Pricing Proof of Concept: What Data to Prepare for reliable results?
October 20, 2026
A pricing proof of concept (POC) does not require perfect data, but rather data that is appropriate for what it needs to demonstrate, with known limitations. The product catalog, sales data, prices and promotions, purchase costs, and inventory levels form the foundation; the granularity and depth of this data depend on the test’s objective.
Before calculating a profit, the recalculated revenue, volumes, and margin must match those in management accounting. A known and documented error does not skew the comparison; a hidden error does.
Before testing a pricing solution, one question always comes up: Do we have enough data to get started? The answer depends on what you want to demonstrate. Verifying the consistency of a product line, simulating a margin increase, or optimizing end-of-season inventory all require different types of information.
This article is intended for pricing, data, and management control teams preparing a proof of concept (POC). It outlines the data to prepare based on the test’s objective, the checks to conduct with the finance team, and the limitations to document in order to start with a viable scope.

Short answer: The data used in a POC depends on what it is intended to demonstrate
The dataset for a pricing proof of concept (POC) consists of all the data extracts—documented and validated by the company—that enable a solution to reconstruct past business activity within a given scope and calculate pricing recommendations for that scope.
It doesn't have to be perfect. It must be suited to the test's objective, its limitations must be known, and the metrics used to measure results must align with financial metrics.
Checking the consistency of a product line, simulating a margin increase, or optimizing end-of-season inventory do not require the same information.
Five categories form the basis for analysis (product portfolio, sales, pricing and promotions, purchase costs, and inventory), but the depth and level of detail for each are determined based on the specific question being asked.
This is the most common mistake: demanding the absolute best in everything right from the start, at the risk of never getting started.
What data is required based on the POC's objective?
The first step is to write down what the test is supposed to demonstrate. The list of data to be extracted follows directly from this, and it is often shorter than expected.
| What you'd like to test | Data to Be Prioritized | Point to Watch For |
|---|---|---|
| Verify the consistency of prices and product lines | Catalog with attributes (format, capacity, brand), current prices, business rules | Reliability of Links Between Comparable Products |
| Analyze Competitive Positioning | Internal prices, outdated competitor data, product matching | Product Comparability and Timeliness of Data |
| Simulate the impact of price changes | Sales figures, prices charged, promotions, availability; historical costs for calculating margin | Actionable price changes and other factors that influence sales |
| Optimizing a price reduction | Remaining inventory, recent sales, prices, costs, sell-through deadline | Actual Availability and Sales Forecast |
| Measuring time savings and the traceability of decisions | Files and tools used today, pricing rules, roles, and approval workflows | Processing time measured before the test; otherwise, there is nothing to compare |
A single POC can combine several objectives, provided they are prioritized: the most demanding one determines the level of detail in the data.
The scope of the test itself (scope, baseline scenario, metrics, duration) is covered in our article on how to verify that a pricing POC creates value.
Is it possible to launch a POC using your data?
Share your file structure and the purpose of the test. Booper helps you—free of charge—identify actionable data, priority checks, and necessary additions before your teams begin the data extraction process.
Let's talk about the data from your POCThe Five Categories of Data: A Foundation, Not a Checklist
Product catalog, sales, pricing and promotions, purchase costs, inventory: these five categories are always considered together. What varies is the level of detail and depth required, and the test’s objective determines these.
Issues related to deployment (images, ticket lines, detailed back-end margins, rules from the old tool) shouldn't hold up the test. Demanding everything right from the start is the surest way to never get started.
The product catalog: the key to everything else
Every other set of data is linked to a product. An incomplete or inconsistent data repository therefore affects sales, inventory, and prices all at once.
What is essential in all cases is a unique identifier that remains consistent over time, a product name, a multi-level product hierarchy (universe, family, subfamily), the brand, and a status that distinguishes active products, new products, and end-of-life products.
The EAN code becomes necessary as soon as the POC matches your SKUs to external data: competitor price data, supplier price lists, and market surveys. For a test that uses only internal data, a stable internal identifier may suffice.
The Case of New Products
This is the most common shortcoming in pricing frameworks, and the one that proves most costly in terms of pricing. A product launched six weeks ago has no sales history, and therefore no observable price response.
A pricing solution can still handle it, provided it has attributes that link it to known products: format, size, product line, brand, and, if possible, the product it replaces.
A company whose product lineup changes rapidly (high-tech, baby products, toys) must treat these attributes as essential; without them, the POC will only test products at the back of the shelf.
