Product Discovery

Learn from people who are not already fans.

Stand Chat helps product managers and product team leads turn willing website visitors into grounded discovery evidence - especially the evaluators, skeptics, and confused prospects who rarely book formal interviews.

Example visitor chat

Visitor

We are checking whether this could replace the custom chat we built into our docs.

Stand

That usually means checking transcript access, routing, and whether product can learn from the chats. Stand keeps labeled history and digest evidence for that. May I ask two quick product questions so the team understands your situation better?

Visitor

Sure.

Stand

When did the custom docs chat last cause a real problem for the team?

Visitor

Last month. Our PMs wanted to see why evaluators got stuck, but the logs were just raw transcripts. Nobody had time to tag them.

Stand

Thanks, that context helps. What happened because those transcripts were not tagged?

Visitor

We missed a pattern around migration questions until sales brought it up three weeks later.

New product signal

Product teams usually hear from committed customers, sales calls, and support tickets. Stand adds another source: people who are evaluating, hesitating, comparing, or getting stuck on your site right now.

The agent answers the visitor first. When the moment is right, it asks consent and switches into Product Discovery mode. In that mode it stops pitching and asks practical follow-ups about what actually happened, how the visitor handles it today, who is involved, and what is at stake.

Product teams can give each agent up to three current research questions: the Big 3 they are trying to answer this week. Stand uses them as interview priorities and records which question each evidence card informs, without turning the chat into a task tracker.

The approach is inspired by the interviewing discipline in The Mom Test, an excellent read for anyone doing product discovery.

Where to use it

Start on one page where visitor questions are product signal.

You do not need to replace your support chat. Put Stand on pages where Product Discovery makes more sense than a support queue, and keep your existing chat where it already works. Avoid running two chat widgets on the same page; that makes the visitor choose between channels before they know who can help.

Docs or developer pages

Learn what evaluators were trying to ship when the docs did not answer the real decision question.

Pricing or packaging pages

Capture what buyers compare, what feels risky, and who else is involved before they talk to sales.

Migration or comparison pages

Drill into the last migration attempt, current workaround, blockers, and what breaks if nothing changes.

Feature pages

Find the gap between what the page explains and what prospects actually need to understand.

How it works

01

Help first

The agent answers the visitor question before asking for research time.

02

Ask consent

If the visitor matches your criteria, the agent asks for permission to ask a few product questions.

03

Notify you

When Product Discovery captures its first grounded signal, Stand can notify you so you can listen in and, if useful, signal the agent to offer to invite you.

04

Aim at research questions

Add up to three active research questions, such as migration urgency, ownership, or alternatives already in use.

05

Drill into reality

Questions focus on the last time it happened, current workflow, workarounds, frequency, stakeholders, and stakes.

06

Stop in time

The agent exits discovery when the visitor seems done, frustrated, rushed, or wants direct help.

07

Report the evidence

Product Discovery chats are labeled in history and summarized in Product Discovery reports with links back to the original conversations.

Setup

Product Discovery is a skill you can turn on for your agent. It can be combined with other skills, such as offering follow-up meetings, answering from a product knowledge base, or conversation labeling that helps you get to other important conversations.

Candidate criteria

Define which visitors are worth a discovery ask: new evaluators, skeptical buyers, migration blockers, accessibility questions, docs confusion, or any segment your PM wants to learn from now.

Research goal

Give the agent the product context behind the interview so follow-ups stay connected to the decision the team is trying to make.

Research questions

Add up to three questions the team is actively investigating. Stand attaches grounded evidence to the matching question when the conversation informs it.

Interview depth

Choose a lighter or deeper interview budget. The agent still stops early when the visitor becomes terse, frustrated, rushed, or asks for direct help.

Avoid guidance

Tell the agent what not to do, such as asking roadmap-leading questions, discussing sensitive commercial topics, or bothering visitors who are clearly seeking support.

Real-time notifications

Notify the agent owner when Product Discovery captures a grounded match, so a PM can offer to join while the visitor is still in context.

Practical examples

Evaluation blocker

Research question
Are migration pains urgent enough to act on?
Trigger
Visitor asks whether your product fits a workflow with legacy systems.
Question
When did this integration problem last slow the team down?
Captured evidence
Last migration attempt, current workaround, team impact, and what happens if it remains unsolved.
Evidence quality
Strong
Digest output
Two evaluators described losing release time during legacy migrations. Both already use internal wrappers and worry about rewriting edge-case components.

Docs confusion

Research question
What alternatives are teams already using?
Trigger
Visitor asks a docs question that hints the page did not answer the real concern.
Question
What were you trying to ship when you looked for this?
Captured evidence
Current job-to-be-done, why the docs failed, what they already tried, and what decision the answer supports.
Evidence quality
Incomplete
Digest output
A visitor was not looking for an API reference. They were deciding whether the team could migrate one screen without committing to a full rewrite.

Buying-process friction

Research question
Who is involved before technical validation?
Trigger
Visitor asks about plans, approvals, or enterprise requirements.
Question
How did the last similar purchase get approved?
Captured evidence
Decision process, security-review timing, budget signal, stakeholders, and the page detail needed earlier.
Evidence quality
Strong
Digest output
The buyer already has budget, but security review starts before technical validation. The product page should make deployment and data boundaries easier to inspect.

Daily, weekly, or monthly report

Digest section

Product discovery

3 interviews

[1] Figma-to-Flow evaluator with inherited custom components

[2] Daniel Bolt: Migration evaluator checking whether existing components survive

[3] Buyer asking about security review before technical validation

Findings

[1,2] Existing custom components are the recurring adoption risk.

[1] Figma handoff is designer-led, but developers own correction.

[3] Security review may start before technical validation.

Research questions

Team shape and UI iteration

Teams describe designer-to-developer handoff through Figma links, Slack, or GitHub issues. Developers expect generation to speed up Flow UI work, but custom component mapping is where manual correction starts. [1,2]

Caveat

Three interviews is still early evidence. Treat repeated findings as signals to investigate, not final roadmap input.

Questions

Is this a replacement for product interviews?

No. It gives your product team more evidence from visitors who are already evaluating, confused, blocked, or motivated enough to chat. Use it to find patterns and better questions, then follow up with deeper interviews when the signal is strong.

Will the agent annoy visitors with research questions?

The agent helps first, asks permission before switching into discovery mode, and asks one question at a time. It stops when the visitor seems rushed, frustrated, terse, asks for support, asks for pricing, requests a demo, or otherwise wants direct help.

What kind of output does the product team get?

Product Discovery chats are labeled in history. Reports include numbered interviews, grounded findings, research-question summaries, caveats, and links back to the original conversations.

Can we use this alongside our existing chat tool?

Yes. Use Stand only on pages where visitor questions are product signal, such as pricing, docs, migration, comparison, or product pages. Keep your existing support chat where it already works. Do not run two chat widgets on one page; that would just confuse visitors.

Learn from the visitors who are evaluating you right now.