Skip to main content

Free guide

Retention Metrics Starter Guide

12 behaviors that predict whether SaaS users renew. Sourced from 11 public churn teardowns.

Retention is not a metric. It is the residue of twelve observable user behaviors that either show up or go silent before a renewal date. The behaviors below are pulled from public churn teardowns of Notion, Linear, Cursor, Slack, HubSpot, Intercom, Asana, Monday, Evernote, Figma, and Calendly. None of these companies are doing badly. All of them lose customers because one or more of these twelve behaviors went silent before the renewal email landed.

The data behind the 12 behaviors

Across the 10 SaaS teardowns RetentionCheck has graded, the median Churn Health Score is 48/100 (D-tier). 9 of 10 scored below the healthy B-tier threshold. Only Linear cleared the band at 72 (B). The behaviors below are the ones whose absence shows up most consistently in those D and F grades. Browse the full teardown index →

01.First-week activation completion

Users who hit your activation milestone (whatever you call it: first project created, first invite sent, first data imported) inside week one renew at roughly 3-4x the rate of users who do not. The behavior is binary, observable in your product database, and shows up before the first invoice clears. If your free trial ends with no activation event, that user is gone, regardless of what the renewal CTA says.

Where it shows up in real teardowns

HubSpot teardown surfaced "had to onboard with a sales rep just to import contacts" as a top-tier complaint pattern. Activation gated behind a human is activation that does not happen.

What to measure

Track % of new accounts hitting the activation event by day 7. Below 30% means your onboarding is the leak, not your pricing.

02.Second-user invite within 30 days

A solo account is a churn risk. The second user added inside the first 30 days is the strongest single behavioral predictor of renewal in B2B SaaS. Once two people on a team rely on the tool, removing it requires consensus, not a one-click cancel. The Slack churn pattern is the inverse case study: nearly every public complaint includes "no one else on my team uses it anymore."

Where it shows up in real teardowns

Slack teardown logged 35+ complaints. The Sept 2025 Hack Club incident accelerated departures, but the underlying behavior was already there: shrinking active user counts inside paid workspaces.

What to measure

Track % of new paid accounts that invite a second user by day 30. Single-seat accounts at day 30 churn at 2-3x the multi-seat rate.

03.Integration connected to one tool in their stack

Customers who connect your product to even one external tool (Slack, Zapier, Stripe, their CRM, GitHub, anything) retain materially better than disconnected accounts. Integrations create lock-in not because they are expensive to unwind, but because they put your product inside someone else's daily workflow. A disconnected account lives alone on a tab nobody opens.

Where it shows up in real teardowns

HubSpot teardown: "doesn't talk to our existing CRM" was a recurring high-severity complaint, even though HubSpot is a CRM. The lesson is broader than HubSpot. If you do not show up where the customer already works, you are an extra browser tab waiting to be closed.

What to measure

Track % of paid accounts with at least one active integration. Compare 90-day retention for connected vs. disconnected cohorts.

04.Active session in the 14 days before renewal

Silent accounts churn at the next renewal. The cleanest leading indicator of churn is: did this account have a meaningful session (not a passive page view, an actual product interaction) in the two weeks before their next billing date? If the answer is no, the renewal is a coin flip. Reach out before the renewal email lands, not after the cancel email does.

Where it shows up in real teardowns

Linear earned a B grade in its teardown, but the failure mode named in 30+ public critiques was "team scaled past 50 seats and admin paused login for unused users to control cost." Silent seats get cut first.

What to measure

For every paid account, log days-since-last-meaningful-session. Trigger outreach when this crosses 14 days within 30 days of renewal.

05.Self-serve resolution vs. support ticket

Users who solve their own problems in your product retain higher than users who require a support ticket for routine tasks. This is not because support is bad. It is because every ticket is a moment where the user noticed friction strong enough to ask for help. Routine-task tickets are a leading indicator that the UX is below the user's threshold of patience.

Where it shows up in real teardowns

Intercom teardown: "AI support failed customers it billed extra for" appeared across 90 Reddit complaints. The first signal was rising ticket volumes for tasks the docs claimed were self-serve.

What to measure

Track ticket-volume-per-paid-seat by month. Rising volume on routine-task categories means the UX is breaking down, not that customers are needier.

06.Feature breadth (3+ distinct features used in week 1)

Users who touch three or more distinct features in their first week renew higher than users who touch one. The behavior reveals whether the customer is actually using the product or just sampling one capability. Single-feature users are competitors-in-waiting: they will leave for the next tool that does that one thing slightly better.

