The business context is not data: what your tool doesn't know
October 9, 2026
The data describes what happened; the context explains what is not done here. This context is implicit: price benchmarks, supplier constraints, unwritten rules, and category-specific guidelines. No search engine can guess it, and no general-purpose AI can retain it from one session to the next.
The least expensive way to formalize it is through rejection: every recommendation that is rejected, along with the reason given, reveals a piece of it. Formalizing it has lasting value, because the context outlives the models and remains a company asset even when the tool changes.
There is one very effective way to derail a pricing project: to get the data side absolutely right. A clean database, complete historical data, reliable product matching against competitors, and real-time data feeds.
And realize that the engine generates recommendations that are technically flawless but commercially untenable.
The problem isn't with the data. It's with everything that's missing from it—things that no one has ever written down because, within the company, everyone already knows them.

Data, rules, context: three different things
These three terms are often used interchangeably, and this confusion accounts for much of the disappointment surrounding pricing projects. Yet they refer to three distinct concepts.
- Data describes what has happened: sales, recorded prices, inventory levels, and realized margins. It is extracted from a system, whether it is endogenous or exogenous.
- The rule describes what the system must do: do not fall below a certain threshold, do not deviate by more than a certain amount from a specific competitor. This is the mechanism described in “How an AI Pricing Engine Works.”
- Context explains why the rule exists, and what would be absurd even without a rule to prohibit it. It is neither factual nor enforceable: it exists only in people's minds.
Data quality is a well-recognized issue, which we address in the “glass ceiling” of AI pricing. Perfect data answers the question, “What happened?”
It says nothing about the question, “What isn’t being done here?” That’s where context comes in.
A classic example that sums it all up
Illustrative example: a shelf-stock item with low turnover, a reasonable margin, no established competitors, and no regulatory constraints. A well-funded algorithm will recommend raising its price by a few cents.
The recommendation is accurate in every measurable respect. Yet it is inapplicable: it is the benchmark that management cites to illustrate the brand’s positioning, and which a local news article used as an example.
The sales department has decided it won't budge. None of this exists in the database. Only five people in the company know about any of this.
Some organizations do not have—or do not know if they have—data management practices suited for AI. Gartner predicts that 60% of private AI projects lacking data ready for this purpose will be abandoned by 2026.
(Gartner, press release dated February 26, 2025)
This figure is often interpreted as a statement about the technology. It deserves a broader interpretation: “AI-ready” data is also data that comes with the tools needed to interpret it.
The four context families that no database contains
1. Price Benchmarks
A handful of products define the brand's price image: those that customers know by heart and compare from one store to another. Sales data doesn't always distinguish them from the others.
Part of this knowledge is reconstructed statistically; that is the work of a well-designed engine. Another part stems from local history, a stated position, and a public promise.
2. Relationship Constraints
That supplier who refuses to have its product line marked down before a certain date. That brand whose contract stipulates a maximum price difference from the store brand. That partner currently in negotiations—it’s best not to make any changes regarding them this quarter.
These constraints have direct financial consequences. They are reflected in contracts, reports, and conversations—but never in sales figures.
3. Unwritten Taboos
These are the most dangerous ones, because no one thinks to mention them. We never go below a certain psychological threshold when it comes to this company. We never align ourselves with this competitor, because the brand doesn't compare itself to it.
They are not seen as rules but as self-evident truths. They only become apparent when the character breaks them, with the phrase, “Of course you don’t do that.”
4. Shopping Intentions by Category
A category focused on market expansion is willing to sacrifice profit margins to gain market share; a category focused on profitability does the opposite. This strategy changes over the course of the year and is decided by the committee.
It is rarely translated into metrics. As a result, the engine optimizes the same objective across the board, even though the company pursues multiple objectives depending on the department.
In the same series: “Managing an AI-Powered Pricing System”
How can you tell if an engine lacks context?
The symptoms are clearly different from those of a data problem.
- The recommendations are sound but have been rejected. No one disputes the calculation; everyone refuses to implement it.
- The acceptance rate varies significantly from one category to another, even when data quality is the same. The categories with high acceptance rates are often those whose managers were involved in configuring the system.
- The same manual corrections come up every week. Each one is a piece of context that’s manually reinserted.
- Objections begin with “obviously.” This is the most reliable indicator of tacit knowledge.
General-purpose AIs have the same blind spot, only more pronounced. A user quoted in the MIT NANDA study sums it up: the tool “requires that you provide it with the full context every session” (MIT NANDA, July 2025).
