"What if" is answered by the backend

The dollar gains 10%: is the SUV still profitable? Margins, scenarios, price tiers and sensitivity curves, all computed server-side, so the browser never subtracts two amounts.

Project
STP: Vehicle Cost Platform, part 2 of 10
Published
October 5, 2026
Reading time
3 min read

"The dollar gains 10%. Is our SUV still profitable?" At most carmakers, the answer to that question comes from the next ERP run, or from a spreadsheet nobody else can verify. In STP it takes one click, and the backend computes it.

From "how much?" to "what now?"

A cost on its own doesn't lead to a decision. Engineers need to know whether the cost is acceptable, and what would change it. So STP grew three things on top of the cost:

  • a margin against the market price;
  • scenarios that change one or more levers and compare against the base;
  • sensitivity curves that sweep one parameter.

They all follow the rule the rest of the product follows: the backend computes, the UI displays, and the browser never subtracts two amounts (never add two currencies explains why).

Margin against the market

Each vehicle belongs to a market segment, for example the "compact hybrid SUV" subcategory of segment C, and each subcategory has a market price set by an admin. The backend converts that price into the requested currency and subtracts the exact, unrounded cost. The UI turns the margin rate into a verdict: Profitable at 10% or more, Fragile below that, Not profitable when the cost exceeds the market price.

The backend computes the margin; the interface only turns its rate into one of three verdicts.

Levers

A scenario starts from the configuration on screen and changes one or more levers:

  • a currency's value against the euro (+10% means the rate is divided by 1.1, so parts paid in that currency cost 10% more);
  • the price of one part;
  • a replacement part, with its quantity per vehicle;
  • removing a part, along with everything below it;
  • another supplier for a part;
  • the annual volume.

The backend computes the scenario next to the base and returns both, with the difference in cost, weight, CO₂e and margin, and the parts that changed the most. Scenarios can be saved under a name, attached to the vehicle model, and reopened on any of its versions. A good one becomes a new working BOM in one click, and the published version it started from never changes.

What the backend returns for a scenario: the base and the scenario side by side, with the difference on every figure. Costs derived from the demo's margins and its 42,550 EUR market price.

Volume and price tiers

Supplier offers carry price tiers from the ERP: a lower unit price from some annual quantity onward. The quantity that counts is the part's, not the car's, so the same volume moves each part by a different amount. At 10,000 cars a year, a part used four times per car is bought 40,000 times, just under a 50,000 tier. At 15,000 cars it's 60,000, past it. Given a volume, each part takes the price of the tier it reaches, and choosing the cheapest supplier compares prices at that tier.

A tier counts the part's units, not the car's: cars a year times how many each car uses. More cars move every part right by the same factor, and a price only steps down where a part passes one of its own tiers. Tiers sized as in the demo catalog; parts and prices are examples.

That's why volume can rescue a margin that exchange rates hurt. On the demo SUV, at 10,000 vehicles a year, the margin is 11.59%. With the dollar up 10% it drops to 9.95%, and the alert lights up. At 15,000 vehicles a year, enough parts reach their next tier to bring it back to 10.04%.

Two levers on the demo SUV, one after the other. The exchange rate pushes the margin under the 10% target; the volume, through price tiers, brings it back over.

Curves

A sensitivity curve varies one parameter (volume from 1,000 to 1,000,000 a year, an exchange rate, or one part's price) and asks the backend for every point, for the base or for any open scenario. The cost curve shows steps where parts change tiers, and the margin curve shows exactly where it crosses the 10% target.

Scenarios and curve points bypass the cache, so each point is a full computation. At the size of a real car that's a few milliseconds per point, and for a 20,000-part BOM a 30-point curve takes seconds. Curves are for exploring a decision rather than something to run on every page load, so that was acceptable.