This report joins two datasets: 4,621 support tickets and 12 months of retention data, matched on account. The join covers 3,940 accounts, and it answers a question neither dataset can answer alone — not "what do customers complain about" and not "who churned", but which complaints actually predict cancellation.

The answer is not the loudest one. Bug reports are the highest-volume complaint theme in the tickets, but bug-report accounts churn at 15.1% — barely above the 14.2% base rate. Customers who report bugs are, if anything, engaged. Billing-confusion accounts churn at 48% — 3.4x the base rate and the single strongest predictor in the join.

Two more findings change how you act on it. First, the risk is not the first billing ticket — it is the third: churn probability stays flat through tickets one and two, then jumps sharply. There is a window. Second, 31% of churned accounts filed zero tickets. They did not complain. They left quietly. Loud complaints are recoverable; silent ones are the harder problem.

What is working: bug-reporting customers stay. What is not: repeated billing confusion is a reliable countdown to cancellation, and almost a third of churn is happening with no warning in the ticket data at all.

Joining the two datasets reframes every complaint theme. Ranked by ticket volume (from the support data alone), bug reports lead. Ranked by churn lift (only visible once you join), billing confusion leads by a wide margin and bug reports barely register.

Bug reports: highest complaint volume, churn rate 15.1% — within a point of the base rate. A customer reporting a bug is a customer still using and still invested in the product. Annoying, not dangerous.

How-to questions churn at 11.8% — below base. Counter-intuitive until you read it: a customer asking how to do something is a customer trying to do more with the product. Curiosity is a retention signal, not a risk one.

Onboarding setup sits at 27.4% — roughly 2x base. Real risk, and it concentrates in the first 60 days. An account that struggles to get started often never does.

Billing confusion is the outlier: 48.2%, 3.4x base, and because the theme is also high-volume it accounts for 48% of all churn in the join. This is the rare theme that is both common and lethal. The chi-squared test confirms the complaint theme genuinely predicts churn (chi2=156.9, p<0.001, Cramer’s V=0.39 — a medium-to-strong effect). Which complaint an account files is not noise. It is one of the clearest churn signals you own.

Churn does not rise smoothly with billing tickets. It is close to flat through the first two — 19% after one billing ticket, 24% after two, both well within reach of a normal recovery. Then the third billing ticket lands and churn jumps to 51% — a 27-point step in a single ticket.

That step is the most actionable number in the report. It means a customer who has filed two billing tickets is not yet lost — they are confused and patient. A customer filing a third has stopped being patient. The first two tickets are the window; the third is the alarm.

This is exactly the kind of pattern that needs the join. The ticket data alone shows you a customer filing repeat billing tickets. The retention data alone shows you they later churned. Only the two together tell you the third ticket is the line — and that a same-day senior-agent escalation on ticket two or three is a concrete, cheap intervention with a measurable target.

The logistic regression confirms it (z=7.42, p<0.001, odds ratio 2.6 per additional billing ticket). Each repeat billing ticket multiplies the odds of churn — and the effect is steep enough that ticket count alone is a usable early-warning flag.

The survival curves run together for the first three months — billing-complaint accounts and clean accounts retain at near-identical rates. Then at month 4 the curves split and never reconverge. By month 12, clean accounts sit at 87% retained and billing-complaint accounts at 52%.

The gap matters less than when it opens. Most billing complaints in the data are filed in months 2–3. The cancellation shows up in months 4–6. The decision to leave is made one to three months before the cancellation is filed — during the window when the account is still paying and still reachable.

That is the practical takeaway of joining tickets to retention. A churn report alone tells you month 4 is bad. A ticket report alone tells you billing complaints cluster in month 2. Together they tell you that the month-2 billing complaint and the month-5 cancellation are the same event, two months apart — and that the intervention window is the gap between them.

The log-rank test confirms the curves are genuinely different (chi2=71.8, p<0.001, hazard ratio 3.1). A billing-complaint account is leaving at three times the rate of a clean account — and the data tells you exactly which month to act in.

The churn-by-theme chart ranked themes by how likely each one is to churn. This one asks the other half of the question: of every customer you actually lost, which theme did they come from? Rate and share are not the same number, and the recommendation depends on both.

Billing confusion is 48% of all churned accounts. It is the rare theme that scores high on both axes — a high churn rate (48.2%) and a high enough account count that it dominates the absolute total. Fixing it is not picking off a risky niche; it is addressing half of everything you lose.

Onboarding setup is 21% of churn despite being a smaller theme — its 2x churn rate is concentrated enough to matter. Together, billing and onboarding are 69% of churn from two themes.

