From Signals to a Bet

Step 1: Break Notes Into Single Observations

Go through your notes and put each separate observation on its own card - physical sticky or a digital board. One card holds one thing someone said or did. Keep the exact words. "I dread Monday reporting" stays as written; don't tidy it into "dislikes reporting," because the tidy version quietly deletes the feeling.

Step 2: Cluster by Affinity

Move similar cards together into groups, and let the groups form from the bottom up - don't start from categories you decided in advance. When a cluster takes shape, name it in plain language: "re-entering the same data in three tools." This is affinity mapping, and the point is that the themes emerge from the evidence, not from your assumptions.

Step 3: Count and Weigh Each Cluster

For every cluster, mark two numbers:

Frequency:how many different people mentioned it. Five people across five separate interviews is a real pattern; one person saying it five times is one person.

Intensity:how much it hurts - a mild annoyance, or a Monday-ruining ordeal.

A cluster that's both frequent and intense is a strong signal. One that's rare or mild is a lead, not a finding. This is how you stop the single loudest interview from hijacking the whole direction.

Step 4: Pull Out Jobs, Pains, and Gaps

Turn the top clusters into structured insight:

• Job: what the person is trying to get done ("keep clients updated on progress").
• Pain: where it hurts today ("the data lives in three tools that don't talk").
• Gap: what's missing between the two.

These three - in the user's own words - are the raw material you'll shape into an opportunity.

Pro tip: Do this with the product trio in the room - PM, designer, engineer clustering the same cards together. You skip the argument three weeks later about "what the research actually said," because everyone watched the pattern form.

Where AI Helps

Prompt: "Here are my raw discovery notes: [paste]. Break them into single observations, cluster them into named themes, and for each theme count how many separate people mentioned it. Keep the exact quotes."

AI clusters fast; you keep the sharp words.It's genuinely good at grouping and counting a messy pile, but it tends to smooth vivid quotes into bland labels. Check its clusters against the raw notes and put the real phrases back - the sharp phrase is the one you'll use later.

Framing the Opportunity

The moment analysis ends, the temptation is to blurt out "let's build a dashboard." That's a solution - and probably not the best one. Frame the opportunity first, and you keep the door open to five better ideas than the one you thought of in the meeting.

Opportunity, Not Solution

An opportunity is an unmet need. A solution is one way to meet it. "Agencies can't see the true status of work across their tools" is an opportunity; "build a status dashboard" is one solution to it, and maybe the fifth-best. Name the need, and you get to compare solutions instead of marrying the first.

So rewrite every solution-shaped note as a need:

• ❌ "Add a notifications feed" → ✓ "People miss updates that happen while they're away"
• ❌ "Build a dashboard" → ✓ "Agencies can't see status across tools"

Map It on an Opportunity Solution Tree

Teresa Torres's opportunity solution tree keeps this honest. It has four levels, top to bottom:

• Outcome - the one metric you're trying to move.
• Opportunities - the unmet needs from discovery, branching off the outcome.
• Solutions - ideas branching off each opportunity.
• Experiments - ways to test each solution.

Drawing it forces every solution to hang off a real opportunity, and every opportunity off a real outcome. Anything that won't attach is a distraction, and now you can see it.

Write the Opportunity as a Statement

Use one fixed shape so opportunities are comparable to each other:

[Segment] can't [job] because [obstacle].

For example: "Small agencies can't give clients an accurate status because the data lives in three disconnected tools." Same shape every time means you can line opportunities up side by side.

Pick One to Pursue

Discovery usually surfaces several opportunities. Choose with what you already learned: how many people have it (reach), how much it hurts (impact), and whether solving it would actually move your outcome. Pick one to carry forward - the rest stay on the tree for later, not in the bin.

Pro tip: If an opportunity names a feature, it's a solution wearing a disguise. Keep asking "what need does that serve?" until the sentence describes a struggle, not a screen.

Where AI Helps

Prompt: "Here are my jobs, pains, and gaps: [paste]. Rewrite each as an opportunity in the form '[segment] can't [job] because [obstacle]', and flag any that are actually solutions in disguise."

AI reframes solutions into needs; you choose the bet.It's good at catching when you've smuggled a feature into an opportunity, but which one is worth pursuing depends on your outcome and context - that judgement stays yours.