How a B2B SaaS team cut guesswork from churn with AI interviews
A mid-market SaaS company was losing mid-tier accounts after month four. Cancel forms said “too expensive” or “missing features,” but CS could not prioritize saves or brief product on what to fix first. They ran Insight agent churn-risk interviews with CRM cohorts, edited auto-metrics in the analysis hub, and acted on explained drivers instead of checkbox labels. Free sessions proved the loop before they scaled credits for ongoing waves.
2 waves
CRM churn-risk interview cycles before the save playbook rewrite
>70%
Invite completion on mid-tier at-risk cohorts
48h
Typical lag from usage-drop alert to personalized Insight invite
1 hub
Shared analysis source for CS and product editable metrics
Challenge
Usage analytics showed a drop in integration connects before churn, but product and CS disagreed on whether pricing, docs, or competitor wins were the real driver. Static cancel surveys returned shallow labels that looked quantitative but explained nothing. Live interviews did not scale past a handful of accounts per quarter, so save plays stayed generic discounts instead of targeted help.
The story
The team’s cancel form had looked “data-driven” for years: price, missing features, switched to competitor, other. Every quarter the pie chart said price. CS still lost mid-tier logos after month four, and product roadmaps kept oscillating between discount experiments and feature parity bets. Analytics showed integration connects falling before churn, but nobody could prove whether that was cause, symptom, or coincidence.
They published a churn-risk Insight agent, imported CRM cohorts by plan and tenure, and triggered personalized invites when usage dropped. Blueprint fields forced structured capture during the conversation; the agent still probed for examples. After the first wave, CS edited an auto-metric so “couldn’t finish setup” stopped collapsing into “too expensive.” The second wave confirmed the pattern: setup friction led, price followed, and named competitors clustered.
With explained drivers in the hub, CS rewrote the save playbook around guided connects instead of blanket discounts. Product scheduled a docs and onboarding fix for the top friction path. The cancel form stayed for compliance; the Insight interview became the research channel for accounts worth understanding. No panel marketplace, no form-builder skin — owned-audience interviews feeding editable metrics.
Timeline
- 1
Discover
Mapped cancel-form labels against usage drops; cloned the churn-risk template and defined blueprint fields for risk reason, competitor, and save willingness.
- 2
Launch
Imported CRM groups, published branding, enabled optional voice, and wired API invites to usage-drop alerts for mid-tier accounts.
- 3
Learn
Compared themes by plan tier in the analysis hub; edited auto-metrics; ran a second wave focused on setup friction and competitor switches.
- 4
Act
Shipped a guided connect playbook, retargeted save offers, and kept weekly digests as the operating rhythm for CS and product.
Approach
- Step 1
Cloned the churn risk template and set session shape to deep dive for at-risk accounts flagged by CS or usage alerts.
- Step 2
Imported CRM groups by plan and tenure; personalized invites after usage-drop events via REST API so research landed in the same week as the risk signal.
- Step 3
Configured blueprint fields for primary risk reason, competitor named, willingness to stay with a save offer, and whether setup help would change the decision.
- Step 4
Enabled voice as an optional mode for busy buyers who preferred speaking over typing on mobile.
- Step 5
Reviewed analysis hub themes across plan tiers; CS edited auto-metric labels when “integration friction” needed splitting from “docs confusion.”
- Step 6
Shared weekly digests with study owners — top quotes, completion KPIs, and cohort compares — so product and CS used one source of truth.
- Step 7
Ran a second wave on mid-tier accounts only after the first themes stabilized, tightening follow-ups around setup and competitor switches.
Results
- Within two waves, “integration setup friction” overtook “price” as the top explained driver for mid-tier churn — completion stayed above 70% on CRM invites.
- CS launched a guided connect playbook; save offers shifted from blanket discounts to setup help within one sprint.
- Competitor-named mentions concentrated in two rivals, giving product a clearer win/loss brief than cancel-form checkboxes ever had.
- Organizers spent minutes reviewing the hub instead of coding transcripts; free-tier sessions proved the loop before upgrading credits.
- Weekly digest adoption replaced ad-hoc Slack paste of quotes; CS and product aligned on the same editable metrics.
- At-risk accounts invited within 48 hours of a usage-drop alert showed higher save-conversation rates than quarterly batch outreach.
FAQ
Did they still keep a cancel form?
Yes — a short compliance form remained. The Insight interview ran as a parallel or follow-up conversation for accounts CS wanted to understand deeply, not as a replacement for legal or billing checkboxes.
How long were sessions?
Most completed in six to eight minutes. Session shape kept probing focused on risk reasons rather than open-ended product tours, which helped completion stay above 70% on CRM invites.
Why not buy a research panel of churned users?
They needed answers from their own mid-tier accounts in the week risk appeared — not strangers paid to role-play churn. KogniFeed is built for owned-audience Insight agents, CRM cohorts, and API triggers.
Could they start without engineering?
Yes. The first wave used CRM bulk invites and free sessions. API automation came after the themes proved useful, so engineering time followed evidence instead of preceding it.