Playbook

Predictive CSAT & NPS Playbook: Spot Renewal Risk Earlier

Score satisfaction and loyalty on every conversation, roll the scores up by account, and see which renewals need attention while there’s still time to act

Why Surveys Miss the Customers You’re About to Lose

Predictive CSAT and NPS are AI-inferred satisfaction and loyalty scores built from every customer conversation instead of survey responses, which typically reach only 5–15% of customers. Joined to account, usage and renewal data, they help customer success spot at-risk accounts earlier.

The survey problem

Why does this tool never work the first or second or third time??????

A customer typed that into a support ticket. The ticket closed unresolved. AI scored the conversation a 1 out of 5.

The CSAT system’s record of this customer? Nothing. No survey was ever filed. On the satisfaction dashboard, and in any health score that leans on survey data, this person does not exist.

And this wasn’t an isolated case.

When we analyzed 500 conversations from our own support workspace, we found a striking disconnect: surveys showed 96% customer satisfaction, while AI analysis of the conversations put that figure closer to 75%. Not one conversation containing explicit churn-risk language produced a survey response.

That’s the danger of relying on surveys alone. You’re not just missing feedback. You’re missing the early warning signs of problems that could affect retention and revenue.

This playbook shows why surveys miss the customers most likely to leave, how predictive CSAT and NPS score every conversation instead, and how to turn those scores into an account-level renewal signal that customer success and CX can both act on.

What is Predictive CSAT?

Predictive CSAT is an AI-generated customer satisfaction score inferred from the full content of every customer conversation, rather than from post-interaction surveys. AI reads each transcript (sentiment, effort, resolution, how the tone changes) and assigns a satisfaction score to 100% of interactions, which removes the response-rate and response-bias problems of survey-based CSAT.

Predictive CSAT (pCSAT): A satisfaction score for every conversation, read from the transcript itself (sentiment, effort, resolution, escalation language) instead of a survey response. Scores exist per conversation, agent, account and segment, and update continuously.

Survey CSAT

Asks a fraction of customers to grade you after the fact

vs

Predictive CSAT

Grades every interaction from the evidence

What is Predictive NPS?

Predictive NPS is an AI-inferred loyalty score built from advocacy and churn signals inside customer conversations (praise, recommendations, repeat frustration, competitor mentions, cancellation threats) instead of the “how likely are you to recommend us?” survey. It applies the same conversation-scoring approach as predictive CSAT to relationship-level loyalty rather than single-interaction satisfaction.

Predictive NPS: A loyalty metric inferred from conversation data across a customer’s full history. Where survey NPS samples stated intent once or twice a year, predictive NPS reads expressed loyalty and churn risk continuously, across every touchpoint.

For a renewal, this is the signal that matters most, and it rarely shows up anywhere else. When a customer tells support they’ve “had to fight to keep this product around” and “can’t justify that fight anymore,” that’s the loudest detractor signal there is, and no NPS survey will ever capture it.

Why are CSAT and NPS surveys unreliable?

CSAT and NPS surveys are unreliable because they measure a small, self-selected sample of customers. Typical response rates run 5–15%, the happiest and angriest respond most, results arrive days after the interaction, and the score carries no explanation. The output is a biased, lagging, context-free number that’s easy to game.

We don't know if our customers are happy currently because we're sampling two percent of the interactions at best.
CX Leader

What is the average survey response rate?

Across survey types, average response rates run roughly 5–15%. Retently’s survey response-rate study puts NPS surveys at about 4.5%, CSAT at 9.8%, and CES at 22.5%. Email surveys typically land at 15–25%, and post-call surveys capture just 3–5% of interactions. Rates have declined for years.

Average survey response rates by survey type
Survey typeResponse rateSource
NPS surveys~4.5%Retently
CSAT surveys~9.8%Retently
CES surveys~22.5%Retently
Email surveys15–25%Clootrack
Post-call surveys3–5%Cresta

What is survey response bias?

Survey response bias is the distortion that occurs when customers who answer a survey differ systematically from those who don’t. The extremely happy and extremely angry respond most; the ambivalent middle stays silent. The resulting score describes your loudest customers, not your customer base, and it skews high, with CSAT scores clustering above 70%.

Nonresponse / self-selection bias: The error introduced when survey participation is voluntary and uneven. If customers who opt in differ from those who opt out (in mood, outcome or loyalty), the sample stops representing the population, no matter how many responses you collect.