Duplicates and code changes
A single product that exists under two different identifiers—or whose code has changed during the year—splits its history in two and skews any analysis.
Code deduplication and chaining of successive codes are described in our article on improving the reliability of product matching; for a proof of concept (POC), it is sufficient to document known cases and provide the mapping table when it exists.
Sales: Mesh Size and Depth
If the POC is to measure the effect of a price change, the relevant metrics are the quantity sold and revenue by product, by day, by retail location, and by channel.
A monthly or national aggregate does not allow us to isolate the price response, construct a control group, or distinguish between a downtown store and a suburban store.
For a POC focused on price consistency or competitive positioning, a more aggregated level of detail is often sufficient.
The receipt line is a valuable addition (it shows products purchased together and discounts applied at checkout) but is not required to get started.
A 24-month time frame is desirable in order to observe two complete annual cycles and two occurrences of each key event.
It adapts to the seasonality of the product range and the objective: twelve months may be sufficient for categories with little seasonality, and a test limited to a few product families does not require the same level of detail as a simulation covering the entire product assortment.
Promotional sales must be clearly identifiable
A sale made during a promotional event does not measure the same thing as a sale at the regular price.
If the sales data does not include a promotion indicator, the promotional calendar can be used to reconstruct it, provided that the data is specific to the product and the store.
Without either of these, promotional volumes will be allocated at the regular price, and the price response will be overestimated.
Supply shortages mask actual demand
A product that is out of stock does not sell, and its sales history shows “zero” where demand actually existed. This bias, known as censored demand, degrades forecasts and distorts the measured effect of price.
It is adjusted based on inventory data, which explains why the two categories go together.
67% of data and analytics leaders do not fully trust the data their organizations use to make decisions, and only 12% consider their data ready for artificial intelligence.
(Precisely and Drexel University’s LeBow College of Business, 2025 Outlook, Data Integrity Trends and Insights, 565 respondents, 2024)
This lack of confidence has a practical implication for a POC: until the company has verified the data extracts itself, it cannot distinguish between a flaw in the solution being tested and a flaw in its own data.
The checks described below are used for this purpose.
Inventory: The Most Difficult Category to Restock
Management systems track current inventory but rarely keep a history of it: a product's inventory level in a warehouse fourteen months ago is not stored anywhere, unless a regular snapshot has been archived.
This is the most common shortcoming encountered when preparing a POC, and one of the reasons people mistakenly give up.
The available frequency determines what conclusions the test can draw. A daily inventory allows for the precise dating of stockouts. A weekly inventory indicates whether the product was available each week, which is generally sufficient to correct the censored demand.
A monthly inventory does not allow for the detection of a stockout that occurred between two inventory counts; it can still be used to track remaining inventory, but its limitations must be documented.
If no snapshot has been archived, the historical inventory is reconstructed retroactively based on the current inventory and inventory movements, with accuracy decreasing as one goes further back in time.
For a markdown POC, this becomes essential: everything depends on the remaining inventory and the sale deadline.
Acquisition Costs and Valuation: A Decision to Be Made with the Finance Department
Without reliable purchase costs, a POC cannot demonstrate a reliable margin increase. It can, however, test price consistency or competitive positioning.
To calculate the margin, the minimum data required is the net purchase price per product and per period, taking into account invoice discounts.
Conditional and year-end discounts, logistics costs, and back-end margins can adjust the margin without blocking a test, provided that their exclusion is documented and consistent across the entire scope.
Weighted average purchase price, last price, or standard price?
This is the most hotly debated issue among the pricing, procurement, and management control teams, and it must be resolved before the POC, not during it.
The weighted average purchase price (WAP) smooths out price increases and reflects the value of the inventory sold; it is generally the figure used in accounting.
The most recent purchase price reflects the replacement cost and carries greater weight in pricing decisions during periods of rising prices. The standard price simplifies matters but deviates from reality.
No rule is wrong, but only one should be applied—the same one used in management accounting—otherwise the margins calculated in the solution will never match those in finance.
Allocate costs over the entire period
To calculate the margin realized in the past, you need the purchase cost at that time, not today's cost. The cost is used to measure profitability; it says nothing about customer response to price, which is estimated based on sales, prices, and promotions.
When the supplier pricing history is not maintained at the product level, it is reconstructed based on invoices or, if those are unavailable, on a “ Markup ” by product family approved by the finance department.
