skills / research

research-synthesis

what it does

Synthesize user research into themes, insights, and recommendations. Use when you have interview transcripts, survey results, usability test notes, support tickets, or NPS responses that need to be distilled into patterns, user segments, and prioritized next steps.

Verified Before and after: 2 models 1 try per case View source Download .zip

Compare the answers

Answers from Claude Sonnet 5.5 and GPT-6.1 Sol · try 1 of 1

Same request, same model, once with the skill and once without. Read both and judge for yourself.

“I ran six interviews about our expense app last week. Can you pull out the main themes and what we should do about them?”

Source material given with the request

Product: Clearclaim, a mobile app for submitting work expenses. P1 (freelance consultant): "I took a photo of a taxi receipt at night, it came out dark and the app rejected it. I gave up and typed it all in by hand." Submits once a month. P2 (sales rep): Receipt photo rejected twice in the last week. "I log mileage in a notebook and add it up on Fridays. Takes 20 minutes." P3 (CFO, 400-person company): "If you don't integrate with SAP, we're out. Non-negotiable." Did not mention receipts or approvals. P4 (field technician): "Thermal receipts get crumpled in my van and the scanner can't read them." Also says approvals are slow, so he stops submitting until month end. P5 (office manager): "I wait about nine days for my manager to approve anything." Has no issue with receipt capture. P6 (sales rep): "Half my receipt photos bounce." Tracks mileage by hand in a spreadsheet.
with the skill
## Research Synthesis: Clearclaim Expense App Interviews **Method:** Interviews | **Participants:** 6 | **Date:** Last week | **Researcher:** [Your name] ### Executive Summary Receipt capture failure is the strongest signal, reported by 4 of 6 participants and leading to abandonment or manual entry. Slow approvals (2 of 6) cause people to delay submitting, and manual mileage tracking (2 of 6, both sales reps) is a possible gap. SAP integration was raised by only one participant, but he is the CFO of a 400-person company and called it non-negotiable, so it is a purchase blocker rather than a usability issue. With six interviews, treat these as hypotheses to validate, not firm rankings. ### Key Themes #### Theme 1: Receipt capture fails in real-world conditions **Prevalence:** 4 of 6 (P1, P2, P4, P6). P5 said capture was fine. P3 didn't mention it. **Summary:** Photos and scans get rejected, and users either retry, type the data by hand, or give up. **Supporting Evidence:** - "I took a photo of a taxi receipt at night, it came out dark and the app rejected it. I gave up and typed it all in by hand." — P1 - "Thermal receipts get crumpled in my van and the scanner can't read them." — P4 - "Half my receipt photos bounce." — P6 - P2 had a receipt rejected twice in the past week. **Implication:** The failure conditions differ: low light (P1), damaged thermal paper (P4), and unspecified (P2, P6). There may be more than one root cause. P6's "half" is a self-estimate, so check actual rejection rates in your data. #### Theme 2: Slow approvals change submission behavior
Show the rest (63 more lines)
**Prevalence:** 2 of 6 (P4, P5) **Summary:** Waiting on managers is slow, and at least one user responds by batching submissions. **Supporting Evidence:** - "I wait about nine days for my manager to approve anything." — P5 - P4 says approvals are slow and he stops submitting until month end. **Implication:** Slow approvals appear to delay submission as well as reimbursement. The bottleneck is the approving manager, who wasn't interviewed. #### Theme 3: Mileage is tracked outside the app **Prevalence:** 2 of 6 (P2, P6), both sales reps **Summary:** Both log mileage by hand (notebook, spreadsheet). P2 spends about 20 minutes each Friday totaling it. **Supporting Evidence:** - "I log mileage in a notebook and add it up on Fridays. Takes 20 minutes." — P2 - P6 tracks mileage in a spreadsheet. **Implication:** This may be an unmet need for high-travel roles. We don't know whether Clearclaim lacks mileage features or whether users don't know about or trust them. #### Theme 4: ERP integration is a buyer requirement **Prevalence:** 1 of 6 (P3) **Summary:** The CFO won't adopt without SAP integration. **Supporting Evidence:** - "If you don't integrate with SAP, we're out. Non-negotiable." — P3 **Implication:** This is a different kind of finding. P3 is a purchaser, not a daily user, and didn't discuss the user experience. One data point can't size the opportunity, but it is a potential deal-breaker for larger companies. ### Insights → Opportunities | Insight | Opportunity | Impact | Effort | |---------|-------------|--------|--------| | Photos rejected in low light and on crumpled thermal receipts | Better capture guidance (flash prompt, flatten tip), more tolerant OCR, and a "submit anyway" path with manual correction | High | Med | | Rejection leads to abandonment or manual re-entry | Never hard-block: let users submit the image and flag it for review | High | Low–Med | | Slow approvals cause batching | Approver reminders, escalation, or delegate approval | Med | Low–Med | | Sales reps track mileage by hand | Mileage tracking or a simple mileage entry flow (after confirming current state) | Med | Med–High | | SAP is a purchase requirement for a large company | Scope the SAP integration and validate demand with other prospects | Unknown | High | ### User Segments Identified Tentative, given the sample size. | Segment | Characteristics | Needs | Size (of sample) | |---------|----------------|-------|------| | Mobile/on-the-road submitters | Sales reps, field tech, freelancer. Capture receipts in difficult conditions | Reliable capture, mileage support | 4 of 6 | | Office-based submitter | Office manager. No capture issues | Faster approvals | 1 of 6 | | Economic buyer (enterprise) | CFO. Evaluates on integration | SAP/ERP integration | 1 of 6 | ### Recommendations 1. **Fix receipt capture and remove hard rejections.** It's the most widespread issue (4 of 6) and causes abandonment. Start by pulling rejection rates from analytics, broken down by cause. 2. **Investigate the approval workflow.** Two participants described slow approvals, one changing his behavior because of them. Interview a few approvers. 3. **Clarify the mileage situation.** Check whether a feature exists and goes unused or is missing, then decide whether to build. 4. **Size the SAP demand before committing.** Ask sales or support how often ERP integration comes up in lost or stalled deals. ### Questions for Further Research - Does Clearclaim have mileage tracking, and if so why aren't sales reps using it? - What is the real receipt rejection rate, and what share is lighting, damage, or other causes? - Why are approvals slow: notification gaps, manager workload, or policy? - How common is ERP integration as a purchase requirement, and which systems? - Do P1, P4, and P5 also have unmet needs they didn't raise? Absence of a mention isn't absence of a problem. ### Methodology Notes - Six participants is too few to size themes. The percentages above are counts within this sample, not population estimates. - The sample skews toward sales reps (2 of 6), and there is only one buyer and no approving managers. - The CFO's comments reflect the purchase decision, not hands-on use, so they shouldn't be weighed directly against user pain points. - Where participants gave only a brief quote, I summarized only what they said. Further interview detail (frequency, device, app version) would strengthen the capture findings. - If you can share the full transcripts, I can add more quotes and check for details I couldn't see in these excerpts.
without
# Themes from the six interviews ## 1. Receipt capture is failing (4 of 6: P1, P2, P4, P6) This is the clearest signal. - P1: dark photo of a taxi receipt was rejected, so they typed everything by hand. - P2: rejected twice in one week. - P4: crumpled thermal receipts can't be scanned. - P6: "half my photos bounce." The failure conditions differ (low light, crumpled thermal paper, unknown), so this is probably several problems, not one. The costs are also real: P1 abandoned the feature, and P4 delays submitting. **What to do** - Pull rejection logs to see the actual reasons and rates, and fix the most common first. - Replace hard rejection with "submit anyway, flag for review," so a bad photo never blocks a claim. - Add capture guidance in the moment, such as a low-light/flash prompt and edge detection. - Test specifically against thermal and crumpled receipts. ## 2. Slow approvals (2 of 6: P4, P5) - P5 waits about nine days. - P4 stops submitting until month end because approvals are slow, so the delay also changes how people use the app.
Show the rest (33 more lines)
The bottleneck is the approving manager, not the app itself, so the fix is probably in workflow. **What to do** - Check the approval time data to see how widespread this is. - Test manager reminders, escalation after X days, and delegate approvers. - Ask P4 and P5 whether they can see approval status, and whether knowing it would help. ## 3. Manual mileage tracking (2 of 6: P2, P6, both sales reps) Both track mileage by hand (notebook, spreadsheet). P2 spends about 20 minutes every Friday. Neither complained about the app here, since they're describing a workaround. I wouldn't assume they want an in-app solution without checking. **What to do** - Run a few follow-up conversations with field and sales users: would they use automatic or GPS mileage logging, and do they trust it for reimbursement? - Check whether other reps in your user base show the same pattern. ## 4. SAP integration (1 of 6: P3) P3 is a CFO and a buyer, not an everyday user, and didn't mention receipts or approvals. This is a purchasing requirement, not a usability problem, and one interview isn't enough to prioritize it. **What to do** - Ask sales how often SAP comes up in lost or stalled deals. - Decide whether you're targeting companies of that size. If so, this may be a gating issue, and if not, it can wait. ## Priorities 1. **Receipt capture.** It has the widest reach, the clearest evidence, and likely a tractable fix. 2. **Approval delays.** Moderate evidence, and a lightweight reminder/escalation test is cheap. 3. **Mileage.** A promising opportunity, but validate demand first. 4. **SAP.** Treat it as a business decision, informed by sales data. ## Caveats - Six interviews show patterns, not prevalence. Back these up with usage and rejection data. - Your sample mixes users and a buyer, and their needs differ. Keep those tracks separate. - Sales reps appear twice, so mileage may be a segment-specific need rather than a general one.

