Your NPS is +32. Now what? Most teams stop at the score. But the answers to “why did they rate us this way?” are in the verbatims — and almost nobody analyses them at depth. Here's how five analysis techniques transform NPS from a metric into an action plan.
The score tells you the weather, not the climate
A Net Promoter Score is a single scalar summarising thousands of distinct experiences. It moves. It is reported. It is rarely explained. The explanation lives in the second question — the open-ended one — and in most programmes that field is read by a human for twenty minutes a quarter and then filed.
The gap is not effort. It is method. Reading verbatims is not analysis; classification, aspect extraction and risk scoring are.
Technique 1 — Aspect-based sentiment
Flat sentiment collapses “the product is excellent but support took nine days” into a neutral score. Aspect-based sentiment analysis splits the sentence into its targets and scores each one independently, so Product reads +0.7 while Support reads -0.5.
That split is what makes a verbatim actionable: it names the owner of the problem.
Want to try this on your own data? Run a free VOC health check — no credit card, EU-processed.
Start freeTechnique 2 — Driver analysis
Once aspects are scored, correlate each one against the numeric rating. Some aspects are loud but irrelevant; others appear in a tenth of the volume yet explain most of the movement in detractor rates. Prioritise by explanatory power, not mention count.
Technique 3 — Root cause classification
“Poor support” is at least six different problems: slow first response, repeated handoffs, wrong answers, no ownership, channel friction and tone. A governed taxonomy forces those apart so the fix has an address.
Technique 4 — Complaint-risk scoring
Not every negative verbatim deserves the same response. Score each one on escalation language, regulatory exposure, customer value and churn signal, then triage the top decile. Eight billing complaints can outweigh two hundred shipping grumbles.
Technique 5 — Trend and cohort comparison
Run the same governed taxonomy quarter over quarter. Comparability is the whole point: if the categories drift with every re-run, you are measuring your analysis, not your customers.
Key takeaways
- The score is a headline; the verbatim is the article.
- Split sentiment by aspect before you prioritise anything.
- Rank by explanatory power and risk, never by raw volume.
- Freeze the taxonomy so quarters stay comparable.
About the author
Elena Moretti
Head of CX Strategy, InsightNarrator
Ten years building CX programmes for European enterprises, previously at a top-5 European retail bank. She writes about turning verbatims into decisions.
Connect on LinkedIn