Will this product launch succeed? AI-driven prediction

A product launch is the business decision people are most overconfident about. The team has lived with the product for months, so its value feels obvious, and the launch plan reads like a foregone conclusion. Then it ships into a market that has never heard of it and doesn't care, and the gap between expectation and reality is brutal — and discovered far too late to adjust.

Predicting the launch up front closes that gap. You describe the product, the market, and the plan, and you get a realistic prediction of the first 90 days: the likely signup and conversion range, the branches, and — most usefully — the early signal that tells you which branch you're in while you can still act. Here's how to predict a launch honestly.

Why launches get over-predicted by their own teams

Internal optimism is a systematic bias, not a personal failing. You can't un-know how good your product is, so you can't simulate the indifference of a stranger seeing it for the first time. A prediction can, because it isn't emotionally invested. It anchors on base rates for launches in your category rather than on your team's enthusiasm — which is exactly why its number will feel pessimistic and exactly why it's more useful.

The honest framing isn't "will it succeed?" but "what's the distribution of first-90-day outcomes, given a market that starts at zero attention?"

The assumptions a launch prediction makes

For a new SaaS tool, MiroFish typically assumes a cold-start audience unless you specify an existing one, a category-typical visitor-to-signup rate, a signup-to-paid conversion in the single-digit-to-low-teens percent range, and a differentiation discount if you're entering a crowded space without a sharp wedge. It also assumes your distribution channel — and this is where launches quietly differ by 10x.

If you're launching to an existing audience of 50,000 engaged followers, that's a completely different prediction from launching to a cold market with a press release and hope. State your real distribution, or the model will assume the pessimistic default.

How the first 90 days branch

A launch prediction usually resolves into:

• Slow burn (~50%): Modest early numbers — tens of signups, single-digit-to-low-teens conversion — that compound only if one channel finds traction. Unspectacular, survivable, and by far the most common reality.
• Wedge catches (~25%): A sharp differentiator lands in a specific niche, word spreads inside that niche, and growth becomes self-reinforcing. This branch almost always traces back to positioning, not features.
• Stalls (~25%): Signups trickle, activation is weak, and the launch never reaches escape velocity. Usually a positioning failure — the product was real but the market couldn't tell why it should care.

The two failure-adjacent branches together carry meaningful weight, which is the sobering, useful part. A launch is not a coin you should flip assuming heads.

What actually decides it

The deciding factor is positioning sharpness, not feature count. Teams instinctively believe a launch succeeds because the product is good; the prediction repeatedly shows it succeeds because a specific person can instantly tell why this is for them. A crowded category doesn't kill you — fuzzy positioning in a crowded category kills you. The wedge branch and the stall branch are usually the same product with different clarity about who it's for.

So the prediction points at a pre-launch action that has nothing to do with shipping more: get the one-sentence "this is for X who needs Y" so sharp that a stranger repeats it back correctly. That's the variable that moves you between branches.

The tripwire to watch

Don't wait 90 days to learn which branch you're in. The early signal is week-two activation rate — of the people who signed up, how many reached the product's core value? Above ~35% activation tracks the slow-burn-to-wedge branches; below it, you're heading for the stall, and the fix is almost never "more traffic" — it's onboarding and positioning. Watching that number lets you intervene in week three instead of grieving in month three.

Launches share their core lesson with pricing changes and hiring decisions: the obvious variable (features, price, résumé) is rarely the one the outcome turns on. Predict it, find the real swing variable, and act on that before you ship.