This approximation is acceptable if it is explicit; it is not acceptable if it is implicit.
Price and Promotion History: The Actual Price Charged
The pitfall here is providing the theoretical rate (the national price list) instead of the actual price charged at each point of sale and through each channel.
In between, there are local adjustments, price matching with competitors, labeling errors, and in-store discounts. A solution that learns from the theoretical price will estimate how customers will react to prices they’ve never seen before.
The ideal price is the one applied per product, store, and day, as shown in the point-of-sale system or the price change history.
Otherwise, revenue divided by quantity yields an average selling price: while this is useful, it may include a mix of different prices over the period, as well as discounts and returns.
The calculation guidelines (excluding tax or including tax, with or without discounts, and how returns are handled) must then be set out in writing.
Added to this is the promotions schedule, including their mechanics and scope. The history of the rules from the old tool can wait: it's relevant to deployment, not testing.
Market data: necessary depending on the objective
Competitor price surveys are not an optional extra when the POC focuses on price positioning: they are the raw material, along with the product matching that links your SKUs to those of your competitors and the date of each survey.
They can be collected online through the platform or integrated from surveys and panels already purchased by the company (our article on the differences between data scraping, field surveys, and panelists explains what each one offers).
For a POC focused on margin or markdown, they remain useful without being a prerequisite for getting started.
Market share by category shows how a business unit is performing relative to its market and helps prevent a decline in volume from being mistaken for a market downturn.
Web metrics (traffic, conversion rates by product) are important for companies where the online channel plays a significant role.
These sources are often already purchased and used within a single department; a POC is a good opportunity to make them available to the pricing team.
Consistency checks: measuring the same thing as the finance department
Before measuring a gain, you must verify that you are measuring the same thing as management control.
This is the step that distinguishes data extraction from a data foundation: recalculating, based on the provided files, the aggregates that the finance department already publishes, and explaining each discrepancy.
It usually takes a few weeks and avoids months of debate over the reliability of the POC results.
Align the definitions before comparing the amounts
Most discrepancies are not due to errors but to different definitions.
Before any reconciliation, the following must be specified in writing: whether prices are exclusive of tax or inclusive of tax, how returns are handled, cash register adjustments, loyalty benefits, the reference date for a sale (order date, delivery date, or payment date), and the method used to calculate costs.
Tolerances set by the finance department, not a priori
Revenue, volumes, and margins recalculated from the files must match the management reports, within the tolerances set by management control.
Every discrepancy is explained: a minor overall discrepancy may mask a significant anomaly in a strategic category.
Illustrative example: a margin gain that wasn't really one
This is a fictional example created for illustrative purposes. Within the scope of a POC, the margin recalculated for the reference year differs by only 0.3 points from the reported total. However, when broken down by department, the small appliances department shows a difference of 1.6 points.
The reason: The cost file contains only the current purchase price, which is applied to past sales, even though the department's main supplier has raised prices twice in the meantime.
The margin for the reference period is therefore underestimated, and the solution being tested would have shown a gain that stemmed solely from the incorrect cost. Adjusting the costs eliminated the discrepancy before the test began.
The total difference was small; it was the reading by section that revealed it.
Rejects by load, by line, and by economic weight
During the import, some of the records are not linked to anything: sales for a product not in the database, purchase prices with no product, or inventory at a closed store.
These exclusions are measured by data family, not just by the number of rows: a few excluded entries may account for a significant portion of revenue or may specifically involve the strategic products being tested.
Measuring bounce rates by line, by revenue, and by category allows us to answer the question, “Is our data good enough to get started?” based on more than just a hunch.
43% of data directors cite data quality, completeness, and readiness as the main obstacle to their AI projects, and 67% struggle to move a pilot project into production.
(Informatica, CDO Insights 2025, survey of 600 data directors in North America, Europe, and Asia, January 2025)
At Booper, this reconciliation with management accounting is part of the data integration process: the platform loads the available data extracts, tracks rejections, and the test does not begin until the revenue, volumes, and margins for the reporting scope have been reconciled with those from the finance department.
The page on data integration and workflow monitoring describes how this works, and our article on integrating pricing data without an IT project shows that it doesn't require any connectors to get started.
What Really Holds Up a Test—and What Can Be Fixed Along the Way
There are three situations that justify postponing a POC: a data model without a stable identifier, data that does not meet the objective, and metrics that do not align with financial metrics.
