Your NPS dropped and the score won’t tell you why, but your feedback will if you ask it in the right order. Start by checking the drop is real, then compare the two periods driver by driver and look at whether praise went quiet as well as whether complaints got loud. Once you know who is affected and where, you can decide whether the drop even deserves a response.
Key takeaways
- The score tells you something happened. The comments, the drivers and the segments tell you what
- Half of all ‘drops’ are a change in the sample, not a change in the experience. Check who answered before you check what they said
- Compare periods driver by driver. The topic that moved is the cause, and it is rarely the loudest one
- A score can fall because people stopped praising something, not because they started complaining
- A three-point drop among your most valuable customers and a three-point drop among first-time buyers are the same number and completely different decisions
The quarterly NPS lands. Down four. Red arrow. Someone has already put ‘NPS’ in the subject line of a meeting invite, and by Thursday there will be a working group.
The working group has a symptom, not a diagnosis. Everyone in the room will have a theory. Delivery, probably. Or the new app. Or ‘the market’. None of them will have read a single comment. Six steps get you from red arrow to root cause before the biscuits run out.
1. Check it’s a real drop
Before you investigate the experience, investigate the measurement. A surprising share of score movement has nothing to do with what happened to customers.
Who answered? NPS response rates are low, and the people who reply skew towards the very happy and the very cross. If the mix of respondents shifted, the score shifts with it. More new joiners than usual, fewer long-standing customers, a spike in one region: any of these will move the number with the experience unchanged. Compare the respondent profile this period against last.
Did you change how you asked? A new in-app prompt, a different trigger, a survey sent at a different moment in the journey. Any change in collection changes who you hear from and what mood they are in. Check the date of the drop against the change log.
Is it seasonal? Peak trading, holidays, a price change everyone already knew about. If the same dip happened this time last year, you may be looking at weather rather than climate.
Is the sample big enough to mean anything? A four-point move on 80 responses is noise. On 8,000 it is news. Know which you have before anyone books a room.
If the drop survives all four checks, it is real. Now you can look for the cause.
2. Compare the two periods driver by driver
Most investigations go wrong here. Teams look at the current period’s top complaints and assume those caused the drop. But the biggest topic is the biggest topic every quarter. Delivery is always in the top three. A topic that has always been there did not suddenly cause a four-point fall.
What you want is the change. Take every topic in your feedback and compare its impact on the score this period against last. Key driver analysis does this by weighting each topic by how many people mentioned it and how strongly they felt, then ranking them. Run it for both periods side by side and the movers stand out.
Delivery will still be the biggest negative driver, and it will have barely moved. ‘App login’ went from not appearing at all to the fourth-largest negative driver in one quarter. That is your cause. A small topic that appeared from nowhere, not a big topic that has always been there.
3. Check whether praise disappeared
A score falls for two reasons, and everyone only looks for one.
Complaints can rise. Or praise can fall. If your promoters used to mention ‘friendly staff’ in one comment out of five and now mention it in one out of fifteen, your score drops without a single new complaint. The thing that was quietly holding the number up has gone.
Run the same period comparison on the positive drivers. A positive topic shrinking is as much of a finding as a negative one growing, and it points to something you stopped doing rather than something you started getting wrong. It is also the finding most likely to be missed, because nobody opens a meeting invite titled ‘compliments are down’.
4. Find out who and where
A driver tells you what. Segmentation tells you who and where, and both change what you do next.
Slice the drop by journey stage. If the fall is concentrated at checkout and flat everywhere else, you have a checkout problem, not a brand problem. Slice by customer value. A three-point drop concentrated among first-time buyers is a conversion issue. The same three points among people who have been with you for five years is a retention emergency. Slice by channel, product line and region for the same reason.
Metadata attached to each response is what makes these cuts possible. A response that carries its store, product and tenure band can be sliced. One that doesn’t sends the working group back to guessing.
5. Decide whether to react at all
Not every drop deserves a response. That sentence goes down badly in CX teams, and it is still true.
