"Where did we put that survey from six months ago?" "What did we end up doing with that data anyway?" — at most organizations, surveys are run and then quietly forgotten.
Industry research suggests roughly 70% of the survey data companies collect never feeds into a decision. This article maps the five patterns behind "collected and forgotten," and lays out a process for deciding how the result will be used before you ever send the survey.
Five patterns of "collected and forgotten"
Pattern 1: Collection itself becomes the goal
"We should run an NPS once a year" — "Everyone else does it" — the survey gets fielded for its own sake. The team feels satisfied just having run it, and using the result becomes a deferred task.
Signals
- Team remembers when it was run, but cannot say what was learned
- More time spent debating the survey design than the survey result
- Behavior post-survey is unchanged
Pattern 2: It never reaches the decision maker
An analyst writes the report, but it never reaches the exec or department head. Or it reaches them and is not opened. Or it is opened but never discussed.
Signals
- Reports sit in internal storage
- Never referenced in executive meetings
- "Here are the results" presentation ends the conversation rather than starting one
Pattern 3: Open-text gets read once and dropped
There is a flash of insight when someone reads the open-text responses, but they are never organized or classified, so they cannot be referenced later. With 100+ responses, reading top-to-bottom does not produce a coherent picture, so they end up shelved.
Signals
- People skim and quote whatever stuck with them
- No theme classification done
- Similar responses are not aggregated
Pattern 4: The result is a lonely number
"This quarter's NPS was 35" — and that is the whole report. No comparison to past rounds, no segment breakdown, no driver analysis, so the number does not tell you what to do.
Signals
- Result summarized as a single number (NPS, CSAT, etc.)
- No discussion of "compared to last quarter"
- No segment-level patterns surfaced
Pattern 5: It does not connect to action
The result gets discussed, but "what we'll do next" never gets decided. Or it gets decided but never shipped.
Signals
- Discussion happened, but no action items
- Action items exist, but no owner
- Action was taken, but no re-measurement to verify impact
The leverage is in pre-collection design
The single biggest fix for "collected and forgotten" is to decide how you will use the result before you send the survey.
Five things to settle before collection
1. Whose decision is this for?
✗ "For the company"
✓ "For the head of CS, who is going to redesign onboarding"
Identify the decision maker at the level of a specific person. That sharpens both the question design and the dissemination plan.
2. What action do we ship for each possible outcome?
Pre-define the response for each scenario:
Scenario A: NPS ≥ 30 → activate promoters as advocates
Scenario B: NPS 10–30 → dig into passive customer pain
Scenario C: NPS < 10 → urgent churn-prevention intervention
Locking this in before collection means you can move the moment the result lands.
3. What is the deadline for a conclusion?
✗ "We'll think about it when results come in"
✓ "Aggregated by 6/30, exec review by 7/15, decision by 7/31"
Without a deadline, "still analyzing" becomes permanent.
4. What are we comparing to?
✗ "Look at this single number"
✓ "Compare to last round, industry benchmark, competitor NPS"
A number without a reference point does not mean anything. Decide what you compare to before collection.
5. What counts as "success"?
✗ "Run a survey"
✓ "If NPS in the 3-month re-measure is +2 or better, this was a success"
A success definition set before collection gives the resulting action a measurable goal.
A standard post-collection workflow
Design the post-collection flow once, run it every time.
Phase 1: Aggregation (immediately after collection — week 1)
- Overall numbers
- Comparisons (last round, year-over-year, full-year)
- Segment breakdowns
- Theme classification on open text
Phase 2: Analysis (week 1–2)
- Confirming the surprises and hypotheses
- Per-segment anomalies
- Representative open-text quotes
- Building the "story" of the data
Phase 3: Sharing (week 2–3)
- Decision-maker report (one-page summary)
- Detailed report for stakeholders
- Excerpts company-wide (transparency)
- Feedback to respondents (the step most often dropped)
Phase 4: Decision (week 3–4)
- Discussion at exec / leadership meeting
- Action shortlist
- Owners and deadlines assigned
- Action items recorded
Phase 5: Execution (week 4+)
- Owners ship
- Progress shared on a cadence
- Course-correction as needed
Phase 6: Re-measurement (3–6 months later)
- Re-run with the same questions
- Verify the action's effect
- Feed forward into the next decision
Run this cycle for every single survey. That is the operating discipline of a data-using organization.
Three ways to actually use open-text
Open-text is where "collected and forgotten" shows up the worst. Classification is what unlocks it.
Method 1: Manual classification
Under 100 responses, a human reading every one and categorizing produces the highest accuracy.
Candidate categories:
- Price complaints
- UI/UX complaints
- Support quality
- Feature gaps
- Things they like vs. competitors
- Brand / trust mentions
Method 2: AI classification via prompt
Above 100 responses, paste a CSV into ChatGPT or Claude and ask for categorization.
Example prompt:
Classify the following open-text survey responses into themes
you derive yourself. Tag each response with its theme.
Then list the top 5 themes by frequency and 3 representative
quotes per theme.
Method 3: A tool with built-in open-text AI analysis
Tools like Repoan with native open-text AI analysis let you extract themes and sentiment in one click. Even 1,000+ responses are structured in minutes.
A report format the decision maker can actually act on
Reports that lead to action share a structure.
Recommended one-page summary
[Survey title] Q2 2026 Customer Satisfaction Survey
[Sample] 500 sent, 215 returned (43% response rate)
[Period] 2026/04/01 – 2026/04/30
[Top summary] (3 lines)
- NPS this round: 32 (+4 vs. last). Trending up.
- Most frequent open-text theme: "Slow support response" (28 mentions).
- Pre-churn segment NPS: -5. Urgent intervention needed.
[Key findings]
1. ...
2. ...
3. ...
[Proposed next actions]
1. Revisit support response SLA (Owner: Tanaka, CS / Deadline: 5/31)
2. 1:1 outreach to pre-churn segment (Owner: Sato, Sales / Deadline: 5/15)
3. ...
[Next re-measurement] 2026/07
Following this format alone makes it dramatically easier for execs to engage with the discussion.
Closing the loop with respondents
Easy to forget, but feeding results back to the people who responded is what keeps your future response rate alive:
- "Based on your feedback, we shipped X."
- "Next round, we'll focus on Y in particular."
- "Your input is going into our decisions, here's how."
Send this within 1–2 months of collection. Companies that do this keep their response rates up over time. Companies that send zero feedback see response rates quietly decay as survey fatigue and distrust accumulate.
How Repoan supports this
Repoan is designed to prevent "collected and forgotten":
- AI-generated summary reports — produced automatically after collection
- Time-series comparison with past rounds — supports the habit of trending, not snapshots
- AI theme extraction on open-text — structures 1,000+ responses in minutes
- Segment dashboards — slice on the fly
- Respondent feedback email feature — building closing-the-loop into the operational flow
Summary
To make survey results count:
- Recognize the five "collected and forgotten" patterns
- Decide the five pre-collection items (decision maker, scenarios, deadline, comparison, success criteria)
- Run the standard post-collection workflow (aggregate → analyze → share → decide → ship → re-measure)
- Open-text only matters once it has been classified
- Build a one-page summary the decision maker can act on
- Close the loop back to respondents
Running a survey is easy. The operational muscle to convert results into decisions is what unlocks the data's actual value.