Reading Retention Disclosures for Hidden Customer Lock-In

If a company tells you that existing customers spent 126% of what they spent a year ago, has it just handed you proof of a moat? Plenty of investors treat it that way. A retention number above 100% reads like a verdict: customers stay, customers buy more, and nobody seems to leave.

I think that's half right, and the wrong half is the expensive one. A retention figure is evidence of an outcome. It says nothing, by itself, about the cause. Customers can stay because leaving is genuinely hard, or because nobody has bothered to look at the alternatives yet, or because the product is simply good this year. Only the first of those is structural. The other two evaporate the moment a credible competitor shows up.

So the working question isn't "how high is retention?" It's "what would it cost a customer to leave, and is that cost written anywhere in the filings?" Companies disclose more about this than most readers notice, scattered across the business description, the metrics section of the MD&A, the revenue footnote, and the risk factors. What follows is a way to read all of it in a fixed order.

Start With What the Metric Actually Measures

Net revenue retention (NRR) compares what a fixed group of customers spent in the current period with what the same group spent a year earlier. Expansion pushes it up. Churn and downgrades pull it down. Every company defines the cohort, the window and the revenue base a little differently, which is why the definition paragraph in the filing matters more than the number next to it.

Snowflake gives a useful illustration. Its fiscal 2025 results (the year ended January 31, 2025) showed a net revenue retention rate of 126%, down from 131% a year earlier. Over the same period, the count of customers with more than $1 million in trailing twelve-month product revenue rose from 461 to 580. Read quickly, that's a company whose customers keep deepening the relationship.

Read carefully, it's more ambiguous. Snowflake bills on consumption, so a customer's spending rises when the customer runs more queries and falls when it runs fewer. A 126% figure partly reflects real dependence and partly reflects how much data the customer happened to process. The metric blends lock-in with usage. I'm less sure than I'd like to be about how much of each, and the filings don't separate them.

Gross retention is the cleaner signal

Net retention nets expansion against losses, so a few big customers growing fast can hide a leaky base. Gross retention counts only the losses. When a company discloses it, that's the number closest to a pure measure of how hard customers find it to leave. When it doesn't, treat the omission as information, and look for logo churn or renewal rates instead.

Check the denominator

Some companies calculate retention only on customers above a spending threshold, or only on annual contracts. A metric that quietly excludes small or month-to-month customers will flatter itself. Look for the phrase describing which customers are in the cohort. Then ask who got left out.

Separate the Contract From the Habit

The business description and revenue footnote usually reveal whether customers are bound by something or merely used to something. This is where switching costs either show up in the paperwork or don't.

Signals that point to structural lock-in

• Multi-year contract terms, with remaining performance obligations (the revenue the company has contracted but not yet recognized) growing at least as fast as revenue.
• Language about integration, data migration, retraining or certification as things customers must go through to change providers.
• Renewal rates disclosed for the full customer base, not for a favorable subset.
• Implementation fees or professional services revenue that signal a heavy setup process on the way in.

Signals that point to habit

• Month-to-month or cancel-anytime terms for most of the revenue base.
• Retention described mainly through satisfaction scores or net promoter scores rather than contract mechanics.
• Expansion driven by price increases and usage swings rather than by additional products adopted.
• A sales and marketing line that keeps climbing while retention holds steady, which suggests the company is paying to keep customers it should be keeping for free.

None of the habit signals mean a business is weak. Plenty of good companies run on cancel-anytime terms and earn their renewals every month. But that's a different thing from a moat, and it deserves a different multiple of your confidence.

Follow the expansion to its source

Expansion revenue can come from three places: customers buying more of the same product, customers buying additional products, or the company charging more. Each says something different about durability.

Additional products adopted by the same customer is the strongest case. It means the first product created a relationship that made the second sale easier, and each added product raises the cost of leaving. Volume growth inside one product is neutral: it's demand, which can reverse. Price increases that customers absorb without leaving are real evidence of pricing power, but only if the churn line stayed flat while they happened.

The filings rarely split expansion this cleanly. The best you'll usually get is a paragraph in the MD&A attributing growth to "increased usage," "new customers" and "expansion into additional products," sometimes with a count of customers using multiple products. If that count exists and is rising, weigh it heavily. If the paragraph only says "increased usage," you're looking at demand, not lock-in.

