Where Can You Locate Your NPS Scores and Customer Survey Verbatims: 6 Places Data Actually Lives

You can locate your NPS scores and customer survey verbatims in your survey platform's dashboard and response export, your CRM's feedback fields, your product analytics or data warehouse, and your app store and review-inbox consoles. The score usually lives in one reporting layer; the verbatim usually lives in another. Finding both means matching the platform that asked the question to the system that stores the answer.

Net Promoter Score is a single computed number. Verbatims are the open-text comments that explain it. They are produced in the same survey response but stored, permissioned, and refreshed differently in almost every stack. Teams lose hours because they search for comments in the reporting dashboard and search for scores in the raw export. This guide maps where each artifact actually sits, how the score is calculated, why the two diverge, and how to build one location that both can be pulled from on demand.

How the score and the verbatim split apart

A survey response is one database row. It contains a 0-10 rating, one or more free-text fields, a timestamp, and identifiers for the respondent. When the survey is submitted, the platform does three things with that row. It stores the raw record, it computes a rollup metric, and it fires the text into any connected text-analysis or ticketing system.

The NPS calculation is arithmetic. Respondents rating 9 or 10 are Promoters. Ratings 7 or 8 are Passives. Ratings 0 through 6 are Detractors. The score equals the promoter percentage minus the detractor percentage. On a 0-10 scale, passives inflate neither side of the subtraction. This is why a sample can look healthy on average and still post a low NPS: a cluster of 6s drags the detractor share upward without moving the promoter share.

Verbatims follow a different path. The platform stores them as unstructured text attached to the response ID. Analytics tools tag them with sentiment or topic labels. Helpdesk integrations convert some into tickets. Each path creates a copy. The score is one derived value per segment and period. The text is many rows per period, often duplicated across systems. That structural mismatch is the root cause of every "I can't find it" problem below.

Where each artifact lives, system by system

Most stacks spread the two artifacts across four to six systems. The table below maps the common locations, what each reliably holds, and the freshness you should expect.

The dashboard is where the computed score is easiest to read and hardest to audit. The raw export is where the score is reproducible and the verbatims are complete. The warehouse is where both can live under one schema. Most mature teams treat the warehouse as the system of record and the dashboard as a read-only display.

Why the numbers disagree between systems

Two tools can show different NPS values for the same week and both be correct. The reasons are mechanical, not mysterious.

First, timing. A response submitted at 11:58 p.m. may land in the dashboard rollup immediately and in the warehouse the next morning. End-of-period cuts therefore differ by up to one batch cycle.

Second, scope. The dashboard may score one survey stream; the CRM may blend several streams into an account-level field. Combining touchpoint surveys with relationship surveys changes the denominator.

Third, de-duplication. Survey platforms often allow one response per contact per period. The warehouse may hold every response ever submitted, including repeats. Repeats from the same contact weight that contact more heavily.

Fourth, partial responses. A respondent who rates the scale but skips the comment counts in the score. A respondent who abandons before the rating does not. Text-analysis tools sometimes ingest comments from abandoned sessions, inflating the apparent verbatim volume without touching the numerator or denominator.

Fifth, the formula itself. A platform reporting a "recommendation score" may use a different scale or exclude passives differently. Confirm the exact arithmetic before comparing any two numbers.

The counterintuitive detail: the dashboard score is usually the least auditable number in the stack. It is a computed value with hidden filters, and reproducing it from the raw export frequently yields a small difference. Treat the export as ground truth and the dashboard as a presentation layer.

The timeline of a single response

Understanding where a response sits at each hour explains most "missing data" reports.

Late responses are the most common source of discrepancy. A response completed after the reporting cutoff appears in next period's score and in this period's verbatim export if the export filters on submission time rather than response ID.

Text routes and why the comment is harder to find

Scores are numbers and numbers travel well. Verbatims are free text and pick up friction at every hop. Three routes matter.

Export route. The raw CSV or API contains the comment in the same row as the score. This is the most complete view and the easiest to reconcile. Filter on the open-text column, not the sentiment tag.

Tagged route. Text-analysis tools ingest comments, assign topics and sentiment, and store labels. The original wording may be truncated or paraphrased. If you need the exact quote for a report, pull from the export and use tags only for grouping.

Ticket route. Helpdesk integrations convert negative verbatims into tickets. The ticket copy may omit the rating and respondent metadata. A comment that appears in the ticket queue may be absent from the analytics tool if the routing rule filtered on rating threshold.

Red flags to watch: A dashboard that caps visible verbatims at a fixed number per period. A CRM field that stores only the latest comment and overwrites prior ones. A text-analysis tool that summarizes or shortens comments. A warehouse table that drops rows with null open-text fields. A helpdesk rule that copies only ratings below a threshold, hiding passives.

Fixing location problems with concrete steps

These actions resolve most "where is it" failures within one reporting cycle.

Inventory the systems. List every tool that receives survey data: platform, CRM, warehouse, helpdesk, analytics. For each, record whether it holds score, verbatim, or both. Rebuild the score once from the raw export. Apply 9-10 promoters, 0-6 detractors, promoter share minus detractor share. Compare to the dashboard. A difference under one point is likely timing; larger differences indicate scope or de-duplication. Set one refresh cadence. Align warehouse batch time with reporting cutoff. Publish the cutoff time in the report header. Pull verbatims by response ID, not by date. Date filters miss late and edited responses. Centralize the quote source. Designate the export or warehouse as the canonical verbatim source. Use tags only for grouping. Reconcile weekly. Match warehouse counts against platform counts for both response rows and non-null text fields. Investigate any gap above a small threshold.

A practical target: if the warehouse and the platform disagree by more than 1 percent on response count for a closed period, treat the period as unreconciled until resolved. For score, a difference under 1 point on a 0-10-derived scale is usually timing; investigate differences of 2 points or more.

Limitations and what the numbers cannot tell you

NPS is a directional indicator with known weaknesses. Response rates vary widely by channel and audience, and small samples produce unstable scores. A handful of responses can swing the number several points. The score does not identify causes; only the verbatims and follow-up do. Combining streams with different survey questions invalidates comparison. And permissioning can hide verbatims from the very people who need them, which is a governance problem rather than a data problem.

The honest boundary: no single system can be assumed to hold a complete, current view of both the score and the text. You must verify each location against the raw export for your own stack. Prices and platform feature sets change, so confirm current capabilities with each vendor rather than relying on a guide. Where your tools differ from the patterns above, the raw export remains the one artifact you can always audit.

This article is general educational information about locating and reconciling survey metrics, not a diagnosis, treatment plan, legal opinion, or investment recommendation. It does not promise any business outcome. Consult a qualified professional for advice specific to your organization's data governance and compliance obligations.