It is fashionable to dunk on NPS, and most of the criticism is fair. The score compresses a rich human opinion into a number, the number swings with sample size, and a quarterly average tells you almost nothing about what to build.
Here is the thing, though: the score was never the valuable part. The valuable part is the sentence underneath it. Ask why did you give that score and thousands of customers, unprompted and in their own words, will tell you exactly where your product delights and where it grinds. NPS is a doorbell. Something is at the door. The verbatims are who.
At Lagoon we run every NPS response through the same four moves before it is allowed anywhere near a roadmap conversation. The pipeline is simple enough to run in a spreadsheet, which is how we did it before we built it into the product:
- Separate the score from the verbatim.
- Tag for theme and for feeling.
- Weight by who, not just how many.
- Turn themes into bets, and say them back.
Here is each move in detail.
Move one: separate the score from the verbatim
The score answers one question only: how is the relationship trending, and for which segment? Track it by plan and by tenure, watch the movement, and resist reading strategy into a single quarter. Relationship metrics are weather reports, not maps.
The verbatims go into a different pipe entirely. Every response is stripped of the number and treated as a piece of raw feedback, sitting alongside support tickets, sales notes and review-site quotes. A detractor comment about slow dashboards should cluster with the support tickets about slow dashboards. Keeping survey feedback in its own silo is how the same problem gets counted three times as three small issues instead of once as a big one.
Move two: tag for theme and for feeling
Each verbatim gets two labels. The first is the theme: what part of the product or experience is this about? Onboarding, permissions, reporting, price. The second is the direction: is this praise, friction, or a request?
The second label matters more than teams expect. Praise is not noise to filter out. Praise tells you which strengths to protect and which ones to put on the homepage. A theme that collects praise from small teams and friction from large ones is not one insight, it is two, and they point at different quarters of work.
Move three: weight by who, not just how many
Raw counts mislead. Twelve mentions from twelve trial users and twelve mentions from your twelve largest accounts are the same bar on a chart and utterly different business problems.
So every theme gets weighted by the accounts behind it: how much revenue they represent, whether they are growing or shrinking, and where they are in their lifecycle. This is the move where a CRM connection earns its keep.
A theme mentioned twelve times is a bar on a chart. Twelve named accounts with renewal dates is a decision.
Keep the weighting boring
Resist the temptation to build a clever scoring formula with nine coefficients. We use three buckets: broad, concentrated in a key segment, or scattered. Anything more precise than that is false confidence wearing a spreadsheet.
Move four: turn themes into bets, and say them back
A weighted theme is still not a roadmap item. The last move is writing the bet: we believe building X will move Y for segment Z, and we will check back in one quarter. The theme is the evidence. The bet is the decision. Keeping them separate means you can be wrong about the bet without doubting the evidence.
Then close the loop in both directions. Inside the company, every roadmap item links back to the themes that justify it, so reviews argue about evidence instead of taste. Outside, the customers whose responses fed a theme hear about it when the work ships. Survey response rates climb when people learn that answering does something.
What this looks like after a quarter
The score still gets reported, and it still swings. But nobody argues about it anymore, because the roadmap conversation runs on the layer underneath: named themes, weighted by real accounts, each attached to a bet with a date on it.
That is the whole trick. Keep the doorbell. Just stop pretending it is the conversation, and build the four moves that let the conversation actually happen.