Demand forecasting: platform, build, or implement. Where each path wins.

To implement demand forecasting for an 80-store chain, there are three paths. Buy a platform like RELEX or SAS. Build internally. Hire someone who implements.

The question we hear most, “which is best”, is the wrong one. None is best. Each wins under a different condition, and the condition is the operation, not the software.

The three paths, as cost shapes

The difference between them is not in the features. It is in the shape of the cost and in who is responsible afterward.

Buy a platform. Recurring annual license cost, predictable, high. It turns on fast. The forecasting criterion is the product’s, tuned for the average of its clients, not for your operation. When the vendor changes price or strategy, you inherit the change.

Build internally. Cost in head count, a dedicated data team. Variable and slower. The system becomes a house asset and a permanent house responsibility. Every improvement depends on who is on the team that month.

Hire an implementer. Project cost, fixed by scope, with a deadline. The internal team keeps the operation and the runbook at the end. The forward-deployed model: someone sits inside, builds against your data, and hands back a running operation, not a slide.

The decision tree

The deciding condition is not technical. It is an operations question: who operates this system on Black Friday 18 months from now?

• If the answer is “the vendor”, buy the platform. You want the system maintained by whoever sold it.
• If the answer is “the consultant”, you never finished buying. You will pay again at every adjustment.
• If the answer is “my internal team, with a runbook and criterion written before the code”, you want an implementer.

The trap in each path

Every path has a failure mode, and the honest move is to name all three.

The platform fails when the license is signed and the internal team does not use it. You buy the ruler and keep measuring by eye. The build fails when the project becomes an ownerless internal product, and the key person resigns with the knowledge in their head. The implementer fails when the criterion stays vague and the engagement stretches, month after month, with no number that says it is done.

That last one is our failure mode, and the defense against it is the criterion written in the first week. Without it, any of the three becomes a hole.

What the choice actually decides

Take the forecasting decision on your desk right now. Ask the team which of the three paths you assumed the last time it was discussed. Most of the time the answer is uncomfortable: nobody chose, the operation drifted toward the option that felt least scary at the time.

Before choosing, it helps to understand how much of the forecast is estimate and how much is discovery. And it helps to remember that aggregate MAPE is not the acceptance criterion for any of the three. The criterion is shelf-out and shrink on the category that pays the bills.

Tell me the shape of the problem, chain size and category. In one hour I will send back which of the three paths fits your operation, with no bias on my side. If it is a platform, I will point you to the three names that show up in RFPs in this space. If it is an internal team, I will tell you the profile to hire. If it is an implementer, the next conversation is the two-week Diagnóstico, which ends with criterion, timeline, and cost defined in the same room, signed before the code.