Where it shows up in real teardowns

Monday.com teardown logged the pattern as "only use it for one thing and that one thing is cheaper elsewhere." The same complaint appears across Asana, Notion, and HubSpot critiques. Narrow usage equals fragile retention.

What to measure

Track count of distinct feature-flag events per account in week one. Cohort renewal rates by week-one breadth (1, 2, 3, 4+ features used).

07.Pricing-page revisits during trial

When a trialing user keeps checking your pricing page, they are not deciding to buy. They are running cost math against value received. The behavior almost always precedes a downgrade conversation or a cancel. By the time they email you asking about a smaller plan, the decision is mostly made.

Where it shows up in real teardowns

Cursor teardown: "June 2025 pricing restructure broke existing-user economics." 40+ complaints surfaced the same pattern. Pre-cancel, users revisited the pricing page repeatedly trying to find a tier that worked. They did not find one.

What to measure

Track pricing-page visits per trialing account. More than three visits in week one is a flag, not a buying signal.

08.Comparison searches for migration paths mid-trial

When a user types "[your tool] vs [competitor]" or "migrate from [your tool] to X" into Google, they are not researching. They are reading the exit interview. Capture this via referral data, on-site search, or support-chat keywords. Migration-path mentions in support tickets predict cancel within 60 days roughly 70% of the time.

Where it shows up in real teardowns

Cursor teardown: Reddit threads pre-cancellation explicitly named the destinations (Claude Code and Windsurf) and the cost delta. The migration path was public before the churn event registered in Cursor's own analytics.

What to measure

Set up alerts in your support tool for keywords: "switching to", "moving from", "alternative to", and your top 3 competitors by name.

09.Recent positive support touchpoint

A resolved support ticket with a positive resolution inside the last 30 days is a renewal accelerator. The opposite is also true: an unresolved or escalated ticket sitting open for more than seven days is a renewal blocker. Support is not a cost center for B2B SaaS. It is a retention signal in disguise.

Where it shows up in real teardowns

Intercom teardown: "support never got back to me" appeared in 90+ public complaints. The migration to resolution-based pricing did not break the customer relationship by itself. The unresolved-ticket pile did.

What to measure

Track time-to-first-response and time-to-resolution by paid cohort. Flag any account with an open ticket > 7 days inside 60 days of renewal.

10.Annual vs. monthly commitment behavior

Annual plan customers churn lower not because they are locked in (they could let it auto-cancel) but because they made an explicit commitment when they signed up. The behavior of choosing annual at signup is itself a retention signal. Monthly customers churn at 3-5x the annual rate in most B2B SaaS data sets.

Where it shows up in real teardowns

Linear's B grade comes partly from annual-plan stickiness on Plus and Business tiers. Teams who renewed annual rarely showed up in the 30+ public complaint sample.

What to measure

Cohort retention by initial plan choice (annual vs. monthly). If your annual % is below 30% of paid accounts, your pricing page is not converting toward retention.

11.Budget signal in support tickets

When a customer asks about multi-year pricing, procurement processes, or whether you can issue an invoice to a new entity, they are folding your product into their next-year planning cycle. This is the most reliable late-stage renewal signal in B2B SaaS. The opposite signal is silence on commercial questions, paired with a sudden ticket about export options.

Where it shows up in real teardowns

HubSpot teardown surfaced "couldn't get budget approval after the 5x-20x cost increases" across the 2024 pricing-restructure cohort. Budget conversations are a retention surface, not a sales-team handoff.

What to measure

Tag support tickets for budget, procurement, multi-year, invoice, PO, vendor. Cross-reference with renewal date. Tagged accounts renew at 2-3x the rate of untagged silent accounts.

12.Engagement continuity through major product updates

When you ship a significant change (UI redesign, AI-feature rollout, pricing tier restructure, deprecation), some users adapt and stay. Others churn within 30-60 days. The behavior of returning to active use within two weeks of a major update predicts retention. Users who never log back in after a redesign are usually never coming back.

Where it shows up in real teardowns

Evernote teardown is the canonical case: "couldn't get used to the new app" appears across post-redesign cancellations. Notion teardown shows the same pattern for AI-feature rollout: power users who did not adopt AI within 30 days disproportionately churned.

What to measure

Before any major release, snapshot active user list. 30 days post-release, measure return-rate. Below 70% return signals the change broke retention, not improved it.

Now run it on your own product

Find which of these twelve behaviors are already going silent in your cancel reasons.

Paste your last 10 cancel reasons. In 30 seconds you'll get a Churn Health Score, the retention killers behind your churn, and one priority fix. Free.

Run RetentionCheck free

Or browse SaaS churn benchmarks and formulas →