Reversibility: 

What do you get if you switch to a different pricing plan?

The question no one ever asks before signing

In a pricing solution consultation, the evaluation grid covers features, model quality, customer references, cost, implementation timelines, and support. Output is almost never included, and when it is, it’s limited to a single line about data delivery.

This clause addresses only a fraction of the problem. Your raw data belongs to you—that much is clear—and no reputable publisher will withhold it. But you already had that raw data before you signed. What you’ve built over three years of use is something else entirely, and that is precisely what the contracts fail to address.

Why Timing Matters So Much

Before signing the contract, you have maximum bargaining power and zero opportunity cost. After going live, you have trained teams, established processes, built interfaces, and a real operational dependency. The same request carries very different weight at these two points in time.

It should be noted that asking these questions up front is not a sign of mistrust, and a good provider understands this. A provider that gives a clear answer about reversibility demonstrates confidence in its value; a provider that evades the question suggests that part of its customer retention strategy relies on the difficulty of leaving.

The four assets you create with your publisher

1. Raw data

Sales, prices, inventory, and reports. This is the only asset typically covered by contracts, and it’s also the least critical: it comes from your own systems, so you can reconstruct it. The only real question concerns the format and turnaround time for data delivery, which ties into the issue of interfaces—discussed in “Integrating Your Pricing Data Without a Major IT Project.”

2. The Rule Set

Several hundred rules written over the years, each with its own priorities, exceptions, and scope. This asset represents a considerable amount of work, as our article on the “rule debt” of a pricing engine demonstrates, and it is very rarely exportable in a reusable format. An export in a proprietary format—or a list of parameters without the logic that ties them together—is of little value elsewhere.

3. History of Arbitration Cases

It is the most valuable—and most consistently overlooked—asset. Every recommendation—whether accepted, modified, or rejected, along with the reason given—forms the record of the company’s decisions. This is what enables a new system to learn quickly, and, above all, it ensures that your teams don’t have to reinvent three years’ worth of best practices. It is rarely included in a clause.

4. Competitive Matching

The table that shows which of your product codes corresponds to a specific product code at a particular competitor. It was built through iterations, manual corrections, and successive validations. Reconstructing it would take months. It is often considered part of the publisher’s know-how rather than customer data, and this is the most common source of contractual friction.

The Gap Between Perception and Reality in Migration

The lesson to be learned from these figures is not that migration is impossible. It is that policymakers’ estimates of migration before it actually occurs are consistently overly optimistic—to such an extent that they render planning ineffective.

There are three reasons for the discrepancy, and they fully apply to pricing.

• The export exists but is unusable. A file containing several hundred thousand lines without any documentation on its structure is technically an output but practically a problem.
• Tacit knowledge cannot be captured. The reasoning behind the settings, the workarounds users know, the special cases handled outside the system—none of this is in the database.
• The transition isn't instantaneous. Two systems must run in parallel during the validation period, which requires that the old contract allow for it and that the schedule have been planned accordingly. This is the same constraint as when migrating from Excel to a pricing platform.

The Seven Clauses to Include in the Contract

Here are some useful points for negotiation, worded so they can be included as-is in a consultation.

• The exact scope of the data return. List the four assets mentioned above by name. A general statement such as “the customer’s data is returned to the customer” will be interpreted in the narrowest possible sense when the time comes.
• The format: well-documented and open. Require a format that is readable without the tool, accompanied by documentation describing its structure. This is the clause that distinguishes an export from a usable output.
• The timeframe, specified in numbers. A number of business days from the date of the request, not an obligation of means. And the option to request a refund during the term of the contract, not just at its conclusion.
• No additional cost. The refund must not be billed as a separate service, or it risks becoming a theoretical right.
• The parallel operation period. The right to maintain access to the old system during the transition—typically three to six months—with the associated pricing terms established from the outset.
• Ownership of the mapping. Explicitly state that the mapping table between your references and those of your competitors is client data. This is the point that will require the most discussion, and it is the one that will be the most expensive to recreate.
• What happens to the data in the event of termination? What happens if the publisher is acquired, changes its service offering, or ceases operations? The continuity clause is rarely requested, and when it is, it is almost always granted.

What's the Point of Demanding It?

The models themselves. The trained algorithms constitute the core of the publisher’s intellectual property, and access to them will never be granted. This is not a shortcoming, however: a model can be retrained using your data, whereas a set of rules and a mapping can only be reconstructed manually. This is why the formalized context matters more than the model—a point we elaborate on in “The Business Context Is Not Data.”

Test its reversibility before you actually need it

An untested clause is an assumption. The most effective approach is to request a full delivery once a year—with no intention of terminating the contract—simply to verify three things: that the stated deadline is met, that the format received is readable by your own teams, and that the scope of the delivery matches what is described in the contract.

This annual request has a second, more political effect: it helps maintain a balanced supplier relationship. A customer whom we know might leave is a customer with whom we negotiate differently, and this is reflected in the quality of support long before a relationship breaks down.

What if the contract has already been signed?

There are three possible courses of action. Request a trial return, even without a clause: the refusal itself provides valuable information. Document internally what cannot be exported, particularly the reasons behind the rules, which effectively ensures reversibility. And include the missing clauses in the next renewal, which is the only time when the balance of power shifts back in our favor.

If the decision to change has already been made, the issue is no longer a contractual one but an operational one, and we address it separately in “Changing Your Pricing Tool Without Repeating Its Mistakes.”

One final word on intent. Negotiating reversibility isn’t the same as preparing to leave, any more than taking out insurance is the same as anticipating a claim. It’s the condition that ensures the relationship is based on the value delivered rather than on the cost of leaving. Both parties benefit: a publisher that retains its customers through their satisfaction builds a more solid business than one that retains them through their data.

FAQ