NPS vs CSAT vs CES for documentation feedback

Start with a page-level helpfulness question. Most documentation teams do not need NPS, CSAT, and CES at the same time. NPS measures loyalty to a company or product, CSAT measures satisfaction with a specific experience, and CES measures how much effort that experience required. The right choice depends on the decision you plan to make after the score moves.

Use CSAT or helpfulness for page-level decisions

CSAT is the closest fit when you need to decide which documentation page to revise. Ask whether the page was helpful or how satisfied the reader was with the answer, then keep the score attached to the page URL.

A binary helpfulness rate is easy to explain:

helpfulness rate = helpful responses / all rated responses × 100

The score tells you where to look, but the attached comment tells you what to change. A page at 62% helpful with repeated comments about a missing authentication example creates a clearer backlog item than a site-wide satisfaction score.

Use the visual feedback option when the problem refers to a specific code block, screenshot, or step on the page. The annotation adds context, but it does not make the rating itself more statistically representative.

Use CES when the documentation is part of a task

CES fits a workflow where the reader is trying to complete a defined task, such as creating an API key, configuring SSO, or finishing an integration. Ask how easy or difficult the task was after the relevant guide or flow.

Do not interpret CES as a general quality score for the whole documentation site. A reader can find one setup task difficult while still valuing the documentation overall. Keep the question close to the task and compare the same page or workflow after a change.

Use NPS for the relationship, not an individual page

NPS asks how likely someone is to recommend a company or product on a 0 to 10 scale. It is useful for tracking the broader customer relationship, but it is too far from a single documentation page to prioritize a rewrite by itself.

The standard calculation is:

NPS = percentage of promoters (9–10) - percentage of detractors (0–6)

If your company already runs NPS, keep it as a relationship metric and analyze documentation feedback separately. Do not claim that a docs change caused an NPS movement unless the study design isolates that change.

Pair every score with a reason

A score without a comment shows direction but rarely produces a useful fix. Collect a short optional reason after the rating and retain the page URL, timestamp, and relevant context with the response.

PushFeedback collects the rating and comment on the page the reader is using. Readers can optionally attach an annotated screenshot, and the dashboard lets the team compare feedback by page.

AI Reports can provide a first pass over larger feedback sets by suggesting themes, representative quotes, and possible actions. Those outputs are model-generated, so a person should check the grouping and recommendation against the underlying feedback before it becomes backlog work.

Measure change on the same surface

Compare the same question, page, and time window before and after a documentation change. Changing the survey wording, placement, or audience at the same time makes the comparison difficult to interpret.

A practical review cycle is:

1. Rank pages by negative feedback volume and rate.
2. Read the comments on the highest-volume pages.
3. Fix one recurring issue with a clear owner.
4. Compare the same page after enough new responses arrive.
5. Record what changed, even when the score did not improve.

This turns the metric into a feedback loop rather than a dashboard number. For the full workflow, see the guide to collecting and analyzing documentation feedback.

Frequently asked questions

Which metric should a documentation team start with?

Start with a page-level helpfulness question or CSAT-style rating because it maps directly to a page your team can revise. Add an optional comment so the score arrives with a reason.

Can documentation feedback replace NPS?

No. NPS measures the broader relationship with the company or product, while page feedback measures a specific content experience. Use each metric for the decision it can support.

Is a negative-feedback rate enough to prove documentation quality?

No. It is a directional signal affected by who responds, where the widget appears, and how the question is phrased. Combine it with comments, support themes, task completion evidence, and search behavior.

How much feedback is enough to act on?

There is no universal threshold. Prioritize repeated, high-severity issues and pages with enough traffic or business importance to justify the work, then keep collecting responses after the change.

Start with one page you can change

The free plan collects unlimited feedback on one website and keeps every text entry readable. See pricing, choose one page-level question, and review the first responses before adding another metric.