-
1
Pick a cadence that matches what you want to learn
The first decision is relational versus transactional NPS, and the two answer different questions. Relational NPS is sent on a fixed cadence, usually quarterly for B2B, to measure how customers feel about the overall relationship. Transactional NPS is triggered by an event such as an onboarding completion, a support ticket closing, or a renewal, and measures how a specific moment landed. Most B2B programs run both: a quarterly relational wave for the headline trend and transactional pulses to catch friction at the moments that matter. Bain and Company's original NPS research and later work from Rob Markey make the same point, that cadence is a strategic choice and running only one of the two leaves half the signal on the table.
- Relational: quarterly, sent to a defined contact role in every active account.
- Transactional: event-triggered after onboarding, support resolution, QBR, or renewal, sent only to the people who touched the event.
- Lock a sample policy up front: never survey the same contact twice in a 90-day window regardless of channel.
- Decide which cadence owns the published headline score so the team is not arguing about which number is the real one.
Tip: If you can only run one cadence in year one, start with relational quarterly. It gives you a trend line, forces the segmentation work, and is easier to defend to the executive team than a stream of event-triggered pulses.
-
2
Write the one question, the follow-up, and the invite copy
The NPS question itself is standardized: "On a scale of 0 to 10, how likely are you to recommend [company] to a friend or colleague?" Do not reword it, do not re-anchor the scale, and do not add qualifiers. The follow-up is where the real signal lives: a single open-ended question such as "what is the most important reason for your score" works better in B2B than a bank of structured follow-ups because it lets the customer tell you what they actually care about. Keep the full survey to two questions. The invite email should be short, from a named human on your team, name the two-question length, and promise a reply, not just "thanks for your feedback." Walker Information's B2B customer experience research consistently shows that invites signed by a real person and promising a real follow-up lift response rates meaningfully over anonymous template sends.
- Use the standard 0 to 10 scale with no label changes and the standard question wording.
- Add one open-ended follow-up: "what is the most important reason for your score."
- Keep the invite under 80 words, from a named human, and commit to a reply inside 72 hours.
- Add an optional second follow-up only for detractors: "what is the one thing we could change that would move your score the most."
Tip: Resist the urge to add a feature-level rating grid or a star bank. Every extra question drops response rate and dilutes the free-text comment, which is the field your product and success teams will actually read.
-
3
Segment who gets asked and when they get asked
In B2B, who you ask matters more than how you ask. The default should be the primary business contact in the account plus one power user, and the sample should rotate so you are not surveying the same two people quarter after quarter. Exclude accounts in their first 60 days because they do not yet have a stable view, and exclude accounts in an active escalation because a survey during a fire makes the score noise, not signal. Split the base into cohorts by ARR tier, product edition, and lifecycle stage, and run the sample so every cohort is proportionally represented. Gainsight's customer success research is clear that B2B NPS at the account level only becomes useful once you can read it by segment, because an aggregate score hides the fact that your enterprise promoters and your mid-market detractors are usually telling you two different stories.
- Include: primary business owner plus one power user per account, rotated across quarters.
- Exclude: accounts under 60 days old, accounts in active escalation, and contacts surveyed in the last 90 days.
- Stratify the sample by ARR tier, product edition, lifecycle stage, and region so every cohort is readable.
- Suppress contacts who have asked not to be surveyed and keep that suppression list under legal and privacy ownership.
Tip: Keep a per-account survey log inside the CRM, not just in the survey tool. If success and sales cannot see when a contact was last surveyed, they will send a transactional pulse the same week as the relational wave and the response rate collapses.
-
4
Launch through email and in-app with sampling that avoids fatigue
A good B2B NPS wave uses two channels: email as the primary send and an in-app banner for logged-in users who have not responded within five to seven days. Email gives you reach to decision makers who do not log into the product every week; in-app gives you the power users who ignore vendor email. Cap total surveys to one every 90 days per contact across both channels so you are not hitting the same person twice. Field the wave over a two-week window rather than a single day so you capture people who travel, and close the window on day 14 so the data does not drift. SaaStr's writing on B2B SaaS surveys and the Walker XMI research both land in the same place: a well-run two-channel wave inside a strict 90-day fatigue cap outperforms high-volume monthly sends on both response rate and comment quality.
- Primary send: email from a named sender on day 0, one reminder on day 5, final reminder on day 10.
- Secondary channel: in-app banner for logged-in non-responders starting day 7, dismissible, never modal.
- Fatigue cap: no contact receives any NPS survey more than once per 90 days, enforced across email and in-app.
- Field window: 14 days, then close the wave regardless of response volume so the data is comparable across quarters.
Tip: Never gate product features behind the survey and never pop a modal that blocks work. Both tactics inflate response rate for one wave and burn the base for the next three.
