company?

No. The case, the counts, and the quotes are invented for this page. Published launch numbers from real ventures come from a different offer at a different price to a different audience, which is why borrowing them leads people wrong.

Would more traffic have settled it?

More traffic through the newsletter would have tested the newsletter, which was the open question. More traffic through the founders' own contacts would have produced a bigger version of a result they already had.

Was the negative-result line set too low?

Possibly, and setting it at two is what made the boundary case happen. A cleaner line would have named the channel as well as the count: two paying customers from people we have never met. That version could not have been argued either way.