Switching Costs

Switching Costs

The cash machine (The Cash Machine) only compounds if customers stay. They do: Salesforce loses roughly eight percent of contract value to attrition each year, and that rate has barely moved for four years — through a demand slowdown, a restructuring, list-price increases, and the arrival of AI. That is the clearest evidence the base is durable. The complication sits beside it: over the same four years revenue growth halved, from the mid-twenties to about nine percent. The moat holds the base; it is not holding the growth rate.

A retention rate that will not move

Salesforce defines attrition as "the reduction or loss of the annualized value of our contracts with customers," measured on a trailing-twelve-month basis [1]. As of January 31, 2026 that rate, excluding Slack self-service and current-year acquisitions, was "approximately eight percent" — and management notes it stayed "consistent as compared to the prior year" [2].

The consistency is the point. Attrition was below 7.5 percent at January 31, 2023 on a narrower base that excluded MuleSoft, Tableau and Slack [3]. MuleSoft and Tableau were folded into the calculation from fiscal 2024, and the rate settled at approximately eight percent at January 31, 2024 [4] and again at January 31, 2025 [5]. An eight percent annualized loss implies the installed base renews at roughly 92 percent of its value each year before any upsell — a high floor for a business whose contracts typically run 12 to 36 months.

Revenue growth derived from reported revenue, FY2022–FY2026 10-Ks; attrition as disclosed. The FY2023 attrition point sits on a narrower base (excludes MuleSoft, Tableau, Slack). Sources: attrition [6][7][8][9]; revenue growth derived from reported financials.

The red line is nearly flat while the blue one falls by more than half. The two lines separate two questions that are easy to conflate: whether customers leave (they largely do not) and whether the business is still expanding quickly (less so). The revenue base is also broad — no single customer accounted for more than ten percent of revenue in any of fiscal 2024, 2025 or 2026 — so the retention figure is not propped up by a few large accounts [10].

What locks customers in

The retention number is a symptom; the mechanism is switching cost. Salesforce's own risk disclosure describes it plainly: prospective customers "have invested substantial personnel and financial resources to implement and integrate their current enterprise software into their businesses and therefore may be reluctant or unwilling to migrate away from their current solution" [11]. Written as a risk to winning new logos, it is equally a description of why existing customers stay: the same integration cost that deters a switch to Salesforce deters a switch away from it.

Two design choices deepen that cost. The first is the ecosystem: third-party developers and independent software vendors build applications that "can be marketed and sold on the AppExchange, our enterprise cloud marketplace," so a customer's Salesforce deployment accretes bespoke apps and integrations that do not travel to a rival [12]. The second is multi-cloud adoption: management states that customers who adopt multiple service offerings "drive higher annual revenues and have lower attrition rates than our company average" [13]. Each additional cloud a customer runs raises the cost of unwinding — the data gravity compounds.

A single peer point frames how strong ~92 percent gross retention is in absolute terms. ServiceNow — a genuine enterprise-SaaS adjacency named among the platform competitors Salesforce faces — reported a renewal rate of 98 percent for each of 2023, 2024 and 2025 [14]. The metrics are not identical — a renewal rate counts contract value retained at renewal, while Salesforce's attrition captures annualized value lost across the whole book — so the two are directional, not strictly comparable. The read is that Salesforce's stickiness is high but not best-in-class among the stickiest enterprise-software franchises.

A caution on the peer set: the indexed competitors here (Adobe, ADP, ServiceNow, Shopify, Snowflake, Uber) are a screened, imperfect match. Salesforce's true CRM rivals — Microsoft Dynamics, Oracle, SAP, HubSpot — are not in this corpus, so a full share-shift comparison is not available from the filings on hand. Only ServiceNow and, partially, Adobe run a comparable subscription-platform model; the retention comparison is limited to that.

The contracted book ahead

Retention shows up a second way, on the balance sheet, as revenue already sold. Remaining performance obligation — all future revenue under contract not yet recognized — stood at $72.4 billion at January 31, 2026, up 14 percent from $63.4 billion, against $24.3 billion of unearned (deferred) revenue [15]. Roughly $35.1 billion of that book is expected to convert to revenue within the next twelve months [16].

Attrition and RPO as disclosed; implied gross retention (100% − attrition) and revenue-under-contract (RPO ÷ FY2026 revenue of $41.5B) derived from reported financials. Sources: attrition [17]; RPO [18].

At about 1.7 times fiscal 2026 revenue under contract, the forward book is a second buttress under the durability question: even before a single new deal, a large share of the next few years' revenue is already signed. The composition and cash quality of that deferred balance — how much of the cash-flow smoothness is prepayment timing rather than earnings — is its own subject and sits with a later chapter; here it functions as evidence that the base is not only sticky but pre-sold.

Where the moat stops

The same evidence bounds the claim. Three facts cut against reading the moat as a growth engine.

First, the arithmetic of the chart above: attrition held near eight percent while revenue growth fell from about 25 percent in fiscal 2022 to roughly 9 percent in fiscal 2026. A moat that keeps customers does not, on its own, keep the top line compounding at its old pace — the retention floor is high, but the expansion above it (new logos plus net upsell) has decelerated sharply. Durability of the base and durability of the growth rate are different properties, and the filings support the first far more cleanly than the second.

Second, the reported ~8 percent is a curated figure: it excludes Slack self-service and current-year acquisitions [19]. Those are the cohorts most prone to churn, so the blended, all-in loss rate is somewhat higher than the headline. The direction of the disclosure — steady, excluding the weakest customers — is favorable; the precision is not absolute.

Third, management names the mechanism that could erode switching costs over a decade. Its own competition risk factor warns that "New AI offerings may disrupt our service offerings," in a market it calls "highly competitive, rapidly evolving and fragmented" with "low barriers to entry" [20], and describes competition as "intensely competitive" where AI-native entrants could bypass established workflows [21]. If an AI agent can sit on top of a company's data without the years of integration that lock a customer into a system of record, the switching cost that produces the eight percent attrition weakens. Salesforce's answer has partly been to buy the capability — its October 2025 acquisition of Regrello and November 2025 acquisition of Informatica were framed as accelerating its "agentic roadmap" [22] — which keeps the platform current but shifts more of the growth from organic to acquired.

On the evidence, the moat is wide where it is measured. Four years of ~8 percent attrition through a downturn, a $72.4 billion contracted book, and rising switching costs from ecosystem and multi-cloud lock-in make a strong case that the installed base is durable — which is most of what the ten-year-revenue question needs. What the filings do not yet settle is whether that durability defends the growth rate against AI-native competition, or only the base. Attrition would be the first number to move if the switching cost thins. It has been the franchise's steadiest number, and the first sustained move above eight percent — particularly if it coincides with AI-native displacement rather than macro softness — would be the earliest sign of that.