Why people and organizations keep using the worse option — and how to see the invisible toll that keeps them there.
Switching costs are everything it takes to move from one option to another — money, time, effort, learning, lost relationships, or simply the discomfort of uncertainty. They explain why a company stays with a mediocre vendor, why someone keeps a phone plan they don't love, and why entire markets can lock onto one product even when a better one exists.
Every option has two prices: what it costs to use, and what it costs to leave. The second price is often invisible until you try to walk away.
High switching costs create "lock-in" — a gap between what's best and what people actually choose. Low switching costs create fragile loyalty, even when the product is great.
Software subscriptions, banking, cloud infrastructure, AI tooling, careers, relationships, habits, and any choice with setup, data, or learning invested in it.
Formalized in industrial economics and information economics (notably Carl Shapiro and Hal Varian's work on lock-in), and a staple of business strategy ever since.
Switching Costs sits near several relatives. The compass below plots it against three models already in this series, on two axes: how much the bias lives in feelings about the past (left) versus structural barriers going forward (right), and how individual (bottom) versus systemic/market-wide (top) the effect tends to be.
Switching Costs is distinct from Sunk Cost Fallacy: sunk cost is about honoring money or effort already spent (a feeling about the past), while switching costs are the real, forward-looking price of leaving — they exist even if you spent nothing getting in.
Switching costs usually come in four flavors, and most real decisions mix two or three of them:
Click into the next stage once you can name which of the four is driving a lock-in situation you've seen.
Businesses actively engineer switching costs on purpose: loyalty points, proprietary file formats, annual contracts, deeply integrated tooling. This isn't automatically predatory — sometimes it reflects a genuinely better, more integrated product. The mental model gets useful the moment you ask: is this cost serving the user, or just trapping them?
The catch is that switching costs are usually invisible at the moment of signing up. You feel the low price of entry; you only feel the full cost of leaving once you try.
Before adopting anything you'll depend on — a tool, a vendor, a platform, a habit — ask what it would cost to leave on day 400, not just what it costs to join on day one. Specifically: can you export your data in a usable format? How much custom work is locked inside this system? Who else would need to be retrained?
Once you can spot switching costs, you can do two things with the model: as a buyer or decision-maker, price in the exit cost before you commit. As a builder or seller, decide honestly whether you're earning loyalty or engineering a trap — the two often look identical from the outside, but your customers will eventually tell the difference.
A mid-size company's accounting software is clunky, but switching means re-entering five years of records, retraining the finance team, and risking a botched year-end close. The company renews every year — not because the software is good, but because leaving looks more expensive than staying, even though it often isn't once you account for the ongoing cost of poor software over several more years.
Health insurance tied to employment creates a switching cost on changing jobs: workers may stay in a role they'd otherwise leave because losing coverage, or re-qualifying for a new plan, feels riskier than a bad job. Policy debates about insurance portability are, underneath, debates about who should bear this switching cost.
Someone stays in a long-term friend group or a city they've outgrown because leaving means rebuilding a social network, routines, and a sense of belonging from scratch. The relationship or location may no longer serve them, but the switching cost — not the quality of the current fit — is what's actually holding the decision in place.
A team builds months of prompts, fine-tuned behavior, and internal tooling around one AI provider's API. A competitor ships a better, cheaper model, but migrating means rewriting prompts, re-validating outputs, and retraining the team on a new interface. The switching cost — not the quality gap — decides whether the team actually moves. Serious AI users increasingly treat "how hard is this to leave" as a first-class question when adopting any platform, not an afterthought.
Scenario: Your team is deciding whether to migrate from one cloud infrastructure provider to another that's 30% cheaper with comparable features.
Fill in your own situation and generate a ready-to-use prompt for thinking it through with an AI assistant, or on your own.
Copied!Turn this model into a reusable instruction set you can hand to an AI assistant, or keep for yourself, any time a "should we switch?" decision comes up.
Copied!