Developing an in-house pricing tool: 
the actual cost of a build

The model is the easy part of a pricing tool. What takes years is the product catalog, product matching with competitors, rules, explainability, and compliance—some of which have expiration dates. Across all tools, 71% of in-house developments end up being abandoned, according to Exclaimer (2025).

The hidden cost is that of continuity: an internal tool relies on two or three people, and it is its success that creates dependency. The build process remains meaningful within a narrow and stable scope, as a complement to a platform. The deciding factor is not the ability to build, but the ability to maintain it for ten years.

This argument holds up perfectly in a meeting: we have a data team, our data is hosted on our own servers, and our business rules are specific.

And in fact, a competent team will release a price recommendation engine in a few months. That's not where the misunderstanding lies.

It focuses on what actually makes up a pricing tool in production: the model is the most visible—and the smallest—part of it.

An unfinished pricing machine under scaffolding, next to a machine that has already been assembled

Why does the build always seem cheaper at first?

Developing your own in-house pricing tool costs little at first but a lot later on: the recommendation model is the most visible—and smallest—part of a production pricing system.

A data team with access to sales history can produce a reliable elasticity model in a matter of weeks and, within a few months, an interface that suggests prices. At this point, the comparison with a vendor’s budget seems overwhelming.

The bias stems from the comparison itself: a working prototype on one side, a system in production on the other. In between, there is everything that turns a recommendation into the price displayed on the shelf.

The Success of the Prototype

A working prototype for a single category demonstrates the feasibility of the calculation, which has never been the difficult part. It says nothing about how the system will perform with 40,000 SKUs, 12 store formats, and 30 users with different access rights.

Our Position
A successful build is riskier than a failed build

A failed project ends quickly and costs only what it has already cost. A successful project, on the other hand, goes into production, becomes indispensable, and relies on just two or three people.

It is the success of the prototype that creates dependency. So the question regarding the build isn't "Will we succeed?", but "What will we do when it works?"

Booper Team, retail pricing since 2014

What a Pricing System Really Includes

Here's what needs to be built, in the order that the teams discover it—which is rarely the order they had planned.

The Product Catalog

To make a recommendation, you need to know what a product is, which product family it belongs to, and which items it replaces. In most stores, this information is available in several places and is inconsistent.

Making this foundation more reliable is a project in its own right, which we discuss in detail in the section on data quality—the “glass ceiling” of AI pricing.

Product Comparison with Competitors

This is the most underestimated component. Determining whether a competitor's product is actually yours—despite differences in labeling, packaging, and product codes—cannot be accomplished simply by comparing strings of characters.

A product-matching error occurs when a product is matched to the wrong item, and the result is immediately reflected in the margin without triggering any alerts.

The Rules Engine and Its Governance

An optimization calculation isn't enough: you need rules, priorities, exceptions, approval workflows, profile-based permissions, and traceability. It's a comprehensive management application, not a script.

The list of what it should cover overlaps with that of our pricing tool comparison.

Explainability

A buyer does not set a price that he cannot justify. For each recommendation, the system must reconstruct the data, the elasticity, and the rule that led to it. When built retroactively, this traceability is much more expensive.

Compliance, Which Has Deadlines

below-cost resale threshold, promotion caps, reference prices for discount ads: these rules are changing, and a mistake can have consequences that go beyond the commercial realm.

A concrete example: According to the DGCCRF, the 1.10 multiplier applied to the “ below-cost resale threshold ” for food products remains in effect until April 15, 2028. An internal tool will need to be modified by that date. The question remains: who will do it, and what will replace it?

71%

In-house tools often end up being abandoned, with 83% of such cases occurring in regulated sectors such as manufacturing and finance. Among teams that prefer to build their own solutions, 8% deliver on time and 11% stay within budget.
(Exclaimer, Build vs. Buy, more than 2,000 IT and security decision-makers, November 2025. Survey covering all types of tools, conducted by a software publisher)

What the Numbers Say About Internal Developments

The most revealing figure isn't the dropout rate; it's the gap between intention and delivery. An internal tool is judged against an estimate made at a time when we knew the least.

The same survey quantifies the maintenance effort: 63% of teams spend 10 to 50 hours per month maintaining their internal tools, and 66% spend an additional $20,000 to $100,000 per year on this. These figures rarely appear in arbitration files.

The Mechanism of Abandonment

The first version is useful, and adoption is growing. Then requests for changes start piling up faster than the team—which has other priorities—can handle them.

Delays are mounting, users are reverting to their spreadsheets, and usage is declining. The project’s abandonment is not a technical failure: it is a tool that is constantly competing with the other priorities of the team responsible for it.

The Cost No One Budgets For: Business Continuity

In practice, an internal pricing tool relies on two or three people. They are familiar with the architectural choices and the reasons for any exceptions—which are rarely documented, since all you have to do is ask them.

