Start with the number, then read the words. The score tells you direction; the open-ended comments tell you why.
Don't stop at the metric itself, focus on what respondents actually wrote, since that's where the fixable detail lives.
Segment before you conclude anything
A single blended average hides more than it shows. Break results down by:
User tenure (new vs. established)
Customer type (free vs. paid)
Feature area
Account or company
A dip in your overall NPS might really be one large account having a bad week, not a product-wide problem, and segmenting is the only way to tell the difference.
Decide what's actionable
Recurring themes across multiple respondents: treat as a systemic issue and put it on the sprint planning agenda.
Low scores from key accounts: treat as a churn risk and follow up personally, quickly.
Score drops right after a release: investigate for a bug or regression before assuming it's a design complaint.
Isolated one-off complaints: log them and watch for a pattern, but don't act on a single data point.
Weight independent mentions of the same issue from different respondents more heavily than one person mentioning it loudly.
Close the loop
Acknowledge respondents where you can, and feed the top themes into your actual planning process. A survey that generates insight nobody acts on doesn't move any of your metrics next quarter.
What's next