Where Context Must Live
Asking teams to write down their rules almost always fails: we don't know how to articulate what we've never needed to put into words. Context is better revealed through refusal than through open-ended questions.
Our belief:a refusal to provide a recommendation without a reason is a missed opportunity for information. It is the company's least expensive—and most overlooked—source of context.
A reason noted in a single sentence at the time of rejection is better than a two-day workshop on formalizing reasons. After a year, these reasons fall into the four categories described above, section by section.
Not all of the challenges facing AI projects stem from algorithms. About 70% are due to people and processes, and 20% are due to technology and data.
(BCG, “Where’s the Value in AI?”, October 24,2024; 1,000 executives in 59 countries)
This ratio says it all: the algorithm is the easy part. What causes a project to fail is what people know but what processes fail to convey.
Context does not mean a rigid rule
Turning every piece of context into a blocking rule is the best way to create rule debt. Much of the context should remain information that is displayed at the time of the decision.
The difference is significant: information informs human decision-making, while coercion replaces it. A binding rule is reserved for matters that are non-negotiable, such as the law or a contractual obligation.
At Booper, this flexibility means that target, minimum, and maximum price limits may vary from one retail location to another, with an alert triggered when a price exceeds these limits rather than a systematic block.
A context embedded in the tool: the strategy is clearly defined, and each scenario retains its rationale.
Saved, compared, and replayable price scenarios.
The store's strategy: to be the cheapest in the area. Booper serves as a benchmark for comparing prices with competitors, and each scenario is recorded: what was decided, why, and the ability to replay it.
- Issue
- Hold the cheapest position in the zone without losing sight of the margin.
- What Has Changed
- Comprehensive competitive benchmarking and a complete history of pricing scenarios.
Users are enough to drive the hypermarket's price alignment.
Store Management · Independent Hypermarket. Case study written by Booper based on an actual client project; client identity has been anonymized.
Why Context Outlives Models
Forecasting and optimization algorithms evolve rapidly. An approach that was considered the gold standard three years ago is now outdated. Therefore, making the model the core of one's investment is risky.
The context, on the other hand, evolves over the years, not with each new version. A company that has formalized it can switch to a new engine without starting from scratch. A company that hasn't done so has to rebuild it every time it changes tools—and ends up paying twice.
What this means when it comes to making a choice
The question to ask a publisher isn't just "What can your model do?", but also "Where does my context fit into your solution, and what do I get out of it if I leave?"
A context buried within a proprietary configuration is one that you created without actually owning it.
To put it simply: data tells you what you’ve sold, the model tells you what might work, and context tells you who you are. The first can be repurchased, the second can be replaced, and the third is built once and for all.
See also in this feature
- Before deploying a pricing tool, ensure your data is reliable
- Integrating Pricing Data Without Major IT Projects: Three Levels
- Eliminating Duplicates in Your Product Catalog: Ensuring the Reliability of Your Internal Database
FAQ
This is the collection of information that outlines what a retailer is allowed to do and what would be unreasonable for it to do: flagship products, supplier restrictions, unwritten prohibitions, and the intended strategy for each category.
It isn't written down anywhere, because he didn't need it as long as decisions were made by people who shared his views.
The data describes what happened: sales, recorded prices, inventory, and margins. The context explains what isn't happening here.
A missing data point results in a visible error. A missing context results in a recommendation that is technically correct but commercially unfeasible.
Because it optimizes what it measures and ignores what it hasn't been told. The most common blind spots: price benchmarks, supplier contracts, commercial restrictions, and the specific objectives of each category.
The typical symptom is a rejection of the recommendations without challenging the calculation.
Rather than a workshop where everyone writes their own rules, it’s better to start with rejections and corrections: every rejected recommendation, along with its one-sentence explanation, reveals a piece of context.
Presenting the team with concrete proposals also sparks discussion in situations where an open-ended question leaves everyone silent.
No. Turning every element into a blocking constraint quickly results in a set of rules that is unreadable and contradictory.
Part of the context must remain as information displayed at the time the decision is made. The blocking rule applies only to regulatory obligations and contractual commitments.
Not sustainable on its own: a general-purpose LLM does not retain context from one session to the next, and it must be provided with the context again for each request, with no guarantee that it will follow it.
The context must be incorporated into the tool that calculates prices, in the form of rules, limits, or displayed information.
Forecasting and optimization approaches are updated with each new version, every few years. A retailer’s benchmarks, constraints, and objectives evolve much more slowly.
A company that has formalized its context can switch to a different platform without having to start from scratch. That’s why every software vendor should be asked: What do we get back if we leave?
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.
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.
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.
