Automotive parts distributor reordered to supplier reality, not calendar habit

An automotive aftermarket parts distributor carried about 1,600 SKUs from more than 40 suppliers, with seasonal demand and promotion-driven spikes. Buyers reordered on a calendar rhythm, monthly for most suppliers, and stockouts hit whenever a single supplier slipped by two weeks. When supplier lead time was measured from actual receipt history instead of the supplier card, reorder timing changed line by line: supplier-related stockouts fell 24% in one season and €52,000 of excess buffer came off the shelves.

The situation

Two branches, a central warehouse, and a purchasing team of three covering brakes, filtration, lighting and seasonal lines such as wiper blades and batteries. The ERP was a legacy system that the original integrator no longer supported, with a thin REST API and no reporting layer beyond exports.

Safety stock fields had been set years earlier and never revisited. Every supplier record carried one lead-time number, typically the figure quoted at contract signature. Buyers placed orders on a fixed cycle because it was the only rhythm they could manage across 40 suppliers by hand.

The result was predictable. Reliable suppliers were over-buffered, tying up cash. Unreliable suppliers were under-buffered, and a two-week slip on one of them emptied a bay before anyone noticed.

What the data showed, and didn't say

Purchase-order and goods-received history contained the real lead-time distribution for every supplier-SKU pair: not one number, but a median and a spread. One filter supplier delivered in 11 to 13 days, every time. A lighting supplier's median was 19 days with a range from 12 to 41.

The ERP never combined that variance with sales velocity to suggest how much buffer each line actually needed. It also never distinguished the supplier that was late once from the one that was late every third order.

Connecting Flowra

Because the ERP had no purpose-built connector, the team used Flowra's custom REST API connector: read-only credentials and a field mapping configured together with the Flowra team in one session. Three years of purchase-order and receipt history, which the API did not expose, were loaded once as a CSV export and mapped automatically.

Flowra reconstructed the stock trajectory per SKU and branch from the transaction flow and computed lead-time bands per supplier from the receipt dates. The first supplier risk panel and reorder review were ready within 48 hours; refreshes run nightly or on demand.

What Flowra surfaced

• Reorder points recomputed per line using average daily sales × lead time plus a safety term of 1.645 × σ × √lead time, targeting a 95% service level, with the supplier's actual median lead time in place of the card value.
• A supplier risk panel for the weekly purchasing call: median lead time, variance band and the number of open POs already past their expected date.
• Buffer recommendations on high-variance suppliers, and buffer reductions on the stable ones, each with the cash impact at cost.
• Expedite flags only where margin and velocity justified express freight, so the team stopped paying for air shipments on slow, low-margin lines.

What the team did

Buyers reviewed the reorder-point changes in three batches over three weeks, supplier by supplier, and applied the ones they accepted in the ERP themselves. Flowra never wrote to the legacy system. On stable suppliers, buffers came down; on the two suppliers with the widest variance, buffers went up and the commercial lead opened a delivery-performance conversation with evidence in hand.

The weekly purchasing call switched from anecdote ("they were late again") to the supplier panel. Alerts on open POs past their expected receipt date arrived by email each morning, so the team learned about a slip on day one rather than when the bay was empty.

Results

"Our supplier cards said 14 days. The receipts said 19, sometimes 41. Once we reordered against the receipts instead of the card, the surprises stopped."

— Purchasing Manager, automotive parts distributor

What made it work

• Lead time from receipts, not the supplier card. The single most valuable input was already in the ERP: PO date and receipt date, per line. The full method is in when to reorder: days of cover vs supplier lead time.
• Variance decides the buffer. A 19-day median with a 41-day tail needs more safety stock than a 13-day supplier that never slips. One global safety-stock rule cannot express that.
• A legacy ERP was not a blocker. A read-only REST connection plus a one-time CSV was enough. No migration, no write access, and the buyers stayed in their own system to place orders. The four-layer model keeps the human in the loop by design.