By the time they leave, the company has a system it no longer knows how to modify. We don't touch it anymore; we work around it, and then we rebuild it.

Three Questions That Reveal the Risk

How many people can modify a rule in the engine without the author's help? How long does it take for a newcomer to become self-sufficient? Who will handle a regulatory change in three months, and what will it replace?

The Cost of Time Versus Value

While the build is in progress, prices continue to be managed as before. This delay comes at a cost, which is rarely quantified.

Illustrative example. A company with €200 million in revenue is aiming for the lower end of the range identified by BOOPER, a +0.5 percentage point increase in margin. Each year of delay represents €1 million in lost margin—often more than the cost difference that originally motivated the build.

When a Build Is Still the Right Choice

Building a system is the right decision in three cases: a narrow and stable scope—a pricing model specific to your industry, covering a few hundred product lines.

A real competitive advantage: a pricing strategy that no one else uses and that you won't find on store shelves.

And above all, focus on complementing rather than replacing: a platform for the foundation (reference data, product matching, rules, compliance, traceability), and in-house development for the layer that supports your unique features.

At Booper
What a platform pools

At BOOPER, product matching combines text recognition and image recognition to match product listings whose names differ from one retailer to another.

below-cost resale threshold, Egalim promotion caps, and the 30-day reference price are managed natively and updated whenever the rule changes. This is precisely the part that no customer should have to rebuild on their own.

The Decision-Making Criterion, in One Question

"Are we capable of building it?" The answer is almost always yes, and it doesn't predict anything.

The real question is: Are we capable of sustaining it for ten years, despite staff turnover, shifting priorities, and regulatory changes? This is the factor that determines whether projects are abandoned.

The related question—whether to use general-purpose AI rather than a specialized solution—is addressed in “General-Purpose AI and Specialized Pricing Solutions.”

See also in this feature

FAQ

Building from scratch is justified within a narrow and stable scope, or when the pricing model provides a competitive advantage. In other cases, the approach that works combines a platform for the foundation with in-house development limited to the specific layer.

The model is inexpensive: just a few months’ worth of work by a data team. The real cost lies in the product repository, product matching against competitors, the rules and their governance,explainability, and compliance.

On top of that, there's maintenance: according to Exclaimer (2025), 63% of teams spend 10 to 50 hours a month maintaining their internal tools.

Because requests for new features quickly exceed the capacity of a team that has other priorities. Deadlines are pushed back, users revert to their spreadsheets, and usage declines. Across all tools, 71% of in-house development projects end up being abandoned (Exclaimer, November 2025).

Product matching with competitors. An incorrect mapping table leads to alignment with the wrong product, and the error appears in the margin without triggering any alerts.

A prototype can be released in a few months. A production system—with rules, permissions, traceability, and compliance—requires much more, and the work never stops: the regulatory rules themselves have expiration dates.

The real cost is the delay: during construction, prices are managed as before.

Not “Are we capable of building it?”, but “Are we capable of maintaining it for ten years?”. How many people can modify a rule without its author, and who will handle the next regulatory update?

‍

Articles
similar
A grid of clear lines that extends into a scatter plot connected by curves
November 17, 2026
AI or Business Rules: When Is AI Pricing Worth the Investment?

Well-defined business rules are often sufficient for a narrow, stable, and loosely interdependent product catalog. AI becomes cost-effective when decisions number in the thousands, products influence one another, and price sensitivity varies.

But you still need a usable sales history. In most cases, the right answer is a hybrid approach: the rules set the framework, and the model optimizes within it.

Read the blog post
Data flows from multiple sources pass through three stages before reaching human validation
November 13, 2026
Pricing Manager and AI: What the Machine Does, What the Expert Decides

AI handles the volume at every stage of the data journey: collection, product matching, verification, forecasting, and simulation. The expert selects competitors, resolves ambiguous cases, provides context, formulates the strategy, and makes decisions.

At Booper, Booper Mind makes suggestions, MPS monitors, and the user decides. Knowing how to interpret elasticity and question a forecast is more important than knowing how to code.

Read the blog post
Scales of Justice placed on a price dashboard, between three colored areas
November 11, 2026
Algorithmic Pricing: Where Is the Legal Line for a Retailer?

Dynamic pricing is permitted in France. Personalized pricing based on automated decisions is allowed, but customers must be notified of it. Algorithm-based price-fixing remains prohibited, even in the absence of contact between competitors.

The strongest safeguard for a retailer: being able to say, for each price, which rule generated it, what data it was based on, and who approved it.

Read the blog post
Ready to
 booster
What about your profit margins?

The pricing intelligence solution for retail leaders. Accuracy, speed, and controlled margins.

Let's discuss your pricing challenges
‍
‍