Here’s the full breakdown behind the opening numbers. In an analysis of one B2B SaaS support workspace (Rippit's own), we ran AI review over a random sample of 500 support conversations and cross-tabbed transcripts against the actual survey records:

  • Happy customers were 4.4x more likely to answer the survey than frustrated ones. 36.9% of conversations with clear positive signals in the customer's own words produced a score; 8.3% of those with clear negative signals did.
  • The survey program caught 2.4% of genuinely dissatisfied customers. AI review found 82 conversations with genuine dissatisfaction. Only 8 ever produced a survey response, and 6 of those 8 still left a 4 or 5. Net: 2 of 82 dissatisfied customers registered as a low score.
  • The survey said 96% happy; the transcripts said 75%. Of 3,681 survey responses in the workspace, 95.9% were 4s or 5s (average 4.76/5, textbook ceiling effect). AI-predicted scores on the scoreable random sample: 7.8% at 1–2, 24.9% at 3 or below, roughly 4x more clear dissatisfaction than the survey record shows.

And the churn-risk finding from the opening: every conversation with explicit churn-risk language (cancellation talk, a move to a competitor) produced zero survey responses. The count is small, but the pattern held without exception. The customers most likely to leave were exactly the ones the survey never heard from.

(Across the whole workspace the survey response rate is 24.2%, which beats industry benchmarks, so these blind-spot numbers are conservative.)

Even respondents mislead. One customer wrote "I just deal with it now, its not ideal I just don’t know what else to do about it", and left a CSAT of 5 on that ticket.

Why are surveys lagging indicators?

Surveys report on interactions that are already over. By the time a bad score lands (if it lands), the ticket is days old, the aggregate report is weeks out, and the customer may be gone. The dissatisfaction was visible in the conversation long before any survey was sent, and long before anyone started preparing for the renewal.

We measured this too. Of the 151 conversations in the same workspace that received a 1–3 survey score, 88% contained clear dissatisfaction language an AI could have flagged from the conversation itself, without waiting for a survey. One customer (no survey ever answered): "ive gotten it for several days now and someone said they fixed it but its not actually fixed." That signal sat unread in a transcript.

88%

of tickets contained clear dissatisfaction language an AI could have flagged from the conversation itself

How bad is survey fatigue?

Bad, and worsening. Survey requests are up roughly 71% since 2020, a typical customer now fields about 12 survey requests a month, and around 70% of respondents abandon surveys partway, a pattern Koji's research corroborates. Every survey you send now carries a small negative CX cost of its own.

Survey fatigue: The declining willingness of customers to start or complete surveys as request volume grows, degrading both response rates and answer quality.

Read that back: the instrument you use to measure customer experience is making the customer experience worse.

How do teams game CSAT and NPS scores?

When a metric becomes a target, it stops measuring. Scores get gamed on both sides: agents pleading for 5s, the dealership-style "anything less than a 10 hurts me" speech, hard ticket types quietly excluded from survey triggers, sends timed for peak gratitude. WalkMe's survey critique catalogs the classics.

Honesty break: we searched our own 500-conversation sample for score-begging and found zero instances. But the gaming shows up one layer down, in the sampling. A customer told us on a call:

There are some bad actors out there that realize that and will apply that tag, knowing that QA is only reviewing less than 5% of their overall tickets.

When coverage is a sample, people hide in the unsampled 95%. At 100%, there's nowhere to hide.

What does a CSAT score actually tell you?

Almost nothing. A "2/5" carries no why, and the open-text box that's supposed to supply it goes more than 80% unanswered.

Sometimes the number is actively wrong. One customer in our dataset wrote "disappointing doesn't even cover it", and the survey they answered logged a middling 3. The transcript reads like a 1. Only one of the two can tell you what to fix, and which account it puts at risk.

The why was in the conversation the whole time. So why not just read the conversation?

Is NPS dead?

What Gartner predicted, and what actually happened

In 2021, Gartner predicted more than 75% of organizations would abandon NPS as a success measure for customer service and support by 2025.

It didn't quite happen, and we won't pretend it did. As CMSWire and Survicate chronicle, NPS survived in usage but lost influence: demoted from north star to one signal among many. And research covered by CustomerThink found ~52% of NPS respondents both promote and criticize the same brand: the single number was always a fiction of averaging.

The question isn't "do we kill NPS?" It's: why is a survey with a ~4.5% response rate still feeding your health score and your renewal forecast?

