How to Find Competitor Churn Rate: What You Can and Cannot Know

Search this question and almost everything returned explains how to calculate your own churn rate, or lists industry benchmarks. That is not an accident of poor writing. It is because the question as asked has no answer: churn is computed from a customer list that exists in one company’s billing system, no regulator requires it, and no panel can observe it.

That makes this page different from the rest of this cluster. Everywhere else the job is finding a figure that exists somewhere. Here the job is building a defensible direction out of signals, and being honest about what each one is worth. The good news is that a direction is usually what the decision needed anyway: nobody changes a roadmap because a rival churns at 11% rather than 8%, but plenty of teams should change something when a rival’s customers start shopping.

1. 1Accept that you are estimating a direction, not finding a number. There is no external source for a competitor's churn rate, and any figure presented as one was modelled. Decide up front that the output is rising, flat or falling with evidence attached, and the research becomes tractable.
2. 2Take the disclosed retention metric, then read what it hides. If they are listed, find net revenue retention in the filings and note the exact definition they use. Then remember it nets expansion against losses, so it can look excellent while customers are leaving in numbers.
3. 3Count observable departures. Technology removal detection names sites that dropped the product. Archived customer pages show logos that disappeared. Neither gives you a rate, because you have no denominator, but both give you named, checkable departures.
4. 4Read the cancellation language, not the star rating. Search reviews, forums and communities for the words leavers use: switched, migrated, cancelled, moved off, renewal. The reasons cluster quickly, and the clusters are more useful than any score.
5. 5Watch what they do about retention. A first retention hire, a win-back campaign, a move to annual-only billing or a sudden loyalty discount are all a company responding to churn it can see and you cannot. Behaviour beats disclosure.
6. 6Measure the one flow you can actually count. Track how many of their customers reach your pipeline, from which segment, and why they left. That is measured, current, first-party and directly relevant, and it is the only churn-shaped number you can fully verify.
7. 7Report the direction with its evidence and re-check quarterly. Write the conclusion as a direction with the named departures and dated signals behind it, never as a percentage. Re-check each quarter and immediately after a pricing change, an outage or a leadership change in their customer organisation.

Listed software companies do publish a retention metric, and reading it correctly is the single most valuable skill on this page. The usual disclosure is net revenue retention: the revenue still coming from a cohort of customers a year later, after upgrades, downgrades and cancellations are all netted together.

That netting is the whole problem. Expansion inside the surviving accounts is added to the same number that losses are subtracted from, so a vendor whose remaining customers keep buying more seats can report a healthy figure while a meaningful share of its logos walks out of the door. Gross revenue retention removes the expansion and is far more revealing, which is precisely why it appears less often.

Three things to take from that table, none of which is a benchmark. First, every company names the metric slightly differently and defines it in its own filing, so you must read the definition before comparing anything to anything. Second, the range spans businesses expanding strongly inside their base and businesses contracting inside it: a figure below 100% means the cohort is worth less than it was a year ago, which is a serious signal. Third, and most importantly, not one of these is a churn rate. They are all compatible with a wide range of underlying customer losses.

You will never get a rate. You can, surprisingly often, get names. A named departure is worth more than an estimated percentage anyway, because it is checkable, it is actionable by a salesperson, and it comes with a story you can go and ask about.

Technology removal detection

For any product that leaves a trace in a public web page, detection services record not only which sites added it but which sites dropped it. That is an observation rather than a model: a script that was present in March and absent in September is a fact about a specific company. Coverage is limited to the sites these services crawl and to browser-detectable products, so it says nothing about back-office software. Within its scope it is the highest-quality churn evidence available from outside.

Logos that disappeared from the customer page

Comparing a competitor’s customer page against archived versions produces a shortlist of companies that were displayed as customers and no longer are. The technique, and its considerable caveats, is set out in the guide to finding a competitor’s customers. What that page does not cover, and what matters here, is the denominator problem: a list of departures without a customer count is not a rate, and never becomes one. Six vanished logos means something very different for a vendor with sixty customers than for one with six thousand.

Leavers who wrote it down

People announce switching in public far more than they realise: review text, community threads, migration write-ups, conference talks about a replatforming project. Search for the vocabulary of leaving rather than for the product name alone. The reasons cluster fast, and a cluster is more decision-useful than a count.

