Automating reference pricing for procurement

The challenge

Reference prices, the internal benchmark every tender is judged against, were produced manually. Data was collected by hand from several departments, formulas were reviewed on a fixed schedule of roughly four times a year, and approvals depended on documents moving between teams.

Downstream, buyers chose between market price, last purchase price and reference price largely at their own discretion. When a reference price drifted away from reality, procurements failed, and there was no structured way to feed that failure back into the formula.

What we did

We designed a phased solution and delivered it mock-up first, so scope was agreed before any production code was written.

• Data integration & harmonisation. An orchestration layer collects, extracts and transforms historical reference prices, bid and won prices, bidder counts, material cost drivers and non-material drivers such as transport and labour, plus public external indicators, into one master dataset.
• Configurable formula engine. Pricing formulas are expressed as versioned value objects, so the client's own developers can add, amend or retire a formula without touching the rest of the system.
• A single price control centre. One screen holds the material, the formula cost tree, historical references, the final price decision, remarks and supporting documents, the anchor point for every pricing decision.
• Dashboards and a feedback loop. Success rates by price type, successful versus unsuccessful procurements against the recommended price, and a channel for buyers to report failures back into formula tuning.

Results

• Time: reference prices generated from formulas and consolidated history instead of manual collection.
• Coverage: a far wider set of inputs behind each estimate, including external market indicators.
• User-centric: one consolidated view for the decision, with the reasoning and evidence attached to it.