What are the alternatives to NPS surveys?

The commonly cited alternatives to NPS are other survey metrics, CSAT, Customer Effort Score (CES), retention rate, but they inherit the same response-rate and bias problems, because they're still surveys. The structural alternative is predictive scoring: predictive NPS and predictive CSAT infer loyalty and satisfaction from 100% of conversations, replacing a biased sample with a census.

The alternative to a broken survey isn't another survey. It's a signal on every conversation, tied to the account it came from.

How AI Measures Satisfaction Without Surveys

How does AI measure customer satisfaction without surveys?

AI measures customer satisfaction by reading the full transcript of every support ticket, chat, email, and call, then using large language models to score signals like sentiment trajectory, customer effort, resolution confirmation, and churn language. Every conversation receives a satisfaction score, no survey required, and scores roll up into continuous metrics by agent, team, and account.

The raw material: every conversation, next to the account it belongs to

Tickets, chats, emails, sales and success calls, joined to CRM, product usage and billing. AI reads every conversation; the account data says whose it is.

The reframe
Every conversation is an unsolicited survey the customer already filled out in their own words.

In our 500-conversation analysis, 59.2% of conversations contained a clear satisfaction or dissatisfaction verdict in the customer's own words, versus 24.0% of the same sample that produced a survey response. Customers "fill out the survey" in-conversation 2.5x more often than they answer the real one.

What the model actually reads

Signals a predictive CSAT model evaluates across each transcript:

Seven signals, split by what they predict
Satisfaction signals: how the conversation went (pCSAT)Loyalty and risk signals: where the relationship is heading (predictive NPS)
1 Sentiment trajectory: how tone moves from first message to last5 Escalation requests: "I want to talk to a human"
2 Customer effort language: "going in circles," "is a two day reply time the new normal?"6 Churn and cancellation language: competitor mentions, renewal threats
3 Resolution confirmation: did the customer confirm the fix, or go quiet?7 Explicit praise and advocacy
4 Repeat-contact frustration: the same issue, raised again

This is not keyword sentiment. As our team says: "We're not going to look just at whether they say 'very happy', it's going to analyze the full context of the conversation." You can even define satisfaction by what happened (not resolved on that call, an extra step needed), whatever words the customer used.

From conversation scores to an account-level signal

In Rippit, every conversation from your help desk or call platform (Intercom, Zendesk, Gong and other integrations, or any source through the API) gets an enrichment: the judgment you describe, here a predicted CSAT score and the driver behind it, applied to every conversation and saved as a column you can filter, count and chart. Your CRM, product usage and billing come in through your warehouse, the Bulk API or a CSV, so every score can carry the account, ARR and renewal date it belongs to.

You describe the score once, and it is answered across 100% of conversations. A pipeline keeps it running: it scores every new conversation as it arrives, rolls the scores up by account on a schedule, and sends the results where your team works. For a new question after the fact, add another enrichment and it runs across every conversation you select.

Per-conversation scores land in worksheets and roll up into charts by agent, team, product area and account. Instead of a quarterly survey aggregate, you get a satisfaction and loyalty signal that updates daily for every account, ready to sit next to its renewal date, usage and owner. (Full mechanics: why Rippit reads 100% of conversations.)

  1. 1

    Enrich

    AS EACH ARRIVES

    Each new conversation gets a pCSAT score, the driver behind it and any other judgment you define.

  2. 2

    Score

    ACROSS 100%

    A pipeline runs the same enrichment on every conversation, including the ones that were never surveyed.

  3. 3

    Roll up

    BY ACCOUNT

    Scores roll up by agent, team, product area and account on a schedule, in worksheets and charts.

From Predictive CSAT to Renewal Risk

How does a predicted score become a renewal signal?

A health score usually weighs usage, seats, payments and the last NPS response. All of those can look fine while an account’s tickets and CSM calls say otherwise, so the health score stays green until it doesn’t. The gap is common: over half of customer success professionals work with fragmented data (Customer Success Collective, State of Customer Success 2026).

A conversation score on its own is a CX metric. It becomes a renewal signal in five steps:

  1. 1

    Score every conversation

    pCSAT from 1 to 5 with the driver behind it, plus predictive NPS signals such as detractor or cancellation language.

  2. 2

    Roll up by account

    Average pCSAT, the lowest scores, the share of low scores and detractor mentions across tickets, CSM calls and shared Slack.