Answers are shown exactly as the model wrote them, from the first try, and nothing is picked by hand.

The models

2 models

Claude Sonnet 5.5

made by Anthropic

claude-sonnet-5-5 · revision not reported by the provider

1 try per case · tested 1 Nov 2026

GPT-6.1 Sol

made by OpenAI

gpt-6.1-sol · revision not reported by the provider

1 try per case · tested 1 Nov 2026

The tasks

written by Carte Blanche
biased-feature-survey

“We surveyed users about what to build next. Write up the findings for the team.”

expense-interviews

“I ran six interviews about our expense app last week. Can you pull out the main themes and what we should do about them?”

export-usability-test

“Summarize what we learned from this usability test on report export.”

nps-verbatims

“Here are last month's NPS responses. What's our score, and what's driving it?”

support-ticket-spike

“Group these support tickets and tell me what we should look at first.”

Install

Download the .zip and unzip it. Put its files, in a folder named research-synthesis, here:

Claude Code~/.claude/skills/Every project. For one project only, use .claude/skills/ inside it.
Codex~/.agents/skills/Every project. For one repository only, use .agents/skills/ inside it.
Cursor~/.cursor/skills/Cursor also reads the Claude Code and Codex folders above.

Then start a new session and ask for what the skill does. Other agents that read SKILL.md have their own skills folder; check their docs.

Coming after launch: get skills from inside your agent, no download needed.