The two low bars are the reassuring ones. Bug reports are 17% of churn only because the theme is large; per-account they churn at base rate. How-to questions are just 5% — almost no one who asks how to do something leaves. And the silent segment is 9% of the join — but that understates it, because the join only contains accounts that filed at least one ticket or were sampled as no-ticket. The honest reading: billing is the visible enemy, and the next chart shows you exactly when it strikes.

The survival curve showed billing-complaint accounts diverge at month 4. This chart shows the same story as a budget: where, in an account’s life, the billing churn actually gets spent.

It is front-loaded and then it stops. For billing-complaint accounts, 22% churn in months 0–3 and another 39% in months 4–6 — 61% of all billing churn inside the first two quarters. After that the rate collapses: 12% in months 7–9, 8% in months 10–12, settling toward the same low single digits that clean accounts hold the whole way through.

Clean accounts, by contrast, are flat — 3–5% in every band, no spike, no cliff. The danger is not "having a billing complaint" in the abstract. It is having one early, before the customer has built enough habit and switching cost to absorb the friction.

That sharpens the intervention to a date range, not just a ticket count. The second-billing-ticket escalation matters most for accounts inside their first six months — that is where the population at risk is concentrated. An account that has carried a billing complaint past month 7 and stayed is, statistically, already most of the way to safe. Spend the retention effort where the cliff is: early tenure, repeat billing contact.

The chi-squared test confirms tenure band genuinely predicts churn within the billing-complaint segment (chi2=58.3, p<0.001, Cramer’s V=0.31 — a medium effect). The first six months are not a hunch; they are where the risk measurably lives.

These are representative first tickets from accounts that later churned — not cherry-picked extremes.

Billing, first ticket (month 2): "I cannot work out why my invoice changed. Can someone explain what I am actually paying for?" Calm, reasonable, recoverable. This account churned in month 5. The complaint was a question; the silence after it was the problem.

Billing, third ticket (month 4): "This is the third time I have asked about my billing and I still do not have a clear answer. I am starting to look at alternatives." The alarm ticket. By the time an account writes this, the survival curve says it is more likely gone than not.

Bug report (stayed): "Found a glitch in the export — flagging so you can fix it. Otherwise loving the product." This is what a non-threatening complaint looks like. The account renewed.

Silent churn (no ticket, from the exit survey): "It was fine, I just never fully got it set up and stopped logging in." 31% of churn looks like this — no complaint, no signal in the ticket data, just a quiet fade. The hardest segment to catch, and the reason ticket data alone is not enough.

Three things, in priority order:

1. Escalate the second billing ticket, not the third. This is the highest-leverage move in the report. Churn is flat through tickets one and two, then jumps 27 points at the third. Auto-route any account filing a second billing ticket to a senior agent with a same-day "here is exactly what you pay and why" resolution. You are intervening inside the window, before the alarm ticket.

2. Fix billing clarity at the source. Billing confusion is 3.4x base churn and 48% of all churn in the join. Escalation treats the symptom; a clearer account page treats the cause. Plain-language tier summaries and line-item charge breakdowns stop the first billing ticket being filed at all. (This is the same fix the support-ticket theme report recommends — here the join shows it is also your single biggest retention lever.)

3. Build a non-ticket signal for the silent 31%. A third of churn never touches support. Ticket data cannot see them. Pair this analysis with a usage signal — logins, core-action frequency in the first 60 days — so the onboarding-stall accounts surface before they fade. The join is powerful, but it only sees customers who spoke.

Note what is not here: do not pour retention effort into bug-report accounts. They churn at base rate. They are talking to you because they are staying.

If billing clarity and second-ticket escalation stay unaddressed, the join projects forward cleanly: billing confusion keeps contributing roughly half of all churn, and the month-4 survival cliff repeats with every new cohort. The 14.2% base rate holds only because low-risk themes dilute it — the billing segment alone is bleeding at 48%.

If both interventions ship: escalating at the second ticket should pull a meaningful share of billing accounts back below the third-ticket cliff, and source-level clarity should shrink the billing-complaint segment itself. Modelling those two moves against the current join lowers blended 12-month churn from 14.2% toward roughly 11% — the difference between the billing segment churning at 48% and at something closer to 30%.

The silent 31% is the part this report cannot close on its own, and that is the honest limit of ticket data: it only sees customers who spoke. Pair this with a usage-based health signal and you cover both halves — the loud churn the tickets predict, and the quiet churn they never will. The whole retention thesis is simple: act on the second billing ticket, fix billing at the source, and build eyes for the customers who leave without a word.