Switching Costs: Why Customers Stay, and How to Earn It

Switching costs are the money, time, data and effort a customer would lose by moving to a competing product. When they are high, customers stay even if a rival is cheaper or slightly better. They are one of the most common moats in software, because every month of use adds history, integrations and habits that do not transfer.

The strategist Hamilton Helmer, in 7 Powers, describes switching costs as arising when a customer values compatibility across multiple purchases from a specific firm over time. In plain terms: the product gets harder to leave the longer you use it. That is different from a product that is merely good. A good product with no switching costs can be copied, and the copy can win your customers in a quarter. For the broader idea, see what a moat is.

Why switching costs matter for startups

For a large company, switching costs protect revenue that already exists. Salesforce reports an annual attrition rate of approximately eight percent (excluding Slack self-service and recent acquisitions), which means most of its customer base renews every year without being won again.

For a startup, the question is different. You have no installed base yet, so switching costs cannot protect you on day one. What matters is whether normal use of your product creates them. If a customer who has used you for six months would lose real data, a configured workflow or a reputation by leaving, a funded copycat has to offer far more than a better feature. If they can export a CSV and leave in an afternoon, your only defense is being better every single week.

Switching costs also work against startups. Incumbents with high switching costs are hard to displace, even with a superior product. That is why many successful startups attack customers who have not yet been locked in: new companies, new teams, or new workflows that no incumbent owns.

The 7 types of switching costs

Helmer groups switching costs into financial, procedural and relational categories. In practice it helps to break them into seven:

• Financial or contractual: multi-year contracts, termination fees, sunk implementation costs, and fees for moving data out.
• Procedural or workflow: approvals, automations and processes that the team has built around the tool.
• Data and history: records, files, past transactions and analytics that are hard to move or lose meaning outside the product.
• Integration: connections to other systems (ERP, payments, identity, apps) that would all need rebuilding.
• Learning: the time staff spent becoming fluent, plus the productivity dip during retraining.
• Relational or reputation: reviews, ratings, trust with a vendor, and status that cannot be carried elsewhere.
• Network: leaving means losing contact with the other people who use the product. This overlaps with network effects.

The strongest positions stack several types at once. A hospital leaving its health record system faces contract, data, integration, learning and workflow costs together.

15 switching costs examples

1. Salesforce: years of customer records, custom objects and automations live in the CRM, and its attrition rate has stayed around eight percent.
2. SAP and Oracle ERP: finance, inventory and procurement processes are configured inside the system, so replacing it means redesigning how the company runs, not just swapping software.
3. Epic Systems: hospital record systems are among the most expensive software projects anywhere; Kaiser Permanente's Epic-based platform cost about $4 billion, and few systems will repeat that work lightly.
4. Workday: HR and finance data plus approval workflows are embedded, and Workday reported gross revenue retention of 98% in its fiscal 2025 fourth quarter.
5. Adobe Creative Cloud: professional files, templates and skills are built around Adobe formats and tools, so designers and agencies face both file and learning costs when they leave.
6. Apple iMessage and ecosystem: in a 2013 email revealed in the Epic Games litigation, Apple's Craig Federighi wrote that iMessage on Android would remove an obstacle to iPhone families giving their kids Android phones.
7. Snowflake: customers load their data, pipelines and queries into the platform and keep expanding usage; Snowflake reported a net revenue retention rate of 125% for fiscal 2026.
8. AWS: data transfer out fees once made leaving expensive, though AWS now waives them for customers migrating away, leaving integration and architecture as the real switching costs.
9. Shopify: a merchant's store, theme, checkout and installed third-party apps all run on Shopify, so moving means rebuilding the stack piece by piece.
10. QuickBooks (Intuit): years of books, bank feeds and the accountant who already knows the file make small businesses reluctant to change.
11. Gusto and ADP payroll: payroll history, tax filings and benefits setups carry compliance risk if a migration goes wrong, so companies are cautious about switching mid-year.
12. Airbnb and eBay: a host's or seller's reviews stay on the platform; Airbnb has over one million Superhosts, a status earned through ratings that do not transfer to a rival.
13. Figma: design systems published as team libraries are shared and synced across every file, so leaving means rebuilding the components a whole company designs with.
14. Notion: company wikis, linked databases and docs accumulate over years, and rebuilding their links and structure elsewhere takes real work.
15. Bloomberg Terminal: traders learn its keyboard functions and use its built-in messaging to reach counterparties, combining learning, workflow and network costs.

Evidence: retention numbers

Switching costs are hard to observe directly, but they show up in retention. Two metrics matter:

High net retention does not prove switching costs on its own (it can also reflect a product customers simply love), but low gross retention is strong evidence they are missing.

Regulation is lowering some switching costs

Some switching costs are artificial, and regulators increasingly remove them. US carriers must let customers keep their phone numbers: wireless number portability has been available nationwide since 2004. In Europe, the Data Act limits cloud switching charges to direct costs until 12 January 2027, after which providers may not impose any switching charges, including data egress fees. AWS's own fee waiver followed the direction set by the Data Act.

The lesson for founders: do not build a moat on fees and friction. Contractual and egress lock-in can be legislated away. Switching costs that come from real value (configured workflows, accumulated history, earned reputation) are the ones that last.

How a startup builds switching costs from zero

1. Become the system of record for one thing. Pick the record your customer cannot lose (patient notes, contracts, books, payroll) and make your product where it lives. This is closely tied to owning the customer relationship.
2. Let customers configure the product. Custom fields, automations, templates and approval rules take time to build and do not exist anywhere else.
3. Integrate early. Every connection to a customer's other tools (accounting, identity, payments, Slack) is one more thing to rebuild after leaving.
4. Spread across the team. A tool one person uses is easy to drop. A tool ten people rely on daily needs a group decision to replace.
5. Accumulate history that compounds. Past data should make the product more useful: better reports, forecasts or recommendations. That is where switching costs meet a data flywheel.
6. Make reputation live in the product. Ratings, reviews, verification and track records that only exist on your platform give users a personal reason to stay.
7. Keep export honest. Customers trust products that let them leave. Real switching costs come from the value of what they built, not from blocking the door.

Common mistakes and fake switching costs

• Confusing annual contracts with lock-in. A contract delays churn by up to a year. It does not stop a customer from leaving at renewal.
• Hostage data. Making export hard creates resentment, invites regulation and becomes a selling point for competitors.
• Counting habits as a moat. A simple interface people are used to is a learning cost measured in hours. A rival can copy the interface.
• Assuming integrations are sticky by default. A single Zapier connection or one-click import is easy to rebuild. Deep, configured integrations are what count.
• Claiming switching costs "at scale". If a customer could leave today and lose nothing, the moat does not exist yet.

How Moat Check scores it

Moat Check asks one question for this test: "Would leaving mean losing history, integrations, workflow or reputation?" Each of the 10 tests scores pass (10 points), partial (5) or fail (0), with one line of evidence, as described in the methodology.

• Pass: leaving means losing history, integrations, reputation or established workflows, and this is true today or by design of the product.
• Partial: switching costs are plausible but weak or not yet in place, for example they only appear "at scale", after a long period of use, or depend on integrations that have not been built.
• Fail: customers can export everything and switch in an afternoon.

The score is a starting point produced by AI and web research applying a fixed framework, not investment advice. Switching costs are one of 10 tests; see all of them on the moats hub.

Check your startup's moat

Questions

Keep reading

• Distribution Moat: How Startups Own Channels Rivals Can't Buy
• Platform Risk: What Happens When the Platform Ships Your Feature
• AI Moats: What Survives When Everyone Has the Same Model
• All 10 types of moats
• What is a moat in business?