Danilo Soto
Founding Partner EREA Decisions Lab. A Division of EREA Consulting Group
Every time we work with a retailer to prepare category and business reviews with suppliers, we encounter the same problems: products are purchased from multiple distributors, brand names contain errors, and back margin data is fragmented across Excel files that are complex to reliably link to the corresponding products, suppliers, and periods. Purchase orders, which appear to be clean records, are often another unreliable source of data: orders that remain open and receipts that cannot be cross-referenced with commercial terms.
These failures become most apparent when the retailer sits down to negotiate. Preparing a fact-based negotiation brief for a business review requires consolidating sell-in, sell-out, front margin, back margin, agreements, inventory and service levels. This exercise is usually manual, slow, and difficult to replicate. Adding to the complexity: when the supplier belongs to a group where one company sells the products, other buys services from the retailer, and a third signs the agreements, consolidating the entire relationship requires connecting entities that the system treats as if they had nothing to do with each other.
In this context, the retailer comes to the negotiating table in a weaker position than the supplier. When attempting to automate the brief, the result is not a reliable brief, but one that appears reliable. This difference ends up shaping the commercial discussion. What many organizations mean by data governance is often reduced to defining concepts: what is an active product, what is a valid purchase order, how is the margin calculated. That is necessary, but it is not enough. Defining concepts without deciding who has authority over them, how they are modified, how errors are detected, and what is corrected first is like building a glossary that will be ignored.
In practice, these tensions surface when the brief is used in negotiations, and the conversation ends up focusing on three points. First, scope: which sales are in-scope and which are excluded: returns, credit notes, special sales, and consignment. Second, timing: when conditions and penalties are recognized, accrued versus paid, sale period versus settlement period. Third, the evidence: which source is accepted as a reference when there are differences: the ERP, the supplier portal, the settlement sheet, or the claims. When these criteria are not resolved in advance, the discussion shifts from the result to how the numbers were derived.
Definitions alone do not work if there is no clear responsibility for them. That is the point at which the problem ceases to be analytical and becomes organizational. The problem does not arise when everything fits together, but when an exception, a change, or a discrepancy between areas arises; at that point, someone has to decide which criterion applies, with what impact, and as of when. In many organizations, that decision has no owner: IT implements what it can, the business adjusts what it needs, and definitions get stuck in that middle ground without an authority to sustain them over time.
In practice, this lack of ownership is compounded by how systems and processes are designed. The business captures data without clear incentives to do it well, or visible consequences for doing it poorly; IT, for its part, prioritizes flexibility and speed of implementation, with few controls over what the user can enter. Data quality suffers as a result.
Definitions and owners are not enough if there is no flow to manage changes: a distributor stops carrying a brand and another takes its place, an agreement is renegotiated in the middle of the year, a brand name is corrected only in some records. Without a process that defines the request, review, effective date, and communication, changes still occur, but haphazardly. Months later, the numbers no longer add up, and no one knows why.
The final piece is detection. Without mechanisms to alert when a rule has been broken, errors accumulate: products associated with the wrong distributor, purchase orders open for months, or brands with multiple variants are often only discovered when someone puts together a brief and time is running out. These anomalies require detection rules, an audit trail, and clear resolution deadlines.
Without data governance, putting together a brief for a group of suppliers becomes a project in itself: the sales team requests data, the analytics team cross-references sources, inconsistencies appear, they are escalated, and compromises are made for lack of time. The result is used, but without a defensible position at the table.
At EREA Decisions Lab, we recognized this challenge for our clients and developed a system that allows briefs to be built with all these variables taken into account.
«Retail is Detail,» and without data governance, building an accurate «negotiation brief» is not possible.
With data governance, the same brief is put together by querying existing definitions and pre-linked data: suppliers are associated with their manufacturer and commercial group, brands are standardized, back margins are distributed, and the relevant entities are connected. The team devotes it’s time to preparing the negotiation strategy. If someone questions a figure at the table, the discussion is about what to do with what the data shows, not whether the data is correct.
Data problems do not disappear there are still data entry errors, inadequately recorded changes, and exceptions. What changes is that there is a system to resolve them: clear responsibilities, defined criteria, controls that detect them, and flows that close them. That is where data governance ceases to be a concept and becomes an operational advantage. A common mistake is thinking that all of this must be resolved before starting. In practice, governance is built along the way, but the sequence matters: starting with the domain where the lack of governance generates the highest cost and expanding from there is what works. Attempts to govern everything from the outset tend to lose momentum.
Technology can help, but it does not replace these decisions. An AI agent operating on ambiguous data does not make better decisions: it just executes the same inconsistencies faster. Tools simply end up reflecting the state of governance that already exists.
At EREA Consulting Group, we have been working with retailers in Latin America for more than 18 years. In our experience, data governance does not make it onto the agenda until something goes wrong, and in retail, it almost always comes down to negotiation. A simple way to understand where an organization stands is to try to put together a fact based negotiation brief today with what it has; the time it takes and the confidence it generates provide a clear indication of where things stand.
In retail, the state of data governance always reveals itself when it comes time to sit down and negotiate.