1. Filed retention metrics

Search the competitor’s annual report for retention, renewal and cohort. Record the figure, the exact metric name and the definition given in the filing, because those definitions differ between companies and sometimes between years at the same company. A definition that quietly changes between filings is itself a finding, and it is the kind of thing an investor relations team hopes nobody reads closely.

2. Earnings calls and investor materials

Transcripts are where retention gets discussed in words rather than numbers, and analysts ask the questions the filing avoided. Listen for hedging: a shift from citing a specific figure to describing retention as stable, a new emphasis on a particular segment’s renewals, or a reference to elongated sales cycles at renewal. Management rarely lies on these calls and frequently changes the subject, and the change of subject is the signal.

3. Technology removal detection

Covered above. Use it in trend form rather than as a snapshot: the count of removals per quarter against the count of additions is a rough but genuine net-adoption signal for the slice of the market these services can see. A quarter where removals exceed additions for the first time is worth investigating properly.

4. Archived customer-page diffs

Take snapshots at six and twelve months and list what is gone. Then do the part most people skip: check each departed company for evidence of where they went, because a logo that reappears in a rival’s directory converts a suspicion into a confirmed competitive loss with a named winner. Be disciplined about the false positives, since page redesigns and segment reorganisations remove logos for reasons that have nothing to do with churn.

5. Review-site cancellation and switching text

Review platforms are unusually rich here because their forms ask what a reviewer used previously and, on some platforms, what they moved to. Read the one and two-star reviews from the last two quarters specifically, and note the tenure the reviewer describes: churn concentrated in the first ninety days is an onboarding failure, while churn at year two or three is a value problem. That distinction changes which of their weaknesses you should be attacking.

6. Communities, forums and support threads

Unmoderated by the vendor and therefore blunter than any review. Migration guides written by users leaving a product are especially informative, because somebody who takes the trouble to document an exit route is describing a friction serious enough to be worth the work. Treat individual posts as anecdotes and repeated threads as evidence.

7. Retention and win-back job postings

Companies hire against problems they can measure internally and you cannot. A first Head of Retention, a churn-prediction data role, a renewals team where none existed, or a customer-success headcount growing faster than sales all say the same thing: retention has become a board-level topic there. The first instance of any of those roles is the signal, not the tenth.

8. Their contract terms and packaging changes

Watch their pricing page and terms over time. A shift to annual-only billing, a longer minimum term, a newly prominent multi-year discount, or the appearance of aggressive win-back offers are all mechanisms for holding customers who would otherwise leave. These changes are public, dated, and much harder to spin than a metric. Tracking that page over time is covered under competitor pricing.

9. Your own inbound from their customers

Count it properly and it becomes the best signal you own: how many opportunities this quarter involve a company currently using them, in which segment, at what tenure, and with what stated trigger. That is measured, current, first-party, and directly about the market you sell into. Most teams have this data and never aggregate it, because it arrives one deal at a time.

10. App store rating decay and review velocity

For products whose users live in an app, the rating trend and the rate of new reviews move well before any disclosure. A rating falling steadily across releases, or a sudden burst of reviews after a redesign, indicates a population that is unhappy and mobile. It is a weak signal on its own and a useful corroborator alongside the others.

A company under churn pressure behaves differently, and behaviour is public in a way metrics are not. Each of these is a decision somebody made because of numbers they could see and you cannot.

• Lock-in gets longer. Monthly billing quietly disappears, minimum terms extend, multi-year discounts become the headline offer. Companies do not add friction to buying unless leaving has become the bigger problem.
• Customer success grows faster than sales. When the hiring mix tilts from acquisition to retention, the constraint has moved from filling the funnel to keeping the base.
• Win-back campaigns appear. Nobody builds a win-back motion for a handful of departures. Its existence implies a population large enough to justify a programme.
• The roadmap turns inward. A run of releases about reliability, migration tooling, admin controls and reporting rather than new capability usually means existing customers are complaining loudly enough to set priorities.
• The messaging softens. A homepage that shifts from growth claims to reassurance, or new emphasis on support quality and partnership, is a company answering an objection it keeps hearing at renewal.

None of these is proof. Three of them arriving in the same two quarters is about as close to evidence of a retention problem as an outsider gets, and it is considerably more reliable than any number you could have bought. Where they matter most is renewals in accounts you share, which is why the conclusion belongs with your churn analysis rather than in a competitor file nobody opens.

