What is vendor lock-in, and how do you avoid it?

A situation where the cost of leaving a supplier exceeds the benefit of doing so — not because a contract forbids it, but because migration is expensive, risky or technically impractical. It is rarely a single decision and almost always an accumulation of small ones.

The forms it takes:

Data lock-in, where your data is in a proprietary format or can only be exported incompletely, slowly or at cost. This is the most serious form, because data is the thing you cannot recreate.

Technical lock-in, where your systems are built against proprietary interfaces with no equivalent elsewhere.

Skills lock-in, where your people know one platform and retraining is a real cost.

Contractual lock-in — long terms, exit fees, and discounts conditional on commitment.

Process lock-in, where workflows have been shaped around one tool's assumptions.

Ecosystem lock-in, where an integrated suite is convenient precisely because the pieces only work well together.

Why it is not automatically bad. Deep integration with one platform frequently delivers real value — better performance, less glue code, fewer vendors to manage. The mistake is incurring lock-in accidentally rather than deciding to accept it, and not knowing what it would cost to reverse.

What actually reduces it:

Own your data, and test the export. An export function that has never been run is not an export function. Do this before you need it.

Prefer open standards and formats where the difference is small.

Isolate proprietary dependencies behind your own interfaces, so replacing one component does not touch everything.

Keep infrastructure defined as code, which makes re-creating an environment a mechanical task.

Know your exit cost. Estimate what migration would take, and revisit it — this single number changes negotiating position substantially.

Negotiate exit terms at the start, when you have leverage. Data return, format, assistance and notice all become non-negotiable later.

Beware egress fees, which are the explicit form of the problem and have attracted regulatory attention.

The realistic aim is manageable lock-in, not none.