  3. 3

    Trend it

    Compare the last 30 days with the 90 before. Direction matters more than level: an account sliding from 4.6 to 3.9 says more than one that has sat at 3.8 all year.

  4. 4

    Join it to the account

    Renewal date, ARR, active users and seats, survey NPS where you have it, and whether champions and executives are still engaged.

  5. 5

    Flag the change and route it

    When risk moves, the CSM gets the account, the reasons and the quotes. CX gets the driver to fix.

A few rules keep the signal honest:

  • Flag on change, not on level. Some accounts always run lower. Alert when an account’s risk moves, so the weekly list stays short and new.
  • Treat silence as a signal of its own. Conversations marked “not enough info” don’t count toward the average. An account that stopped talking to you while logins fall needs the usage data to carry the flag.
  • Weight by account, not by volume. One account that files 40 tickets shouldn’t drown out a quiet enterprise customer with three bad CSM calls.
  • Score bots and humans separately, then combine. Checkr built separate models for human-agent and chatbot conversations and combined them into one signal.
  • Backtest before you trust it. Run the rules against accounts that churned and renewed last year, then tune the thresholds (how to build a churn radar).

Who acts on the signal?

The same score serves different owners. CX owns the driver; customer success owns the account:

Who acts on a renewal signal built from predicted scores
OwnerWhat they getWhat they do next
CSM or account ownerAccounts whose risk changed, with the reasons, a customer quote and links to the conversationsBook an exec check-in, loop in support, add a next step to the renewal
CS leadershipA roll-up of renewals that changed risk, with the ARR at stakeAssign exec sponsors and adjust the renewal forecast
CX or support leaderLow-score drivers by channel, team and AI agentFix the driver: a billing error, a slow handoff, a wrong help article
CS operationsAccounts where the health score and the predicted scores disagreeReweight the health score so it stops staying green too long

What does good output look like for one account?

Here’s one week of output for four accounts.

Predicted scores joined to account data, one row per account
AccountPredicted scores, last 60 daysHealth scoreAccount dataRenewal risk
Mid-market · $120K ARR · renews in 61 dayspCSAT 4.3 → 3.4; detractor language on 2 CSM callsGreenSeats −40%; pushback on priceMedium → High
Mid-market · $65K ARR · renews in 150 dayspCSAT 3.8 → 4.5; praise for SSOGreenSeats +12; hiring in CSMedium → Low
SMB · $18K ARR · renews in 33 daysNot enough info: no conversations in 60 daysGreenLogins −60%Low → Medium
Enterprise · $310K ARR · renews in 200 dayspCSAT steady at 4.2YellowUsage dip during a holiday code freezeLow, no change

Two of the four would have surprised a team watching only the health score: the first account is green and slipping, and the last is yellow and fine. Here’s what the first one looks like when it lands with the CSM:

Health score

Green

Logins within range, invoices paid, last survey NPS an 8 from the champion in the spring.

Nothing to look at.

Renewal signal · Mid-market · $120K ARR · renews in 61 days

Renewal risk moved from Medium to High

High risk

pCSAT fell from 4.3 to 3.4 across 14 conversations in 60 days. Main driver: invoice errors, raised in 5 tickets. Detractor language on 2 CSM calls. Seats down 40% since Q2. The renewal opportunity has no next step.

“We’re paying for seats we don’t use, and the invoices are still wrong.” Finance lead, CSM call

Next step: book an exec check-in this week and ask support to fix the invoice issue first. Sent to the CSM in the customer’s Slack channel; the billing-error driver goes to the support lead.

What makes it good: it names one account, says what moved and why, quotes the customer, names an owner and a next step, and links every claim to the conversations and records behind it. A dashboard tile that says “pCSAT down 0.3” does none of that.

Which agents run on the same scores?

Agents to run on predicted scores
AgentWhat it addsRunsSends to
Predictive CSATpCSAT on every conversation, with a daily list of low scores and the driver behind eachEvery morningA pCSAT sheet in Google Sheets or Notion; the low scores in Slack or email; driver reports to Notion or Google Drive, or a Linear issue, over MCP
Renewal Risk WatchEvery renewal whose risk moved, with the reason, the evidence and a next stepMondaysA Slack message tagging the CSM in the customer’s channel; a leadership roll-up in Slack or email; a CRM update through a custom MCP connection
Expansion RadarThe other side of the same signals: seat, new-team and new-use-case requests from accounts ready to growWeeklyA Slack note to the account owner; a CRM opportunity through a custom MCP connection
Agent Coaching BriefFor each rep, the habits behind low scores, with example conversations, and a 1:1 brief for the team leadMondaysSlack DM to each team lead; full briefs in Notion, Confluence or Google Drive

Customer Voice

How CX teams describe the problem in their own words

We don't have to guess how teams measure satisfaction today. We're on sales and CS calls with CX, QA, and support leaders every week, so we pulled a random sample of 700 of those calls, ran AI enrichment across every transcript, and let the market speak for itself. The single most common pain, by a wide margin: coverage. Here's what it sounds like.

