Switching costs get discussed mostly as a retention mechanism. Lock users in, raise the cost of leaving, and you keep your revenue even when competitors improve.

That framing is defensive and a little cynical. There’s a more useful way to think about them.

Switching costs are a map of where value actually lives in your product.

High switching costs cluster around the places where users have made the product theirs. Uploaded their data. Built workflows around your interface. Trained team members on your conventions. Tagged, categorized, annotated.

Low switching costs cluster around commodity functions — things the product does that any other product also does just as well.

When you audit your own product for switching costs, you’re really asking: what has the user invested in here? The answer tells you where to double down and where your moat is thinner than you think.

The asymmetry in the title is this: switching costs are much higher for internal-data products than external-data products, almost structurally.

If you’re using an external-data tool — one that queries public markets, web indexes, or third-party databases — your switching cost is basically just habit and workflow friction. The data lives elsewhere. The tool is a window. Swap the window, same view.

If you’re using an internal-data tool — one that has ingested, indexed, and organized your own documents — your switching cost is your own history. The annotations, the summaries, the search indexes built over months of your files. That’s not portable. Switching means starting over.

This is why internal-data products can charge more and retain better even when their features are comparable to competitors. They’re not competing on features at renewal time. They’re competing against the cost of recreating everything you’ve already built inside them.

This asymmetry has an important implication for how to think about early traction.

External-data products can get early users easily — low friction to try, low friction to abandon. The adoption curve is smooth but so is the churn curve. You can measure product-market fit in days, but you can also lose it in days.

Internal-data products have a slower adoption curve. Users have to commit enough to upload something, organize something, build something inside. The first session is harder. But once they’re past that threshold, retention is almost automatic. They’re not evaluating you every month — they’re sitting inside infrastructure they built.

The early metrics look worse. The long-term economics look much better.

If you’re early in building something, it’s worth asking: am I building a window or a container?

Windows are easier to build and easier to sell. Containers are harder to enter but much harder to leave.

Neither is wrong. But you should know which one you’re making.