1. 1Never convert a departure count into a rate. You do not have the denominator and you are not going to get it. Report the named departures and the trend in their number, and refuse the percentage even when somebody asks for one.
2. 2Require a second signal for every departure. A logo missing from a page plus a removed script, or a review describing the switch plus an appearance in your own pipeline. One weak signal is a hypothesis.
3. 3Check the metric definition. Net and gross retention differ by exactly the amount you care about, and companies choose which to publish. Read the definition in the filing rather than the headline figure.
4. 4Separate churn from a shrinking market. If every vendor in the category is losing customers, you are looking at a category problem rather than a competitor weakness, and attacking them on it will not work.
5. 5Date everything and watch the trend. A single quarter of departures is noise. The same signal appearing across three quarters, from independent sources, is a finding you can put in front of an executive.

Reading filings, reviews, public forums and your own records is entirely legitimate. Churn research carries one specific risk the other data points do not, and it comes at the end rather than the beginning: what you say about what you found.

• Do not publish a churn figure you cannot evidence. Telling prospects that a named competitor loses a third of its customers, when the number is an inference, is a false statement of fact about a business. Depending on the jurisdiction that is trade libel, unfair competition or misleading advertising, and it is the fastest way to turn a research exercise into a legal letter. Say what you observed, name the source, and stop there.
• Do not misrepresent yourself to reach private discussions. Customer-only communities, support forums and user groups are where cancellation conversations happen, which makes signing up as a fake customer genuinely tempting. It is deception, it usually breaches the platform terms, and the professional standard is to disclose your identity and organisation before any research interaction.
• Be careful with unpublished information about a listed competitor. A large account leaving a listed company is price-sensitive, so learning about it before the market does puts you in possession of information securities law restricts you from acting on or repeating. Nothing in the table above can produce that, since every source there is already public. A conversation with somebody on the inside can, which is the reason to be clear about what you are asking for.
• Approaching their unhappy customers is normal competition. Selling to a company that already buys from someone else is the ordinary business of selling. The constraints are on contacting individuals, which data protection and marketing rules govern regardless of who the person currently buys from.

• The churn rate itself. It exists in one billing system and is disclosed by nobody. Proxy: a direction supported by named departures, retention behaviour and your own inbound, reported as a direction.
• Their customer count. Without it, departures can never become a rate, and most vendors stopped publishing customer counts for exactly this reason. Proxy: detectable install counts as a countable universe, with the universe named.
• Churn by segment or cohort. The useful cut, and never disclosed. Proxy: the tenure and company size described by reviewers who left, which gives you a rough shape of where the losses concentrate.
• Voluntary versus involuntary churn. Failed payments and customers going out of business look identical from outside to customers choosing to leave. Proxy: whether the departed company is still trading, which takes one registry check.
• Renewals that were nearly lost. The account that stayed after a hard negotiation is invisible, and it is where their real vulnerability sits. Proxy: discounting behaviour at renewal, reported by shared customers and by prospects comparing quotes.

Split it by speed. The slow signals move quarterly and deserve a quarterly review: filings and transcripts, retention hiring, contract-term changes, and the archived customer-page comparison. The fast signal is continuous and you already own it, which is their customers turning up in your pipeline, so aggregate that monthly rather than reading it deal by deal.

Add an immediate check after three specific events, because each reliably sends a competitor’s customers shopping: a price increase, a significant outage or security incident, and a change of leadership in their customer organisation. The output is not a number. It is a short, dated note saying whether their retention looks better or worse than last quarter and what you saw, ideally beside the switching costs that decide whether their unhappy customers can actually leave.

Churn is the hardest thing on this cluster to keep current, for a reason specific to it: there is no event to watch. Every other data point has a moment. A price changes, a round closes, a role is posted. Churn only ever appears as an accumulation of fragments arriving on different clocks, a review here, a customer logo removed there, a retention role opened, a contract-term change buried in a filing. Assembling that picture by hand takes an afternoon, and it is stale within weeks because each fragment moves independently.

That is a collection problem before it is an analysis problem, and collection is what competitive intelligence software does well. Flares watches review sentiment, customer pages, product changes and hiring together, so the fragments land in one place with dates attached and the composite stays assembled instead of being rebuilt from scratch each quarter. It will not produce a churn rate, and you should be suspicious of anything that claims to. It shortens the time between a competitor’s customers becoming unhappy and your knowing about it.