Midwest steel distributor

From 1,000 purchasing decisions a day to a review-and-approve queue

A mid-size steel coil distributor in the Midwest supplies its sister roofing manufacturer and about 400 outside customers, and it grows more than 30% a year. With lead times of three to six months, every buying mistake shows up a season later. Convexity forecasts each item and recommends what to buy, within the cash the business has set aside.

Illustrative diagram of the workflow described below.

Impact

~10% less excess inventory — on a multi-million-dollar coil position
Fewer stockouts — item-level forecasts flag spikes like the July that came in at double the plan, months before stock runs out
Buyer time saved — about 1,000 manual purchasing decisions a day become a queue of recommendations to approve
Faster purchase orders — draft orders and transfers by location, edited by the buyer and written back to the ERP
Better decisions — spend capped by period and service levels set by item class, so cash goes to the items that matter most

The problem

One buyer making about 1,000 purchasing decisions a day across two screens
Monthly plans built as last year’s sales times 1.3
Lead times of three to six months, so mistakes surface a season later
Strong seasonality and no sales and operations planning
Cash is limited, so buying everything is not an option

Convexity features used

Demand forecasting →Item-level forecasts that model seasonality and swings, over three- to six-month lead times.
Inventory optimization →ABC/XYZ classes, service levels, safety stock and min/max per item, within a budget cap.
Purchasing workbench →A prioritized queue of recommended orders, sent back to the ERP once approved.
AI assistant →What-if questions (a demand spike, a tighter budget, a longer lead time) answered in plain English.

Company

Location: U.S. Midwest
Industry: Construction materials, steel coil
Customers: A sister manufacturer and ~400 others
Lead times: 3 months domestic, 6 months import

Book a demo

Before Convexity

One buyer made around 1,000 purchasing decisions a day across two ERP screens. The plan for each month was last year’s sales times 1.3, with no view of seasonality or of which colors and gauges were taking off.

Then one July, demand came in at double the plan and inventory ran out. With domestic steel taking about three months and imports about six, there was no quick way to catch up.

What Convexity does

Forecasts for every item replace the flat growth factor and show a breakout months before the stock runs out. Each item gets a class, a service level, safety stock and min/max levels, and spending is capped by period so the most important items are protected first.

The buyer works from a prioritized queue of recommendations and approves each one before it goes back to the ERP. The recommendations run beside the old process until the team trusts them.

Results and what’s next

The buyer moves from about 1,000 manual decisions a day to reviewing a short queue; demand breakouts get flagged months earlier; and a conservative 10% less excess inventory frees cash for the items that sell.

FAQ

Questions about the Midwest steel distributor

It assumes every item grows at the same rate and ignores seasonality. When demand came in at double the plan one July, the lead times of three to six months meant there was no way to catch up.

No. It recommends purchases within a budget, and the buyer approves each one before it goes back to the ERP.

Start with your workflow

Bring one inventory, purchasing or order-management problem. We will map the required data, demonstrate the relevant workflow and identify what needs configuration or further validation.