Where Do You Want Your Vendor Lock-In?

Every martech architecture decision leaves vendor lock-in somewhere in your stack. The real choice is where, and that call belongs to you and your CIO together, made before you sign rather than after you want out.

Ask a CMO where the customer data lives and you’ll get a vendor name. Ask the CIO the same question and you’ll often get a different one. They’re looking at the same stack through a different lens.

I see that split in my own search data. Two queries reach this site from people working the same decision. One opens “i am a chief marketing officer or marketing manager.” The other opens “I am an IT manager or chief technology officer.” After that the wording of the query is identical: how should an organization evaluate vendor lock-in risk when selecting a digital experience platform? Two roles, same question, asked separately, answered by a search engine instead of by each other.

Fix that before the architecture conversation starts, because that’s where the lock-in gets assigned.

What sits at the center of your stack

Something is at the center, whatever your architecture diagram says. It’s the system everything else is arranged around. Turn it off and most of your stack stops working. For some companies that’s the digital experience platform running the website. For others it’s the CRM, the data warehouse , a customer data platform, or the elusive holy grail not many marketing organizations have implemented: a shared record of activity that every tool reads from. For plenty of companies it’s a set of agreed rules and definitions with no system behind them at all.

Whatever sits there is what you’re stuck with, and each one leaves you stuck in a different way. Build around the warehouse and you’re tied to the data platform and a bill that climbs with volume. Build around a customer data platform and you’re tied to the stitched-together profiles it holds, which you’d have to rebuild somewhere else. Build around your own record of activity and you’re tied to your engineers and the knowledge that walks out when they do. Build around a vendor’s decision layer, the system that picks the next action for each customer, and you’ve rebuilt the all-in-one suite you were trying to escape.

Most stacks lean on 2 or 3 at once. The one you could least afford to turn off is your center, and that’s the vendor with the most leverage over you at renewal.

Assembling your own has a cost worth naming. David Raab, who named the customer data platform category, points out that it takes more technical effort and more customer-data expertise than buying one system, and every added component is another place the whole thing can slow down or break (1. Raab, 2024). That’s the honest trade for the control you get.

The decision layer is the newest of these, and the one with the least prior work to copy. In August 2026 I shared a panel run by the Association of National Advertisers (ANA), the US trade group for brand marketers, with Danilo Tauro , who has worked at P&G, Amazon Ads, and Uber Advertising and now runs CartographAI, a platform for evaluating martech and adtech vendors. He made the case that the sticky part of the stack is moving up a layer, from the individual tools to whatever runs the agents and keeps the business knowledge they build up over time, which Tauro calls accumulated context (2. Tauro, 2026).

So ask any AI vendor one question: can our own systems use your AI on our data, or does it only work inside your product? Inside-only means everything it learns about your business stays with them the day you leave. If your systems can call it from outside, you can change vendors and keep what it learned.

What your vendor already owes somebody

The European Union settled a version of this argument already. The Data Act took effect in September 2025, and it writes exit terms into every cloud and software contract inside its reach (3. European Union, 2023).

The vendor has to provide open interfaces so your data can move to a new system. It also has to keep the service running while you move, give you at least 30 days after that to pull your data out, hand over a written list of exactly what can be ported, and support your exit as a contractual obligation rather than a courtesy. From January 12, 2027, they can’t charge switching fees, though a fixed-term contract can still carry a proportionate early-termination penalty. A vendor whose service is expensive or complicated to leave has to say so before you sign (3. European Union, 2023).

The Data Act reaches past Europe’s borders. A provider with no European office is still in scope if it sells to European customers (4. Faegre Drinker, 2025). Those are the customers it owes these terms to, which doesn’t entitle your US-operating company to them. But a vendor already meeting this standard in Europe can’t tell you it’s impossible, and that changes what you’re arguing about.

Where the contract runs out

Those rights cover your records, your configurations, your files. What they leave open is everything an AI system learns by running your business for a year. An export returns your records. What stays behind is the model that learned which of your customers convert, and the workflow logic a year of corrections shaped. Whether that counts as a portable digital asset, or falls under the vendor’s trade-secret exemption, is unsettled (4. Faegre Drinker, 2025). The regulation doesn’t answer it. So ask before you sign: what does their AI record about its own decisions, how long do they keep it, and can you take it with you? If they can’t export it, your next vendor starts from zero.

Governance is the real protection. Establish before signing who owns the data, who gets to use it, where it came from, and who decides what connects to what, and the vendor never accumulates something you have to bargain for later. If you skip that step, no contract clause can add it later. Dependency also shows up in how your team behaves long before a renewal comes up, which I’ve written about in the real warning signs of martech vendor dependency .

Why marketing and IT price this differently

The split I saw in those 2 searches has a mechanism behind it. The person who negotiates the contract knows the terms: renewal dates, the built-in price rises (escalators), what it costs to terminate early. The person who knows what leaving costs, which downstream systems break and how many weeks the rebuild takes, never sees the contract. Each holds half the picture, and in most companies the halves never get compared.

So the risk gets priced badly in both directions at once. Procurement worries about the termination fee, which is the small number. The rebuild is the large one, and it goes unpriced because the people who could estimate it were never asked.

That’s fixable in one move, and it belongs in the negotiation rather than the renewal. Ask the vendor for a written offboarding plan. Walk it through your own engineers before you sign, and get an estimate in hours for standing the same capability up somewhere else. That figure is your real switching cost. Run it again at every renewal, because it grows quietly with every integration you add.

Then name what sits at the center of your stack today. Ask your CIO the same question on their own, compare the answers, and make the call together. If the answers differ, that’s the meeting to book before the contract goes out for signature.