How to build a dealer portal that dealers open every week

The graveyard of enterprise software is full of portals that were technically functional and completely ignored. When regional managers do not open a dealer portal, the problem is in the design. The portal that gets used is built around the three questions a dealer asks every week, explains its recommendations in plain language, and turns every screen into a workflow action.

This is an Operator Note from our work building Disandina's dealer management platform for the Giant Bicycles distribution network in Colombia. Giant sells through regional dealers across Colombia's major cities and departments. Before our engagement, dealer ordering ran on email threads and WhatsApp groups. Regional managers had no visibility into which models were moving, which dealers were underperforming, or where inventory was piling up.

What goes wrong with email-based dealer ordering?

Email-based dealer management creates three compounding problems. There is no audit trail: when a dealer claims they ordered 20 units of a model that is now out of stock, there is no reliable way to verify it. There is no demand signal: the distribution team sees what dealers order, not what they sell, and orders lag demand by weeks. And there is no early warning: a dealer whose orders drop 40% over two months might be switching to a competitor, and the signal stays invisible until it is too late to act.

What do dealers need from a portal?

We interviewed eight regional managers before writing a line of code. The finding that changed the product direction was that dealers did not want another system to log into. They wanted faster answers to three questions they asked every week:

• Which models should I reorder this week?
• What is my sell-through rate on the units I have in stock?
• How am I performing compared with dealers in similar markets?

The portal is organised entirely around those questions. The homepage is a reorder recommendation list rather than a dashboard of metrics. Every screen answers one of the three directly. There are no reports to configure, no date ranges to set and no exports to run.

How does the TensorFlow reorder model work?

The reorder engine uses a TensorFlow model trained on 18 months of sales data across the dealer network. Its inputs are current inventory by SKU, trailing 30-day sell-through rate, a regional seasonality index (cycling season in Colombia varies a lot by altitude and climate zone) and the dealer's historical reorder lead time.

The output is a ranked list of reorder recommendations with a confidence score and a suggested quantity. We deliberately kept the model out of the black box. Each recommendation carries a plain-language explanation: "You have 3 units of the Giant Contend AR 2 in stock. At your current sell-through rate you will be out of stock in 11 days. Regional dealers typically reorder 7 to 10 days before stock-out. Recommended reorder: 8 units."

Operator insight: the plain-language explanation is the adoption mechanism. When a dealer understands why the system recommends what it recommends, they trust it enough to act. When it is just a number, they ignore it and reorder on instinct. After three months, 71% of dealer reorders followed the model's recommendations. Before the portal that number was unmeasurable because there were no recommendations, but dealer stockout incidents dropped by 34%.

Which design principle made it work?

Every feature in the portal has a direct workflow equivalent. Accepting a reorder recommendation creates a purchase order in the Shopify B2B backend automatically. Viewing sell-through rates links to the specific SKU pages where performance is low. The comparison view with peer dealers is opt-in rather than surfaced by default. Dealers are competitive, and they need to choose to engage with that data.

A B2B portal built for the brand's reporting needs will have great dashboards and zero adoption. A portal built around the dealer's daily workflow will have simpler screens and become indispensable. Build for the person who opens it at 8am before the shop opens, not for the executive who looks at it once a quarter.