Dynamic pricing for used parts without giving up control

In the new parts business a price is a calculation: purchase, margin, done. With used parts every item is unique, with its own condition, origin and shelf time — and the market often consists of three offers, two of which have been standing for months. Automating that needs fewer compute cycles than rules that still make sense when the data is thin.

The six quantities a price is built from

• Identity. Without a confirmed OE number you compare apples with pears. It is the precondition, not one field among many.
• Condition. A grade with criteria, not “good”. An A part and a C part of the same number are two products.
• Market situation. Existing offers for the same number, with the number of observations — not just an average.
• Acquisition. What did the vehicle cost, what the removal, what the storage? Without that figure any floor price is a guess.
• Stock and shelf time. Three identical parts on the shelf is a different case from the last one.
• Channel. Marketplace fees, shipping costs and return rates differ markedly per channel.

Why a range has no midpoint

The price evaluation GET /parts/oe/{oeNumber}/price returns min, max and average plus priceRecommendation with confidence as HIGH, MEDIUM or LOW. What matters is result: priced means usable offers were found, no_listings_found means there were none — the response then carries zero values, and that is not a price but an empty finding. With LOW the range rests on few observations; its midpoint then carries no decision, but its lower end does. Automated pricing therefore works from the minimum, not the average — the same caution with which the economic evaluation forms its purchase recommendation.

A workflow that holds

1. Confirm identity and grade condition before anything is calculated.
2. Fetch the market situation and store result and confidence with it — not just the number. A recommendation without its reliability cannot be assessed afterwards.
3. Build the floor from acquisition, removal, storage, shipping and channel fees. That figure is internal and does not come from the market.
4. Set the price as the maximum of floor and market-based proposal, capped by the maximum change per step.
5. Stage by shelf time: fixed steps by days in stock, not daily micro-adjustments. The latter cost calls and create noise in the channel.
6. Measure the outcome: sell-through, margin after fees, returns, time to sale — per part group, not as an overall average.

What belongs to it legally

Towards consumers, German price indication rules require a clear, attributable total price; anyone showing struck-through prices or discounts must state the reference price correctly. Misleading statements about price, availability or essential characteristics are unlawful under unfair competition law — including when an automation produced them. And since 12 September 2025 the Data Act (EU) 2023/2854 has applied: it governs access to data from connected products, supplemented for the automotive sector by dedicated Commission guidance. For pricing this means less that new data sources are permitted than that the origin of the data you use must be evidenced.

Limits

• An evaluation is not a price guarantee. The output is explicitly indicative.
• Market data is not the market. It shows offers, not transactions; an offer standing for a year distorts every statistic.
• No automation on safety parts. There the inspection decides, not the price.
• Coordination is prohibited. Pricing rules are internal; aligning prices with competitors is a competition-law risk.
• Every lookup costs. Price evaluations belong at decision points — initial pricing, shelf-time steps — not in a daily loop across the whole stock.

The test for a pricing system is not how often it changes prices but whether the sales director can say, for any price, how it came about. If that question has an answer, the automation is a tool; if not, it is a risk.