Product discovery: how to learn what not to build before you build it

Product discovery is how a team decides what to build: story-based interviews, three competing ideas, tests of the riskiest assumptions and the right expert call.

Instant Expert EditorialPublished Updated 6 min read

Product discovery is how a team decides what to build before spending engineering time on it, and one of its most useful results is "don't build that." You learn what customers need, come up with more than one way to meet the need, and test the assumptions behind each idea until you can choose. The output is a decision. Often it is "build a smaller version of this."

Discovery runs alongside delivery (building and shipping), and most product teams do some of both every week.

Discovery exists to shrink four risks

Marty Cagan of SVPG groups the risks in a new product idea into four types (The Four Big Risks):

  • Value: will customers buy it, or will users choose to use it?
  • Usability: can users figure out how to use it?
  • Feasibility: can your engineers build it with the time, skills and technology you have?
  • Business viability: does it work for the rest of the business, such as your sales channel, legal and contract constraints, the cost of acquiring customers and how you charge?

Cagan argues that value risk is usually the hardest, and that strong teams tackle value and business risk early instead of after launch. GOV.UK's service manual makes a related point about its own discovery phase: you shouldn't start building during discovery, and stopping at the end is not a failure if the research shows that is the right call (How the discovery phase works).

The example: a customer asks for pick-route suggestions

Here is a made-up example to follow through each step. You run a small team building warehouse management software for regional distributors. A large customer asks for automatic pick-route suggestions, so pickers walk less.

That request is a solution. Discovery asks what problem it solves, and who else has it. (The customer is large, so the temptation to just build it is also large.)

1. Start from an outcome and real stories

Teresa Torres's opportunity solution tree is a common way to structure this work (Opportunity Solution Trees). At the top is the outcome your team is trying to move, ideally a customer behavior in the product. Below it sit opportunities (customer needs, pain points and desires), then possible solutions, then the assumption tests that show whether each works.

Opportunities come from the stories customers tell in interviews, not from the team's guesses. Torres suggests three to four story-based interviews before drafting a first tree, and revisiting it every three to four interviews after that. A story-based interview asks about a specific past event, with a prompt like "Tell me about the last time..."

For the warehouse team, the outcome might be "more orders picked per labor hour at customer sites." Interviews with pick supervisors about their last busy shift might surface:

  • Pickers walk to the wrong spot for items stored in two places.
  • New temporary staff don't know the layout.
  • Rush orders interrupt planned batches.

Route suggestions might help with the first. They do nothing for the other two.

Torres offers a useful check: if there is only one way to address an "opportunity," it is a solution in disguise. "I want route suggestions" has one answer. "I waste time walking to the wrong aisle" has many.

Customer interviews covers how to plan and run these conversations, and problem interviews vs. solution interviews helps you tell which kind you are running.

2. Compare three solutions, not one

Pick one target opportunity and take three ways to address it forward. Torres recommends three because comparing options leads to better decisions than asking whether a single idea is good enough.

For "pickers walk to the wrong location," the team might consider:

  • route suggestions on the handheld scanner;
  • a re-slotting report that recommends moving fast-selling items closer to packing;
  • showing an item's second storage location on the pick screen.

3. Test the riskiest assumption first

Every idea rests on things that must be true for it to work. Torres groups these assumptions into five types: desirability, viability, feasibility, usability and ethical (Assumption Testing). The four common kinds of test Torres describes are prototype tests, one-question surveys, mining data you already have, and short engineering investigations called research spikes. Decide what result counts as success before you run each one, as Torres recommends.

To choose what to test first, Torres points to David Bland's assumption mapping: rank each assumption by how important it is to the idea and how much evidence you already have. Test the important ones with the weakest evidence.

IdeaRiskiest assumptionQuick test
Route suggestionsPickers follow a suggested route even when a supervisor calls out a rush orderPrototype on a handheld, watched during a real shift
Re-slotting reportSupervisors can move stock without approval from someone elseAsk in the next three interviews; check with a former operations manager
Show second locationMost wasted walking comes from items stored in two placesLook through existing pick logs for split-location items

Route suggestions may still win. You now know which assumption would sink each idea, and you can often check it with a small test before any engineering starts.

4. Ask a domain expert about constraints users can't see

Users are the right source for what they did and what went wrong, and often the wrong source for constraints outside their job. A pick supervisor may not know who approves re-slotting across sites, how distributors evaluate add-on software modules, or what a labor agreement says about tracking pickers. Those are viability and feasibility questions.

That is when someone who has worked across many sites earns a call: a former distribution center manager, or someone who has implemented warehouse software. Use them to check constraints and sharpen your user questions, not to replace users: their picture of what pickers do minute to minute is secondhand. Expert interview covers how to run one, and market research interviews with customers, suppliers and operators covers whom to ask.

5. Decide, then keep going

After a round of tests, some assumptions usually held and some failed. Torres suggests asking what you learned about customers, as well as which idea came out ahead. You might build a smaller version of one idea, redesign around what failed or pick a different opportunity. GOV.UK frames the end of discovery the same way: decide whether there is something worth building and whether it is worth the cost.

Once you have chosen, usability testing checks whether people can use what you design. If the idea started as a customer request, see how to test a feature request.

Your next step

Write your team's outcome at the top of a page. Under it, list the opportunities you have heard in real interviews, and circle one. Write three ways to address it, then the one assumption behind each that you have the least evidence for. Pick the cheapest test for each.

If the assumption you can't test from inside the company is about an industry constraint, describe the people who would know to Instant Expert, such as "former distribution center operations managers." A narrow constraint question often fits a written request, such as "who has to approve moving fast-selling stock closer to packing?" They answer in text or by voice note, and you're charged only for answers you get. Send a 30-minute call request when an answer opens more questions. You set what each person receives, at least $5, and our fee is added on top. The operations professionals in warehousing directory page is one place to start.