Automating internal supply orders

Every month, the 174 Sam’s Club Mexico wholesale stores ordered the supplies they run on through a macro-driven Excel file passed around by email. Consolidating those orders and generating purchase orders relied on manual work spread across several teams.

I led research and design for a module that automates that process. I mapped an undocumented workflow from 14 interviews with 20 participants across nine teams, designed the solution and validated it in 7 usability tests. The module launched in production and was in the adoption phase when the project closed.

Internal supplies are the materials a store uses to operate and never sells to customers: trays, domes, bags and packaging for the bakery and fresh departments. They are operating expenses, not stock for resale.

Each month, the 174 wholesale stores ordered these supplies by filling out a macro-enabled Excel file circulated over email and Teams. The process depended on several teams and on a support team that had shrunk from 16 people to 4.

Ordering supplies took five stages. Operational efficiency, the team that set each store’s budget, updated the catalogue by hand. Internal communications sent it to the four district assistants, who looked after about 45 stores each on average; each assistant passed it on to their stores and followed up until every completed file was back. Operational efficiency then consolidated everything so that replenishment, the team with access to the purchasing system, could run the macros and generate the purchase orders.

Nobody had this process documented. I mapped it through interviews with the teams involved to understand who did what, in which order and how long it took.

The main cost was in consolidating orders and generating purchase orders:

“From Monday, November 4 to today, the 14th, I’ve put in 43 hours.”

The file was also shared, so one store could overwrite another. Files arrived incomplete and each category ordered in different units, so entering kilos where pieces belonged was common.

With spending of about 26 million euros a year, some stores had stockouts while others were overstocked at the same time, with no clear way to measure either.

I worked with product, developers, business analysts, content design and operations stakeholders. Decisions weren’t made alone: my responsibility was to bring the evidence that steered the discussion.

My research and design leads mentored me and reviewed the approach and deliverables. That gave me autonomy over day-to-day decisions and the responsibility to defend them with stakeholders.

Starting hypothesis: replenishing more often would improve supply and budget control at each store.

Sources: 14 interviews with 20 participants from nine teams, sessions with product and stakeholders, a review of the operational process, analysis of the Excel file and an understanding of the business rules.

I used AI to speed up the analysis of the interviews and generate per-team syntheses and draft findings, always reviewing every result. Because I ran the interviews and took part in the stakeholder sessions, I could quickly spot anything that didn’t match what I had observed and fix it. I also used it to explore design alternatives based on the findings, and some ideas became starting points for discussion with the UX team.

The project arrived with the assumption that internal supplies would sit inside the existing fresh-produce flow. It made business sense, but it had never been validated with users.

I validated the solution with 7 usability tests across three flows: 4 users on order creation, 2 on approvals and 1 on configuration.

In testing, 3 out of 4 users of the ordering flow couldn’t find the tool where it was. To them, packaging was inventory, not fresh produce. That is why the solution became a dedicated module.

The tests helped make key decisions:

Going over budget was normal and expected: approvals by amount already existed, but they happened over email. The design brought that rule into the product with a tiered approval flow, from store manager up to district deputy director depending on the amount, with the justification visible to the approver.

An order can’t be submitted until it passes validation. Out-of-range items are flagged right in the cell.

The test sample was smaller than ideal because of operational availability, so some recommendations were left as future validations.

When my involvement ended, the module was in the adoption phase, so this case documents the process design and the decisions behind it, not measured business results.