-
5
Close the loop with promoters, passives, and detractors inside 72 hours
Close-loop is where B2B NPS stops being a survey and starts being a program. Every response gets a reply within 72 hours, routed by score band to a specific owner. Detractors (0 to 6) get a call booked by the account's CSM or AE within the window, with an executive sponsor copied for enterprise accounts and a short written summary of what you heard back to the customer. Passives (7 to 8) get a short email from the CSM asking the one question that would move the score to a 9 or 10, which is often the richest research in the program. Promoters (9 to 10) get a thank-you plus a specific low-friction advocacy ask, such as a G2 review, a reference call, or a case study intro. Bain's close-loop research is unambiguous that the 72-hour window is the single biggest driver of whether customers answer the next wave honestly.
- Detractor: CSM or AE call booked within 72 hours; written summary of what you heard; action owner and date committed in the CRM.
- Passive: short email from CSM with the single question "what would it take to move you to a 9 or 10."
- Promoter: thank-you email plus one concrete advocacy ask (review, reference, case study, referral).
- Enterprise detractor: executive sponsor copied on the first touch and included on the call if the score was 0 to 4.
Tip: Track close-loop completion as a program KPI, not just response rate. If more than 10 percent of detractors are not contacted inside 72 hours, pause the next wave until the team catches up, because customers who were ignored on the last survey will not answer the next one.
-
6
Roll up responses by segment, CSM, and product area
A single headline NPS number hides almost everything the program is capable of telling you. Roll the data up along the cuts the business already runs on: ARR tier, product edition, lifecycle stage, region, primary CSM, and the product area the customer named in the free-text comment. Score each cohort separately so leadership can see that your enterprise segment is a 54 and your mid-market self-serve tier is a 12 rather than averaging to a comfortable 33 that nobody can act on. Theme-tag the comments against a stable taxonomy such as pricing, onboarding, performance, support, specific product areas, and account coverage, and track those theme counts quarter over quarter so you can see what is trending up and what is trending down. Gainsight's benchmarking work and Walker's XMI reporting both treat the segmented view as the real deliverable and the single number as a communication device.
- Cut by ARR tier, product edition, lifecycle stage, region, and primary CSM in every readout.
- Theme-tag free-text comments against a stable taxonomy of 8 to 12 codes, not a free-form list that grows every quarter.
- Track theme counts quarter over quarter so trending issues are visible before they show up in churn.
- Attribute each response back to the account record in the CRM so the data is joinable with usage, support, and renewal signals.
Tip: Pick the taxonomy once with product, success, and sales in the room, then freeze it for four quarters. Changing codes every wave makes trend analysis impossible and gives every executive a reason to argue with the numbers.
-
7
Publish the score with context, not as a single number
When you share NPS internally and externally, lead with the cohort view, not the headline. The quarterly readout should open with the score by segment, the top three theme movers up and the top three down, the close-loop completion rate, and the actions the team committed to last quarter with a status check on each. Reserve the aggregate number for the appendix, because the first time an executive sees a cohort view they understand immediately that enterprise and self-serve need different responses. If you publish NPS externally in marketing or on the site, publish the methodology alongside it (sample size, segments included, cadence, response rate) so a prospect can read it as evidence rather than a slogan. Harvard Business Review and Bain have both pushed the field away from vanity NPS headlines toward published methodology for exactly this reason.
- Lead the readout with segmented scores, theme movers, close-loop completion, and action status from last quarter.
- Put the aggregate score in the appendix with sample size, response rate, and segments included.
- Share the segmented view with every CSM and AE for their book so they can act on their own numbers, not an org-wide average.
- If published externally, include methodology so prospects can judge the number on its merits.
Tip: Never celebrate a rising headline number without the response-rate footnote. A score that went up because detractors stopped answering is worse than a score that went down for an honest reason.
-
8
Review the program quarterly and keep it honest
An NPS program decays if nobody reviews it. Hold a quarterly program review with product, customer success, sales, and marketing where you look at four things: did the sample stay balanced across segments, did close-loop hit the 72-hour service level, did the actions committed last quarter actually ship, and did the themes change in a way that implies the business responded. Retire themes that have been closed for two consecutive quarters, promote new themes once they cross a count threshold, and revisit the cadence and the fatigue cap if response rate has drifted more than 5 points in either direction. The goal is a program that gets more trustworthy every quarter, not one that produces a bigger number. Rob Markey's follow-up work on NPS as an operating system makes the same point, that the program's value is in the loop between question, answer, action, and the next question, and reviewing that loop quarterly is what keeps it alive.
- Review the four health metrics every quarter: sample balance, close-loop SLA, action completion, theme movement.
- Retire closed themes after two clean quarters and promote new themes once comment counts cross the agreed threshold.
- Revisit the cadence, the fatigue cap, and the sample policy annually and publish any changes before the next wave.
- Tie at least one action from each quarterly readout to a named owner with a visible date, and open the next readout with its status.
Tip: If three readouts in a row produce no committed actions, pause the program. A survey that never changes anything trains customers that answering was a waste of their time, and the response rate will tell you exactly that within two more waves.
