Does our ICP need this?

Find out if your ICP needs a feature before you build it

Describe the problem the feature solves as an offer, send it to the people who'd use it, and let the replies tell you whether to build.

Start your first experiment Read the playbook

An example hypothesis

Operations leads at 50–200 person logistics companies reply at 10% or more to “stop re-keying orders from email”.

How it's usually answered, and why that misleads

Ask current users

Existing customers ask for everything, and they are a biased sample of the market you want to grow into.

Run a survey

Saying yes to “would you use this?” costs nothing. Replying to an email that names your problem costs attention.

Buy ads for a landing page

Clicks tell you about the ad and its audience, which is rarely the people who would use the feature.

How it works in Sellvance

1

Frame the problem, not the feature

“Stop re-keying orders from email”, not “AI order import”. People reply to problems they recognise.
2

Email the people who feel it

The role that does the painful work, at companies where the work exists.
3

Set the bar for building

The reply rate that would justify the engineering time, decided before you see a single reply.
4

Read the verdict, then decide

A no in two weeks is cheaper than a feature nobody uses in two quarters.

AI qualification that keeps only the companies and roles that fit

What you get

✓ A pass, a fail or a grey zone on whether the problem gets a response
✓ Replies describing the problem in the prospects' own words, useful for the spec
✓ A list of people who replied, who are the first to show the feature to

Before anything is built

Only offer what you could honestly deliver or discuss. A reply means someone wants to talk about the problem, so be ready to have that conversation, and say plainly that the feature is not built yet.

When to stop, decided before you send

Outcome	When	What to do
Works	Delivery passes and replies are conclusively above your target.	Scale it: start a separate campaign with a bigger daily limit and more mailboxes.
Doesn't work	Delivery passes and replies are conclusively below your target.	Change one thing and run a new experiment on fresh companies.
Grey zone	The planned sample is sent and the verdict is still open.	Extend once, by at most half the sample. Still open? Treat it as a no.
Invalid	Delivery fails, so replies can't be read.	Fix the channel and restart. Nothing was learned about the market.