One habit helps here: line up three years of the same disclosure side by side. A single retention figure is a snapshot, and snapshots flatter. Three years shows direction. Snowflake's slide from 131% to 126% isn't alarming on its own (a larger base makes high percentages harder to sustain), but the direction is the thing to track, and you can only see it if you keep the older filings open next to the new one. Ask whether the decline is gradual or sudden, and whether the explanation management gives in the MD&A matches the pattern in the numbers.

It also helps to read a second company in the same industry using the same checklist. Retention that looks exceptional in isolation often looks ordinary against a peer, and a number that survives the comparison is far more convincing. If you can't find a peer that discloses the same metric, that is itself a clue: the metric may be one the company chose precisely because it flatters.

Read the Risk Factors as a Confession

Risk factors are written by lawyers who'd rather over-disclose than under-disclose, which makes them oddly candid about retention. Look for sentences like "customers may reduce their spending or terminate contracts with limited notice." That sentence tells you how binding the contracts really are, in the company's own words, and it sits in the same document as the headline retention figure.

Compare that language against the first section of the filing. If the business description talks about deep integration and the risk factors admit customers can leave on short notice, one of the two is overselling. Then check the cost side: a company that's confident in its lock-in doesn't need to discount heavily at renewal, and discounting language in the filing (renewal concessions, pricing pressure, competitive pricing) is a warning that the ability to hold customers depends on price.

A Four-Step Reading Order

Put together, here's the sequence I'd run on any company that advertises strong retention. It takes about an hour per company once you've done it a few times, and it forces the cause question before the number gets to anchor you.

1. Find the retention definition. Note the cohort, the time window and which customers are excluded.
2. Find gross retention, logo churn or renewal rates. If none are disclosed, say so in your notes and lower your confidence.
3. Classify expansion as additional products, volume or price. Weight them in that order.
4. Read the risk factors for the contract-termination language, and set it against the business description's lock-in claims.

Then run the result against the financial fingerprints in our guide on how to find wide moat stocks. Lock-in that's real should be showing up in stable gross margins and returns on capital. If retention looks great and margins are sliding, trust the margins.

A worked pass shows how short the process is. Suppose a software company reports 118% net retention. Step one finds that the cohort covers only customers with annual contracts above a set size, which excludes the long tail. Step two finds no gross retention figure, so confidence drops a notch. Step three finds expansion attributed mostly to usage, with only a modest count of multi-product customers. Step four finds a risk factor admitting that most contracts allow termination for convenience after twelve months. The headline looked like a moat. After four steps, it looks like a good product with decent habits, which is worth owning only at a price that doesn't assume more.

Where This Framework Breaks

Three honest limits. First, young companies have short histories, so a single year of 120%-plus retention may be the first lap of a long race or the high point of an adoption wave. The data here is thin and I'd resist drawing strong conclusions from fewer than three years of disclosure.

Second, concentration distorts everything. If a handful of customers drive most of the expansion, the headline number describes them, not the base. We covered how that cuts against an otherwise strong position in how customer concentration undermines a wide moat.

Third, lock-in cuts both ways. A customer base that can't leave also resents the vendor, and that resentment tends to surface as regulatory complaints or the first competitor willing to subsidize a migration. And when retention starts slipping, it usually shows up among the first of the early warning signs a moat is eroding, well before margins move.

None of this makes retention data useless. It means the number is where the analysis begins. The filings give you enough to work backward from the outcome to the mechanism, and the mechanism is the part that decides whether the outcome lasts. What would it take to displace them? If the honest answer is "a slightly better product," the retention figure is describing satisfaction. If the answer involves rebuilding integrations, retraining staff and renegotiating contracts, you may be looking at a structural advantage.

Key Takeaways

• A high retention number is an outcome. Whether it reflects a moat depends on why customers stay, and the filings usually say.
• Prefer gross retention and renewal rates over net retention, and check which customers the metric excludes.
• Rank expansion by source: additional products first, volume second, price increases last.
• Read risk-factor language on contract termination against the business description. Where the two disagree, believe the risk factors.

This framework describes how to read disclosures; it isn't a rating or a recommendation on any security. Verify every figure against the company's filings before relying on it.