The coverage ceiling. Every team hits the same wall: humans can't read everything.

There's no way to manually review those. I mean, if we're lucky, we review a couple hundred, I mean, literally one percent.
Operations leader at a payments company handling ~40,000 tickets a month
We're currently only able to audit about anywhere from two to five percent of our ticket volume. Our co-founder really wants to be able to push that number up to 75 percent of our tickets, which currently, given the way things are working manually, is just impossible…
CS manager at an e-commerce brand running 100,000+ conversations a week

The survey black hole. When leaders describe their survey programs, they describe what's missing.

January through the end of July, only 25 percent of the conversations that we have with customers had that CSAT rating. So there's 75 percent that we're not getting customer feedback on.
Chief Customer Officer, managed hosting provider
There's 80 percent of tickets on average that do not receive a survey, either because we may not have one attached to that particular issue type, or players decide not to give us a survey, or it times out. We're finding that there could be value in those 80 percent.
Quality manager, mobile gaming

On another call that same quality manager put it more bluntly:

… We have a black hole of tickets, that are about 80 percent of tickets, that are not even looked at.
Quality manager, mobile gaming
One is to ask customers: how was it? Which we actually turned off, because our customers literally hated it. We got more messages saying 'please stop asking' than actual responses
Director of customer support, B2B SaaS
We had less than five percent, I think it was less than three percent, of people who volunteered to take a phone survey. They just call back after and ask to speak with a manager if they had a terrible experience.
CX leader, e-commerce retail

The score–reality gap. Sometimes the survey number and the transcripts tell opposite stories.

Our CSAT scores that we measure are good. And I don't understand, to be honest with you, why they're so great. When you listen to our phone calls, some of our agents, not all of our agents, are just… they're rude. They're terrible.
Quality leader, insurance

Satisfaction without asking. The most sophisticated teams have already named the destination.

The critical piece of this project involves automating sentiment analysis for the care organization. So not having to rely on email surveys, or even IVR 'did you get your questions answered' kind of things, but just listening to the call and understanding what the customer felt.
Customer insights lead at a telecom with ~110,000 care interactions a month
… It seems like the general trend is away from those surveys and more towards this, like, predictive CSAT, because it's the biggest blind spot.
CX leader, healthcare & wellness (their survey program: 15–20% response rates, on only 75–80% of conversations)

Churn flagged too late. Loyalty signals live in conversations; teams find them after the customer decides.

Churn risk, by the time we flag it, most of the time, we're reactive. And so we're trying to go back through and analyze data to the point where we can understand the earliest indicator of churn, and then sort of change the direction of that inevitable path.
VP of enterprise customer experience, web infrastructure
We track our retention loss week over week… And so for our support team specifically, we have done engagement calls, retention calls, tried to figure out: who is maybe on the verge? And getting to them proactively, before they've already made up their mind and they're set to leave.
CS leader, virtual healthcare

The next blind spot: AI agents. The newest use case in our calls isn't measuring humans at all.

For the next six months, it's our main priority for the business, because we're moving all of the games in customer support for AI. And we have a lot of executives asking: why does the level of transfer to agent keep increasing, slash, decreasing?
Support leader, mobile gaming

Read those again. Not one of these leaders asked for a better survey.

Every use case (coverage, blind spots, churn, contact drivers, AI oversight) is the same underlying ask: read all of our conversations and tell us what's actually happening.

That's the job surveys were hired to do and never could. Predictive CSAT does the reading; joining each score to the account is what tells you who is about to leave, and why.

Making the Switch

What's the difference between CSAT and Predictive CSAT?

Survey CSAT asks the few customers who respond to grade an interaction after the fact. Predictive CSAT scores every conversation from the transcript itself. The biggest difference is coverage: 5–15% of interactions against 100% of conversations. The table compares the two on all seven dimensions.

