IGNITE EXTENDED / LESSON 04
How can you uncover a need without pitching your solution?
An empathy log and two evidence-backed needs.
For decades Stanley sold vacuum bottles and thermoses to tradespeople and outdoor customers. Its 40-ounce Quencher tumbler got little attention, and by 2019 it was not a priority. In 2017 three women in Utah had started The Buy Guide, a shopping blog and Instagram account whose audience was 97.7% women, and they made the Quencher their main recommended product. When they heard it might be dropped, they told followers to buy whatever stock was left. Stanley had no affiliate program, so it asked them to place a wholesale order. The first 5,000 cups sold in about four days and the next 5,000 in about an hour. The founders then told Stanley executives that the cup was being marketed to the wrong people. In 2020 Terence Reilly, formerly of Crocs, became Stanley's president. He made the hydration category a priority, partnered with The Buy Guide on new colors and moved to limited-edition releases. Stanley's sales grew from $73 million in 2019 to $402 million in 2022, and CNBC reported a projection above $750 million for 2023.
Sources: CNBC (Dec 2023): How a 40-ounce cup turned Stanley into a $750 million a year business · Retail Dive: Stanley Quencher tumblers' viral success
Stanley's sales data said the Quencher was a weak product. What would an empathy-led team have watched for and asked about, and which of Stanley's assumptions about who its user is did that evidence overturn?
An observation records what a person did or said in a specific situation. An interpretation proposes why it happened. A need describes something the person is trying to accomplish. Keeping these separate makes it possible to revise an explanation without losing the original evidence.
The empathy material emphasizes implicit needs because customers often describe a familiar solution rather than the underlying purpose. A request for a larger calendar may express a need to coordinate several people with confidence. The underlying need admits more possible solutions than the requested interface.
Moving upward on the ladder asks why a requested feature matters. Moving downward asks how the person currently accomplishes the task and what a concrete episode looked like. Both movements are necessary. Purpose without behavior becomes vague, while behavior without purpose can lead to a narrow feature fix.
Use the person’s account to test each step. Avoid asking “why” in a way that pressures them to invent a rational explanation. Questions about the last occurrence, the surrounding circumstances and the consequence of failure often provide stronger evidence.
Use the loop to turn a surprising customer episode into a testable explanation. The example follows a coordinator checking availability across several calendars.
Record what happened before interpreting it: the coordinator opened three calendars, messaged two colleagues and postponed confirmation. Keep the setting, sequence and tools in the evidence log.
Explore that recent episode: what triggered the check, which information was missing and what would happen if the meeting conflicted? Ask for behavior and consequences before requesting opinions about a proposed app.
Propose a need that explains the behavior: confidence that required participants can attend. This is broader than a request for a larger calendar and allows several possible solutions.
Seek another explanation and a contrasting episode. The obstacle could be outdated records, restricted permissions or a colleague who never updates availability. Each explanation suggests a different response.
The next distinction, observation versus interpretation, protects the loop: changing your explanation must not erase the original event.
Preserve what happened separately from your explanation. That separation lets another researcher propose a different explanation without losing the underlying evidence.
“The customer copied the same address into three forms.”
“The customer needs a single shared record.” The interpretation may be useful, but it is not the observation itself.
Illustrative application: a coordinator opens three calendars. “She distrusts the shared calendar” is one explanation. Permission restrictions or missing external participants are alternatives. Ask about the recent episode and observe another attempt before turning distrust into the problem the product must solve.
A need statement is an informed interpretation. Keep the observation that supports it and the alternatives still unresolved.
An extreme user experiences a relevant condition more intensely than an average user. They may have unusual frequency of use, physical constraints or a particularly demanding environment. Their workarounds can expose needs that are difficult to see in a convenient sample.
An extreme case helps generate and refine a hypothesis. It does not estimate how common the need is in the market. After identifying the mechanism, investigate where else it appears and which customers would pay for a solution. This connects empathy research to later segmentation work.
A hypothetical question asks people to predict their future behavior. An episode question asks them to reconstruct something that happened. Prediction can be useful, but it is vulnerable to politeness, memory shortcuts and a desire to appear consistent.
The interview guidance favors short questions and listening for the person’s language. Follow a surprising detail instead of rushing through a fixed list. Record contradictions and workarounds. They often reveal the difference between a stated preference and the constraints governing actual behavior.
Move up by asking why; move down to recover concrete behavior.
Shadowing a clinic's front desk puts you next to protected health information. Plan the session so you never capture it.
Write down roles, times, counts and actions, for example "called patient 3 of 5," and leave out names, phone numbers and diagnoses. Photograph tools and workarounds only after checking that no screen or list showing patient data is in the frame.
Get the clinic's permission in advance, tell staff what you will and won't record, and keep notes in a shared folder that nobody outside the team can open. Whether HIPAA or state privacy law applies to a particular visit depends on current law and the facts, so get legal review before you start.
“Yesterday I sent the delivery instructions by email. The driver called because he could not find them. I texted a photo, but the office later used the old address. I do not need more messages; I need to know everyone has the right version.”