A real drop, driven by a new topic, concentrated among your most valuable customers, needs action this week. A real drop driven by a one-off event that has already passed, among a low-value segment, might need a note in the report and nothing else. Reacting to everything is how CX teams spend a quarter chasing a dip that would have recovered on its own, while the slow leak nobody noticed carries on.
The test is the same as for any driver. Would fixing this move the score, for how many people, and what are they worth? If the answer is ‘not much, not many, not a lot’, log it and move on.
6. Get it to the person who can fix it
Once you know that ‘app login’ appeared from nowhere and is concentrated among mobile users who joined this year, the investigation is over. What remains is delivery.
The finding, with the comments attached, goes to whoever owns the app. Not in the quarterly deck. In the tool they work in, this week, with a case number. A diagnosis that stays in the CX team is a very well-informed way of doing nothing. It also makes you the person who knew and said so in a deck.
And the whole chain goes into the report leadership reads. The score, the driver that moved, who it affected, and what is being done. ‘NPS is down four’ is a red arrow. The same drop with a cause, an owner and a next step attached is CX intelligence.
Where the answer needs to live
None of this works if the comparison takes a fortnight. By the time a team has exported the comments, read them, and argued about the categories, the next period has started and the trail has gone cold.
The period comparison, the driver ranking and the segment view need to be on the same dashboard as the score, updated as the feedback arrives, and available to the people who are going to be asked the question. If your NPS sits in the built-in dashboard or in Power BI next to churn and revenue, the answer to ‘why did it drop?’ should be one click below the number. The same view lets you prove impact later: it found the cause, and it shows the score recovering once the cause is fixed.
Mistakes that turn a diagnosis into a quarter
Starting with the biggest topic
The biggest topic is the biggest topic every period. Look for the topic that changed.
Only reading detractor comments
Detractors are where people look first. But a drop can come from promoters becoming passives, which shows up as praise going quiet rather than complaints getting loud.
Treating the company-wide score as the unit of analysis
A single number hides everything. Segment before you form a hypothesis, let alone a working group.
Skipping the measurement check
Investigating a drop caused by a survey trigger change is a very thorough way to waste a month.
Forming a group before finding a cause
Working groups are for fixing things. Finding out what to fix is analysis, and it should take days, not a committee.
From red arrow to root cause
An NPS drop is a question, not a crisis. Rule out the false drops, compare the drivers between periods, check what stopped being praised, find who and where, decide whether it matters, and hand the answer to the person who can act.
Do that and the Thursday meeting is a five-minute update with a cause and an owner. Skip it and it is an hour of people looking at a red arrow and saying ‘delivery, probably’. Same meeting. Very different biscuits-to-decisions ratio.
That’s what SmartCX is built for: driver analysis that compares periods, segment views that show who and where, and dashboards where the answer sits one click below the score. See it in our product tour.
Frequently asked questions
Why did my NPS drop when nothing changed?
Check the measurement before the experience. A shift in who responded, a change to how or when the survey was sent, seasonality, or a small sample can all move the score with the experience unchanged. Then compare drivers between periods to see whether praise fell rather than complaints rising.
How much of an NPS drop is significant?
It depends on your response volume. A few points on a small sample is noise. On thousands of responses, a two or three point move is worth investigating. Watch for drops that persist across two periods and that concentrate in one segment or journey stage.
How do you find the cause of an NPS drop?
Compare the drivers of the score across the two periods and look for the topics whose impact changed, not the ones that are largest. Then segment the drop by journey stage, customer value and channel to find where it is concentrated, and read the comments behind the driver that moved.
Should you act on every NPS drop?
No. Act on real drops driven by a new or growing topic, concentrated among people who matter to the business. Log drops caused by one-off events or measurement changes and watch whether they persist.
What is the difference between a driver and a topic?
A topic is what people talk about. A driver is a topic weighted by how many mention it and how strongly they feel, ranked by its effect on the score. The biggest topic is not always the biggest driver, and the driver that moved between periods is the one behind the change.
SmartCX is the VoC platform that helps you turn feedback data into CX intelligence the whole business can act on. Watch the product tour or talk to us about finding the cause behind your next red arrow.