DimensionSurvey CSATPredictive CSAT
Coverage5–15% of interactions, self-selected100% of conversations
LatencyDays to weeks after the interactionScored as conversations close
BiasExtremes over-represented (happy customers 4.4x likelier to respond, per our data)A census: the silent middle is counted
ContextA number; the open-text box is ~80% emptyThe full transcript and the reasoning behind every score
Fatigue costEvery send erodes goodwillZero asks; customers never see it
GameabilityHigh: begging, timing, trigger exclusionsLow: the model reads everything
Trend granularityWeekly or quarterly aggregatesDaily, by agent, team, account and product area
Survey CSAT vs. predictive CSAT on seven dimensions.

A survey metric can be precise and still wrong: it precisely measures the wrong population. Sampling gives you an answer. It doesn't give you the right answer.

How accurate is predictive CSAT?

Industry benchmarks show predictive CSAT models agreeing with actual survey scores 80–90% of the time. At Checkr, predicted satisfaction matches survey results about 85% of the time.

In our own validation on one B2B SaaS support workspace (Rippit's own), AI-predicted scores matched survey scores exactly 76.7% of the time and within one point 93.3% of the time — and where they diverged, the survey was the flattering record, not the honest one.

76.7%

AI-predicted scores matched survey scores exactly

93.3%

matched within one point of the survey score

Sit with that: in all 8 cases where the two diverged by 2+ points, the survey score was higher than the transcript supported. Respondents are politer than their transcripts. The model wasn't wrong; the survey was.

The deeper reframe: a slightly noisy score on 100% of conversations beats a "precise" score on a biased 5% sample. And full-context AI catches what keyword sentiment misses: the customer saying "fine" through gritted teeth.

Now the honest limitations, because candor beats hype:

  • Some conversations carry no signal. As our own team puts it: "In some cases there are going to be no pointers within the conversation. So what do we anchor it on then?" In our sample, 38 of 500 conversations were too short to score. A good system says "not enough info" instead of guessing.
  • Sarcasm and terse exchanges remain harder than plain frustration.
  • Calibration takes a few weeks, validating against a human-audited sample and historical surveys before trusting the trend line.
  • Our validation set skews happy, because it inherits survey-respondent bias; agreement on angry customers is undertested.

Should you stop sending CSAT surveys?

Not immediately. Run predictive CSAT alongside your surveys for a quarter, correlate the two, then demote surveys to periodic calibration checks and relationship pulses while predictive scores become your operational metric. Surveys still capture direct, volunteered customer voice; they just shouldn't be the system of record for satisfaction.

The migration

  1. Months 1–3

    Score 100% of conversations in parallel and investigate divergences (in ours, divergence usually meant the survey was too generous).

  2. Month 4

    Move operational decisions (QA, coaching, at-risk alerts) to predictive scores, add the account roll-up to your health score and renewal reviews, and report the predictive metric as the headline number.

That's not theoretical. It's what Checkr's Director of Shared Services described after the switch:

We still survey our customers, but we don't actually look at that. We look at this one hundred percent. And we report it all the way up to the C-suite.
Director of Shared Services, Checkr

What this looks like in practice

One predictive score runs on 100% of conversations. Support leaders use it to fix what’s driving low scores; sales and CS leaders use it to see which accounts are at risk.

What predictive CSAT on 100% of conversations means for support leaders and for sales and CS leaders
One score, two teams: support leaders and sales and CS leaders.

Mini case: Checkr's 100% vs. 7%

Checkr built a custom predictive CSAT (pCSAT) program in Rippit analyzing sentiment across 100% of conversations, versus the 7% their CSAT survey actually covered, with separate models for human-agent and chatbot conversations combined into one unified signal, reported to the COO and CEO.

Checkr case: predictive CSAT in Rippit covers 100% of conversations versus a 7% CSAT survey response rate

14×

more coverage than the survey saw

The Payoff

Speed to action, not just coverage. When about 600 billing tickets showed a 47% predicted satisfaction score, the finding reached the Chief Product Officer, with customer quotes, within a week.

Within less than a week, we had new information to people, and people are moving on it. That would have taken months before.
Kristin Johnson, Director of Shared Services, Checkr

Brex

Finds churn risk in 2 to 3% of tickets each week and reaches those customers within hours, often before they could answer a survey.

