UX Research · For leads · 5 min read
Showing the impact of research
For leads: "we ran 24 interviews" is effort, not impact. How to show what research changed: track every recommendation to adoption, link the metrics design moves to the metrics leadership watches, and stay honest about what you can and can't claim. With a worked Recommendation-Adoption Score.
Nobody funds research because it happened. They fund it because something changed. A quarterly slide with the number of studies tells leadership the team was busy. A slide that says "these four recommendations shipped, and support contacts about splits fell" tells them what the busyness was worth. Showing impact is a habit you build from the first study, not a report you write when the budget is at risk.
What you'll be able to do after reading:
- Tell activity metrics from outcome metrics, and report the second
- Track recommendations from delivery to adoption
- Calculate a Recommendation-Adoption Score
- Link a design change to a business metric without overclaiming
1. Outcomes, not activity
NN/g's Lola Famulegun names the two most common mistakes: reporting effort ("24 interviews and 3 usability studies this quarter") instead of what changed, and reporting UX-only measures such as a SUS score to executives who don't track them (NN/g, 2026). Keep those numbers; they belong in research reports. For leadership, frame the work around what they already ask about: revenue, cost, risk, speed to market, and retention and satisfaction.
The bridge is the difference between two kinds of metric:
- Upstream metrics describe how the design performed: task success, error rate, SUS (see Measuring usability).
- Downstream metrics describe what changed for the business: support volume, conversion, churn. "Downstream metrics tell you what the work was worth" (NN/g, 2026).
- 1Finding
People don't trust a split they can't check.
5 of 8 participants - 2Recommendation
Show the working under every split.
Owner: Splits PM - 3Change shipped
The working is visible, released in May.
Verified by the researcher - 4Upstream
Splits questioned in the chat, in a follow-up test: fewer.
How the design performs - 5Downstream
Support contacts about splits, the quarter after: lower.
What it was worth
Most of the data already exists inside the organisation, with finance, analytics, support or marketing. If you don't have access, getting it is the first step (NN/g, 2026).
2. Track every recommendation
Brian Utesch and Tammi Fitzwater call the gap between a recommendation and a product change breakage: value lost because a recommendation is dropped, diluted, deprioritised or ignored between the readout and the release. "Breakage thrives when no one measures adoption" (NN/g, 2025). To measure it, they built the Recommendation-Adoption Score (RAS) at Cisco (NN/g, 2026):
- Pick a scope (a product, a programme, a study) and gather its recommendations, ideally from the last 12 months.
- Give each a status. Adopted (shipped and verified by the researcher) earns 3, 2 or 1 points for high, medium or low value. Committed (on a roadmap, resourced) earns 0.66, whatever its value. Communicated (delivered, no plan) earns 0. Cancelled is left out.
- RAS = actual value ÷ possible value × 100, where possible value is what every non-cancelled recommendation would earn if adopted at its value level.
The authors' bands: 0–29 poor, 30–54 fair, 55–79 good (where most organisations should aim), 80–100 great. At Cisco, the score started in the poor range and reached good after about a year; the main benefit was that adoption became visible and started new conversations with product teams.
| Recommendation | Value | Status | Points earned | Points possible |
|---|---|---|---|---|
| Show the working under every split | High | Adopted | 3 | 3 |
| Let the household "banker" send one reminder to everyone | High | Adopted | 3 | 3 |
| Rename "Settle" to "Mark as paid" | Medium | Adopted | 2 | 2 |
| Remove the empty-state tutorial | Low | Adopted | 1 | 1 |
| Share a split to the group chat | High | Committed | 0.66 | 3 |
| A move-out flow that closes balances | Medium | Committed | 0.66 | 2 |
| Split by usage for energy bills | High | Communicated | 0 | 3 |
| Monthly summary e-mail | Low | Communicated | 0 | 1 |
| Two more, rejected at planning | – | Cancelled | left out | left out |
RAS = 10.32 earned ÷ 18 possible × 100 = 57, "good" on the authors' scale (55–79). The two committed items alone would lift it to about 78 once they ship and are verified.
Two conditions make it work. Every recommendation must be specific, linked to evidence, and have "a named owner who has both authority and accountability" (NN/g, 2026). And a low score is a prompt to check your own work first: were the recommendations clear, and did they arrive in time? See Research reports people read.
3. Stay honest
- Don't overstate. Many things change a metric at once: a campaign, a season, a price change. Say "after the change, support contacts fell 30%", not "research cut support contacts 30%". NN/g's advice is the same: even directional data persuades when it's credible (NN/g, 2026).
- Count what didn't happen. A study that stopped a feature nobody needed saved real money; write down the estimate the team would have spent, and who agreed with it.
- Use the trend, not one score. New research lowers the RAS for a while, which is normal; a drop that lasts two or three quarters signals a real problem (NN/g, 2026).
4. Make it a habit
An impact log is enough to start: one row per recommendation, with the study, the owner, the status, the date, and any metric before and after. Update it at every planning cycle, and bring three rows of it, not the whole table, to every quarterly review. Over a year, those rows are your case for headcount, tools and time (see Building a research practice).
5. Common mistakes
- Reporting the number of studies as if it were impact.
- Showing SUS to executives with no link to anything they track.
- Claiming a business result the research only contributed to.
- Recommendations with no owner, so nobody can adopt them.
- Tracking only the successes.
6. Try it (no tools needed)
- Take the last study you ran or read, and list its recommendations.
- Give each a status: adopted, committed, communicated or cancelled.
- For one adopted recommendation, write the upstream metric it should move and the downstream metric that would show it mattered.
Reflection questions: Which of your recommendations had no owner? Who in finance, analytics or support could give you the downstream number?
Checklist
Learn more
- Lola Famulegun, Stop Reporting UX Activity and Report Business Outcomes (NN/g, 2026)
- Brian Utesch and Tammi Fitzwater, The Recommendation-Adoption Score (NN/g, 2026) · Insights Aren't Outcomes: Research Recommendation Breakage (NN/g, 2025)
- Laura Klein, How to Get Research Recommendations on the Roadmap (NN/g, 2026)
Books, courses and videos to go deeper
- It's Our ResearchTomer SharonBook · Morgan Kaufmann, 2012 (Amazon)
- Research That Scales: The Research Operations HandbookKate TowseyBook · Rosenfeld Media, 2024 (Amazon)
Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.
Related
Building a research practice · Research reports people read · Defining success · What research is for