A practical decision framework for determining the right level of UX and product discovery based on impact, confidence, and risk.
During my time at Sendcloud, I worked as a User Researcher focused on helping teams build the right products. Much of my work involved exploring opportunities, validating ideas, and running experiments. Along the way, the I noticed that decisions about when to invest in discovery were often made ad hoc and varied from team to team.
To bring more consistency to this process, the Sendcloud UX team created an early version of the Discovery Decision Model. The idea was simple: provide teams with a structured way to decide when discovery work was needed and foster alignment within the product department. I want to acknowledge the great work from Joost Wolzak that preceded this model. I’ve continued to refine and expand the model ever since.
Use this model when a problem or solution is proposed to be picked up by UX. Score the proposed change on impact by combining how many users it affects with how often it occurs. Then score your confidence based on the quality of evidence behind the problem or solution. Use the matrix to combine both scores and choose the appropriate discovery approach: no/little discovery, quick check, medium-sized discovery, or extensive discovery.
Go through this model and define the discovery need together with your team’s Product Owner / Manager to get consensus from the get-go.
1. Impact Score (Reach & Frequency)
The impact score determines how heavily the change weighs for your users and business.
Use the matrix by first choosing the row that best matches how often the issue or change occurs, then choose the column that best matches how many users are affected. The cell where they meet shows the impact score as a number of stars. Use that star score to find the related impact category to understand whether the change is a rare exception, niche experience, standard workflow, major feature, or core experience.
Impact matrix
Select a cell and get the result automatically in the bottom bar.
| Volume → Frequency ↓ | Almost no users | A few users | Some of the users | Most of the users | All of the users |
|---|---|---|---|---|---|
| Almost never | ⭐ | ⭐⭐ | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ |
| Rarely | ⭐⭐ | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ |
| Some of the time | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| Most of the time | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| All of the time | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| Volume → Frequency ↓ |
Almost no users | A few users | Some of the users | Most of the users | All of the users |
|---|---|---|---|---|---|
| Almost never | ⭐ | ⭐⭐ | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ |
| Rarely | ⭐⭐ | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ |
| Some of the time | ⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| Most of the time | ⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| All of the time | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
Impact categories
⭐⭐⭐⭐⭐ Core Experience
- All of the users, most/all of the time
- Critical changes to primary platform functionality.
⭐⭐⭐⭐ Major Feature
- Most of the users, some/most of the time
- Affects the daily habits of the majority of your users.
⭐⭐⭐ Standard Workflow
- Some of the users, some of the time
- Impact is localized to specific user segments or tasks.
⭐⭐ Niche Experience
- A few users, rarely/some of the time
- Edge cases or settings that users rarely interact with.
⭐ Extreme Exception
- Almost no users, almost never
- Rare anomalies that have negligible impact.
2. Confidence Score (Quality of Signals)
The confidence score determines how reliable your assumptions are based on the evidence you currently have.
Select a row and get the result automatically in the bottom bar.
| Score | Qualification of Evidence | Type of Signal / Source |
|---|---|---|
| ⭐⭐⭐⭐⭐ | Validated behavior | Live data, large-scale A/B test results, or convincing usability tests with a working product. |
| ⭐⭐⭐⭐ | Broad consensus | Extensive UX research (interviews/surveys) combined with strong quantitative analytics data. |
| ⭐⭐⭐ | Targeted signals | Multiple separate qualitative signals, user feedback from Customer Success, and supported by market trends. |
| ⭐⭐ | Anecdotal evidence | 1 or 2 pieces of user feedback via CS, sales, or a gut feeling within the team. |
| ⭐ | Pure assumption | Only a stakeholder opinion or a new, untested idea. |
3. Matrix: What Discovery Is Needed?
Now, place the two scores for impact and confidence side by side to determine your action plan. It will fill into one of five categories. Use the category details below to create a plan of action!
| Impact → Confidence ↓ |
Low (1-2) | Medium (3) | High (4-5) |
|---|---|---|---|
| High (4-5) | 🚀 No / Little Discovery Needed | 🚀 No / Little Discovery Needed | 🔍 Quick Check (Sanity check) |
| Medium (3) | 🔍 Quick Check (Sanity check) | 🧭 Medium-sized Discovery | 🧭 Medium-sized Discovery |
| Low (1-2) | 🔍 Quick Check (Sanity check) | 🛑 Extensive Discovery (High Priority) | 🛑 Extensive Discovery (High Priority) |
Discovery categories
🚀 No / Little Discovery Needed
- When: Low impact or very high confidence (everything has already been proven).
- Approach: Design and build directly. Use common-sense best practices. Roll out through a regular release.
- Who to involve: Product Designer, Product Owner / Product Manager, Developer(s)
- Risk if wrong: Minimal. The impact is limited or the solution has already been proven. Mistakes can be corrected easily through normal iteration.
🔍 Quick Check (Sanity check)
- When: Low impact, but also low confidence (e.g. 1 CS ticket about a small button).
- Approach: Do a quick internal review or present the design to 2 to 3 colleagues/users. No heavy process.
- Who to involve: Product Designer, 2-3 Subject Matter Experts, 1-3 representative users (if easily accessible)
- Risk if wrong: Low. You may invest some effort in solving the wrong problem or introducing a small usability issue, but the consequences are localized and reversible.
🧭 Medium-sized Discovery
- When: Medium impact and medium confidence.
- Approach: Validate the problem. Organize 3 to 5 quick user interviews or send out a short, targeted survey to test the volume.
- Who to involve: Product Designer (lead), Product Owner / Product Manager, 3-5 representative users, Domain expert(s), Data analyst (optional, for usage signals)
- Risk if wrong: Moderate. You may spend significant development effort on a feature that only partially addresses user needs or solves the wrong problem, reducing adoption and business value.
🛑 Extensive Discovery (High Priority)
- When: High impact, but low to medium confidence (high risk if you get it wrong).
- Approach: Start a full discovery track. First conduct qualitative research into the real problem (interviews), then test concepts with interactive prototypes before a single line of code is written.
- Who to involve: Product Designer (lead), Product Manager / Product Owner, User researcher (if available), Multiple user groups, Domain experts, Engineering lead / architect, Business stakeholder(s)
- Risk if wrong: High. You risk investing substantial time and resources into the wrong solution, creating user frustration, low adoption, rework, and potentially business-critical consequences.