We prefer to start a POC using imperfect but well-documented data rather than wait for the data to be fully cleaned. A known defect, which is identical in both the test set and the control set, does not skew the comparison.
A hidden flaw, yes. The only one that justifies waiting has to do with the measure itself: margins that don't align with those in the financial sector, and no one can say why.
The rest (inventory to be replenished, partial new-item attributes, promotions to be linked, estimated costs) is handled during the test, provided that it is documented and identical in both the test and control sets.
Comparing Multiple Editors: Same Data and Same Protocol
When a company compares several solutions, the dataset is the first prerequisite for a fair comparison.
Once prepared and reconciled with management accounting, it is provided in identical form to each publisher along with its documentation: file descriptions, mesh size, depth, valuation rules, known rejections, and documented defects.
But that is not enough: the objectives, price constraints, evaluation period, and metrics must also be identical. Only when both of these conditions are met can differences in results be attributed to the solutions.
This foundation has a second benefit: it outlasts the POC. Whether or not a solution is selected, the company has a documented pricing dataset that has been reconciled with the finance department—something it often never had before.
Many people discover on this occasion that the problem they had attributed to their existing tool was partly due to the data feeds that powered it.
See also in this feature
- The Booper Solution: Booper Pricing Software
- Selecting and Implementing a Pricing Tool: From Identifying the Need to Going Live
- Alternatives to Pricefx, Competera, and Revionics: What Are the Options in France?
- SaaS Platform for Retail Pricing Management: The Benefits
- Food Retailer Case Study: Managing Prices Across a National Network
FAQ: Data from a Pricing Proof of Concept (POC)
It depends on what the test is intended to demonstrate. There are five key areas to examine: the product specifications, sales, pricing and promotions, purchase costs, and inventory.
A proof of concept (POC) on price consistency primarily draws on the price list and current prices.
A proof of concept (POC) on competitive positioning includes dated competitor data and product matching.
Simulating the effect of a price change requires detailed sales data, promotions, and inventory availability, and measuring a margin gain requires historical purchase costs. A markdown is based primarily on remaining inventory.
No. The data must be appropriate for the test's objective, and its limits must be known and documented.
There are three situations that justify postponing a POC: a data source without a stable identifier; data that does not allow the question at hand to be answered (sales data that is too aggregated to measure a price effect, or missing cost data when the test is intended to demonstrate a margin gain); and metrics that do not align with those used in management accounting without an explanation.
The rest is handled during the test, provided that it is identical in both the test and control groups.
A period of twenty-four months is desirable in order to observe two complete annual cycles and two occurrences of each key event, which helps distinguish the effect of price from that of the calendar.
This depth varies depending on seasonality and the objective: twelve months may be sufficient for categories with little seasonality.
The sample size is just as important as the depth of the data. To measure the impact of a price or build a control group, you need the product, the day, the point of sale, and the channel, along with identifiable promotional sales.
Yes, by reconstructing it retroactively based on current inventory and recorded transactions (receipts, sales, returns, inventory counts, markdowns), with accuracy that decreases as we go further back in time.
Before restocking, it's worth checking for archived snapshots in the data warehouse, management reports, or the replenishment tool.
Frequency matters: a weekly inventory is generally sufficient to identify stockouts, whereas a monthly inventory is not. For a markdown proof of concept (POC), an inventory is essential.
Whichever method is used by management accounting. The weighted average purchase price smooths out price increases and reflects the value of the inventory sold; the most recent purchase price reflects the replacement cost; the standard price simplifies the calculation.
There should be only one rule—the same one used in finance—and costs must be time-stamped: applying current costs to past sales distorts the margin for the reference period and can create the appearance of a gain that does not exist.
The decision must be made in writing before the test.
First, by standardizing the definitions (excluding or including tax, returns, discounts, loyalty benefits, allocation dates, and cost valuation), and then by recalculating revenue, volumes, and margins—by month and by product category—based on the provided files, to align them with management reporting.
Tolerances are set in consultation with management, and every deviation—no matter how small—is explained. Rejections during loading are measured in production lines and revenue, since even a few excluded SKUs can have a significant impact.
Yes, and that is the first requirement for a fair comparison: a base system set up once, aligned with financial systems, and provided identically to each publisher along with its documentation. But that is not enough.
The objectives, price constraints, evaluation period, and metrics must also be identical; only when both of these conditions are met can differences in results be attributed to the solutions.
This base passes the test and remains useful for deployment.
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.
.avif)