What is directly reported, and which needs are still hypotheses?
Nielsen Norman Group compared AI-generated "synthetic users" with real participants. The synthetic users claimed they had finished online courses that real learners admitted abandoning. They praised discussion forums that most real learners avoided, and they treated every need as important. NN/g recommends using them only for desk research or to pilot an interview guide, never in place of real users.
An LLM can turn a question list into a draft guide, flag leading questions, or tag transcripts by theme. You still have to keep each tagged passage tied to the original episode and check which contradictions a summary has smoothed over.
If an interview or shadowing session captures patient details, any cloud transcription or storage vendor handling that data may be a HIPAA business associate and need a business associate agreement (BAA). Under HHS guidance, cloud providers that store or process electronic PHI are business associates. The simplest safeguard is to design the session so no patient identifiers are recorded. Whether HIPAA applies depends on current law and the facts, so get legal review.
Sources: Nielsen Norman Group (2024): Synthetic Users: If, When, and How to Use AI-Generated Research · Nielsen Norman Group (2025): Synthetic Users: AI Participants (video) · HHS: Business Associates guidance
Use the fictional interview on the preceding slide; treat it as the full available interview packet.
Record three reported events, two possible explanations, an underlying need and three follow-up questions. Do not propose an application yet.
Maya is sure about the problem because she lived it. As operations manager of a 6-location Austin PT group, she watched front-desk staff spend about an hour a day phoning patients to fill gaps. Daniel wants proof that this is not one group's habit. They are about to talk to clinics that might pilot Relay. A typical small clinic has 4 therapists, about 280 visits a week and about 42 empty slots from late cancellations and no-shows (illustrative). The risk is that every conversation turns into a Relay pitch.
Produce an empathy plan and an interview guide. The plan covers who to observe and interview, where and when, and which extreme users to include. The guide has 8–10 episode-based questions. Add a two-column evidence log with three sample entries that keep observation apart from interpretation.
Once Relay is described, interviewees start judging the idea and stop reporting what they do. Polite approval then gets logged as evidence. Keep the product out of the conversation, or hold it back until the episode questions are finished.
Maya's years at one Austin PT group are a source of hypotheses. They are not a sample. Write her beliefs down as assumptions, then interview clinics she has never worked with, including types she doubts will fit.
Owners talk about revenue, coordinators know the workflow and patients decide whether a slot gets filled. A frame built from owner interviews alone can miss the reason the product fails at the front desk.
A Relay empathy plan (who, where, when, extreme users), a front-desk interview guide of 8–10 episode questions and an evidence log template that keeps observations apart from interpretations.
Next lesson: Problem framing, prototyping & learning
Check your understanding
Question 1
On a shadowing visit, a Relay interviewer records four entries. Which one is an observation and not an interpretation?
Only the second entry records what happened: who acted, when, what was done and the result. The other three explain behavior. They may be right, but they need checking, and "stressful" sounds descriptive while actually guessing at a feeling no one stated.
Question 2
Maya says front-desk staff at her former group spent about an hour a day phoning patients to fill gaps. What does this evidence establish for Relay?
A vivid first-hand episode can reveal a mechanism and justify investigation, so it isn't worthless. It comes from one organization, though, and one person's recollection. It cannot estimate prevalence or market size. That needs observations across clinics of different types.
Question 3
An interview guide includes: "Would you use an app that automatically texts your waitlist when someone cancels?" Which replacement best follows the lesson's guidance?
The last option asks someone to reconstruct a real episode, which shows actual behavior, tools and outcomes. The pricing question still asks for a prediction about a pitched solution. The frustration question gets a yes/no opinion. The third is a leading question that invites polite agreement.
Question 4
After three interviews, a Relay team drafts need statements for front-desk coordinators. Which one is a need, as opposed to a disguised solution?
The first option names what the person is trying to achieve and the constraint they work under. It leaves room for texting, policy changes, standing short-notice lists or other answers. The other three each assume an interface. An interview can confirm a need. It can't confirm that a particular feature is the answer.
Question 5
A coordinator at one clinic fills nearly every same-day gap by texting a personal group of "flexible" patients she keeps on her own phone. How should the Relay team treat this interview?
Extreme users are valuable because they make a mechanism visible: here, a pre-screened short-notice group reached through a fast channel. One case can't show how common the practice is, so it doesn't measure the market. Throwing it out loses the most informative interview, and copying it ignores the privacy problem of patient data on a personal phone.
Question 6
A startup sells a medication-reminder device to nursing homes. It has interviewed 15 residents and no one else. What is the most important gap in its evidence?
As the MRI example shows, the person who buys, the person who operates and the person who is affected can have different needs. A device that residents like can still fail if aides find it adds work or administrators can't justify the cost. More resident interviews don't fix that gap.
Workbook
Module output: An evidence log, problem frame, prototype, segment choice, positioning and route to market.
Library
10 appear in the slides with notes; the rest are kept here for reference.
Source S06 · Soule · 8 pages
