The Lock-In You Can't See Yet: A Framework for Reading Exit Cost Before You Sign

Lock-in never arrives as one dramatic moment. It accumulates quietly, contract by contract, until the "reasonable price" from year one becomes leverage the vendor holds because they know exactly what it would cost you to leave. · Juan Vegarra

Every technology vendor relationship starts the same way: a compelling demo, a reasonable price, and a sales team assuring you that switching later, if it ever comes to that, will be straightforward. It almost never is. Lock-in doesn't arrive as a single dramatic moment. It accumulates quietly, contract renewal by contract renewal, integration by integration, until the day you realize the reasonable price from year one has become a negotiating position the vendor holds because they know exactly how expensive it would now be for you to leave.

The numbers on this have become sharper in 2026 than they were even two years ago, largely because AI features have added new dependency layers on top of the ones executives were already used to tracking. One 2026 analysis puts the average cost of an enterprise vendor migration at roughly $315,000, and finds that organizations trapped in a lock-in situation without prior planning face switching costs up to sixteen times higher than organizations that built exit planning into the original contract.^[1] The same research identifies enterprise software buyers in 2026 as routinely accumulating switching costs across four or more separate AI vendor relationships simultaneously — spanning categories like productivity suites, CRM, and cloud infrastructure — with almost no methodology in place for measuring what that accumulation means for their negotiating position at the next renewal cycle.^[1]

Three kinds of lock-in, and why they compound

It's worth being specific about the mechanisms, because "lock-in" as a single word tends to flatten three genuinely different risks into one, and each requires a different mitigation. Data lock-in is generally the most expensive to escape, because it involves not just the records themselves but the metadata, configuration, and historical context that gave the data its operational value. Workflow lock-in is often the hardest to see coming, because after two or three years of building internal processes around a vendor's specific interface and logic, "switching" doesn't just mean migrating data — it means rebuilding the processes your teams have quietly grown to depend on. And integration lock-in is what makes the timing of an exit especially painful: every API connection built over the years assumes the vendor's specific data model, and migrating that isn't a weekend project. It's routinely a full quarter of dedicated engineering time, sometimes more.^[1]

Questions that belong in every platform review

The organizations that avoid this trap don't avoid vendor relationships altogether — that's neither possible nor sensible in most categories. They ask a specific set of questions before signing that most procurement processes skip entirely, because standard procurement is built to evaluate price and features, not exit cost.

The first is data portability, asked at a level of specificity that makes most vendors visibly uncomfortable. Not "can we export our data" — every vendor says yes to that question. The real question is: in what format, at what completeness, and at what cost in engineering time would it actually take to operationalize that exported data somewhere else. A CSV export of raw records, stripped of the configuration, workflow logic, and historical context that made the platform valuable, isn't portability. It's a paper receipt.

The second is integration depth versus integration surface area. A vendor deeply embedded in three genuinely critical workflows represents a different risk profile than a vendor embedded shallowly across fifteen minor ones. Depth creates real switching cost even when the surface area looks manageable on a vendor map, because unwinding one deeply embedded workflow can take longer and cost more than migrating ten shallow ones combined. This should be mapped explicitly during any significant deal review, not estimated by gut feel about how "important" the vendor seems.

The third is pricing structure under growth, not pricing at signing. Vendors routinely price the first contract aggressively to win the deal, then structure the pricing model to capture disproportionate value as usage grows — per-seat costs with no volume discount, usage-based pricing with no ceiling, or bundling that makes partial adoption practically impossible once a company is embedded. Recent price increases across the SaaS market illustrate the pattern: one major CRM vendor raised prices twelve percent, a productivity suite vendor raised prices fifteen percent, and a web platform vendor raised prices by twenty-three percent in a single cycle.^[2] Ask specifically what the contract looks like at three times and ten times current usage, not just at today's volume, before signing.

The fourth, and the one boards specifically should be asking management about directly, is what happens on vendor consolidation. What is the company's contractual and operational position if this vendor is acquired by a direct competitor, deprecates the specific product line the company depends on, or is acquired by an organization whose data practices the company would never have chosen to do business with directly? This is not a hypothetical question in a software market moving through as much AI-driven consolidation as the current one.

The honest counter-case

None of this argues against vendor partnerships, and treating vendor dependency as inherently dangerous leads straight back into the build-versus-buy trap covered earlier in this series, where organizations build expensive in-house alternatives to avoid a manageable dependency and end up with something worse than what a mature vendor already offered. Some lock-in is simply the price of a genuinely good product, and a company that refuses any meaningful vendor dependency will end up either building everything itself at enormous cost, or perpetually choosing immature, unproven vendors specifically because they're easier to leave — which trades a real, bounded switching-cost risk for a much larger and less bounded execution risk. The goal isn't zero lock-in. It's priced, understood lock-in, entered into deliberately rather than discovered three years later during a renewal negotiation the vendor has been quietly preparing for the whole time.

Treating exit terms as seriously as entry terms is the discipline this comes down to, because the exit terms are the ones you'll actually need in year three, when the relationship has changed, the pricing has shifted, and the leverage has moved decisively to the other side of the table. What does your organization's exit position actually look like on the platform deals signed this year — and does anyone know the answer before the renewal conversation forces it into the open?

Juan Vegarra is the author of An Outsider's Playbook (forthcoming). The views here are his own. More essays · Write me