Latest / Customer Journey with Fexingo: Touchpoints, Mapping, and Multi-Channel Marketing / The Touchpoint Where Your NPS Score Lies to You
Transcript
- Lucas: So there's this moment that happens in almost every subscription business, and I bet most marketers don't realize it's actively distorting their most sacred customer metric. Luna: Which metric are we talking about? Lucas: Net Promoter Score. Specifically, the timing of when you ask that single question. I was looking at data from a mid-market SaaS company — about 50 million in ARR — and their NPS was sitting at 62, which is stellar. But when they sliced it by when the survey was triggered, something weird emerged. Luna: Okay I'm hooked. What did they find? Lucas: Virtually all their survey responses came within 24 hours of a support ticket being closed. Because their survey was automated to fire after any customer interaction. So what happened was: someone has a problem, they get it solved, they're relieved — and they give a 9 or 10. Meanwhile, their most loyal users, the ones who never contact support, were almost never surveyed. Luna: So the NPS was inflated by surveying only the just-rescued customers? Lucas: Exactly. They re-ran the survey with a random sample across the entire customer base — not tied to any touchpoint — and the real NPS was 47. That's a 15-point gap. And management had been making decisions based on that 62. Luna: Wow. So the happy path customers were invisible. How common is this? Lucas: Extremely. Most journey maps have a single survey touchpoint — usually post-purchase or post-support. But the customer's experience is a sequence of moments, and where you place that one question biases the answer. It's called 'touchpoint timing bias' — it's a well-documented problem in survey methodology, but almost nobody in marketing applies it. Luna: So the fix isn't just one survey at one point. It's sampling across multiple stages. Lucas: Right. And it's not about abandoning NPS — it's about recognizing that a single number from a single moment is a snapshot of that moment, not the relationship. If you only ask right after onboarding, you get enthusiasm. Right after a price increase, you get hostility. Neither is the truth. Luna: This is genuinely useful. If today was actually useful to you, the way these stay ad-free is listener support — buy me a coffee dot com slash fexingo. Lucas: Yeah, that's the only reason we can keep digging into stuff like this. Appreciate anyone who chips in. Luna: Okay, back to the data. So that SaaS company — did they redesign their survey approach? Lucas: They did. They implemented what I'd call 'journey-stage sampling'. They set up three triggers: one after a support interaction, one after a billing event, and one completely random each quarter for a cross-section. The random sample gave them the 47 baseline. The support-triggered sample stayed high at 60-plus. And the billing-triggered sample? That was 38. Luna: So they learned that billing is a pain point they weren't addressing. Lucas: Exactly. Before, they thought the billing experience was fine because nobody complained — but nobody complained because the survey only went to people who just had a good support interaction. The billing issues were silent. That's the danger of a single touchpoint. Luna: It's like only interviewing voters who just left a campaign rally and calling that public opinion. Lucas: Perfect analogy. And yet most companies do this. They have one survey tool, one trigger, one number on a dashboard. They don't realize the number is a function of the trigger, not the customer. Luna: So what should a marketer do tomorrow if they suspect this bias? Lucas: First, audit your survey trigger. Ask: when does this email go out? What event causes it? If the answer is 'after a support ticket closes' or 'after purchase', you have a biased sample. Second, run a one-time random sample — just send the survey to 500 random active customers with no trigger. See if the number changes. Luna: And if it does change by more than, say, 5 points? Lucas: Then you know you have timing bias. Third, you build a multi-trigger system. You don't need to survey everyone at every touchpoint — just enough to get a representative picture. For most B2B SaaS companies, 200 responses per segment per quarter is plenty. Luna: What about the survey design itself? Does the question wording matter here? Lucas: Absolutely. The classic 'How likely are you to recommend us?' is fine — but if you ask it right after a support interaction, people interpret 'recommend' as 'recommend our support team', not the product. So the context of the touchpoint leaks into the meaning of the question. Luna: So you're saying the survey isn't measuring loyalty, it's measuring satisfaction with the last touchpoint. Lucas: That's exactly right. NPS was originally designed as a relationship metric, not a transaction metric. But most implementations turn it into a transaction metric because of the trigger. The original article by Fred Reichheld in 2003 recommended surveying customers at regular intervals — quarterly or annually — not after every interaction. Luna: So we've drifted from the original methodology. Lucas: Completely. And the marketing automation platforms made it easy to trigger surveys based on events, so everyone did it. But the consequence is that your NPS becomes a measure of your support team's performance, not your brand's overall health. And then you make product decisions based on that. Luna: I've seen companies celebrate an NPS of 70 and then wonder why churn is still 5 percent a month. This might be why. Lucas: Exactly. High NPS, high churn — that's a classic sign of timing bias. The people who are about to churn rarely answer surveys, especially if the survey is post-support, because they've already disengaged. So your NPS is based on your happiest engaged users, while the silent churners are invisible. Luna: So it's not just a measurement problem — it's a strategic blind spot. Lucas: It is. And the fix is relatively simple technically, but it requires organizational will. You have to convince stakeholders that the shiny 62 NPS they've been bragging about is partly an artifact. That's the hard part. Luna: Do you have an example of a company that successfully made that transition? Lucas: Yeah, there's a B2B analytics company — about 200 employees — that had an NPS of 55 based on post-support surveys. They implemented a quarterly random sample and got 41. The CEO was initially upset, but they used the real number to prioritize a product simplification initiative. Within two quarters, the random-sample NPS climbed to 48, and churn dropped 15 percent. Luna: So the real number was lower, but they could actually act on it. Lucas: Exactly. A biased high number is worse than an accurate low number, because it breeds complacency. If you think you're at 62, you don't feel urgency to improve. If you know you're at 41, you act. Luna: So the key takeaway: your NPS might be lying to you, and the lie starts with the trigger. Lucas: That's the headline. And the solution is multi-touch sampling plus regular random surveys. It's not about getting one number — it's about getting a distribution across the journey. Luna: Alright, I'm going to go audit my own survey triggers now. Lucas: Good. And if you find a gap, you'll know exactly what to do.