Klaviyo

Was sitting on 2.5 million conversations it had never really looked at. Now it reads a month of launch feedback in about two hours.

As one customer put it:

We wanted to focus on 100 percent of our contacts, which of course, with CSAT/NPS survey data, you're not able to do.

What is Rippit?

Rippit is a Growth Data Platform: anyone can build AI agents on all of their customer and business data, with no engineer needed. Connect your data in one click or through our MCP, add your own research, and describe what you want in plain language.

Rippit reads every row and runs as an agent, sending findings to Slack, email, Notion or Drive on a schedule. Predictive CSAT is one application: the same scored conversations, joined to CRM, usage and billing, give CX the drivers to fix, customer success the renewals to protect and Product the issues behind both, without a separate system for each.

Support & CSM conversations (tickets, CSM calls)Product adoption & NPS (usage, seats, surveys)CRM & renewalsStakeholder engagement (champions, execs)and more

FAQ

What data does predictive CSAT need?

Your existing conversation data: tickets, chats, emails, and call transcripts. No new instrumentation and no surveys; Rippit connects to Intercom, Zendesk, Gong and other integrations, or any source through the API, and scores each conversation as it arrives. To turn the scores into a renewal signal, connect your CRM, product usage and billing data too, so every score carries its account, ARR and renewal date.

Does it work on voice calls?

Yes. Call transcripts from Gong, Attention, Fireflies or Granola, or any call platform through the API, are scored the same way as tickets and chats.

How is predictive CSAT different from sentiment analysis?

Sentiment analysis labels emotional tone, often from keywords. Predictive CSAT is a full-context judgment: an LLM weighs sentiment plus effort, resolution, escalation, and repeat-contact signals across the whole transcript, and outputs a calibrated score with reasoning attached.

Can it replace NPS for board reporting?

Increasingly, yes. Checkr, for example, reports pCSAT to the C-suite. Run both for a quarter, show the correlation, then present the predictive metric as the headline number with survey NPS as a periodic pulse.

How long until scores are reliable?

Most teams validate within a few weeks: score a backlog, human-audit a sample, correlate against historical surveys. Our own validation hit 93% agreement (±1 point).

Can predictive CSAT predict renewal risk?

Not on its own: it predicts satisfaction, one conversation at a time. Rolled up by account, trended and joined to renewal dates, usage and champion activity, it becomes one of the earliest renewal-risk signals you have, because customers usually say something is wrong before usage or the health score moves. Backtest it against accounts that churned and renewed before you trust it.

Can’t I just use Claude or ChatGPT for this?

Keep them as the brain. On their own, they read a small slice of your conversations, start over each session and get expensive at volume, which rules out a daily score on every conversation. A Growth Data Platform reads every row, saves each score as a field on the conversation and the account, and lets the assistant query them over MCP.

Conclusion

The survey era measured whoever felt like talking

Survey-based CSAT and NPS measure the small, self-selected slice of customers who feel like answering, late, without context, skewed to the extremes. Predictive CSAT and NPS measure every scored conversation, not just the customers who answer a survey, with the reasoning attached. Joined to the account, they tell customer success which renewals need attention while there’s still time to act, and tell CX which drivers to fix.

The customer who asked why the tool never works the first time never answered a survey. Neither did the one who gave up the fight to keep the product. Their verdicts were sitting in your conversation data the whole time.

Stop asking. Start reading.

Here’s an agent that flags those renewals every week.

Agent · runs every Monday

Renewal Risk Watch

Catch at-risk renewals months before the renewal call. Every Monday it joins support conversations and sales and CS calls to CRM, usage and billing, one row per account, and sends:

  • Every renewal whose risk changed, such as Low to High, with the reasons: usage down, escalations piling up, a champion gone, no next step on the renewal
  • A customer quote and links to the tickets, calls and usage behind each flag
  • A suggested next step, such as an exec check-in before the QBR
  • A leadership roll-up of the renewals that changed risk and the ARR at stake
Sends to
Slack, tagging the CSM in the customer’s channel once Rippit is invitedLeadership roll-upCRM risk field, through a custom MCP connectionTasks over MCP

Build your first agent on your customer and business data.

Try Rippit free
or request demo

Sources

  • Customer Success Collective, State of Customer Success 2026
Share this post

Keep reading

More from the Rippit Library.

View the library →

Stop asking, start reading. Uncover true customer sentiment.

Replace biased 5% survey samples with a satisfaction score on every conversation.

Try Rippit free