Data-driven pricing for service contracts, across 1,500+ devices
Sarlin priced maintenance contracts by hand in Excel, leaning on a handful of experts' intuition. We built an AI-powered pricing engine that enriches fragmented equipment data and predicts costs across more than 1,500 industrial devices - so account managers quote objectively, consistently and at scale.
The challenge
Sarlin needed a consistent way to price service contracts across a large and varied equipment base.
Manual pricing
Each quote required time-consuming Excel work.
Expert-dependent decisions
Pricing relied on senior specialists' knowledge, making it difficult to standardize and teach.
Fragmented equipment data
Inconsistent component names and missing specifications made pricing automation difficult.
What we built
We built a working pricing engine in two layers: an AI-powered tool that cleans and enriches the equipment data, and machine-learning models that turn that data into a defensible price - all behind a secure web app account managers actually use.
- 01Foundation
AI-powered data enrichment
A web tool that pairs rule-based extraction with a large language model to read inconsistent device names, standardize them into clean categories and manufacturers, and fill in missing technical specifications such as power, pressure and voltage - turning fragmented records into a reliable pricing dataset.
- 02Modeling
Four specialized cost models
Rather than predicting one lump sum, we built separate gradient-boosted models for material, labor and travel costs, plus an integrated total. Splitting the problem this way captured the real cost drivers and kept every prediction explainable.
- 03Transparency
Benchmarking built in
The engine lets pricers cross-check each estimate against the real cost history of identical equipment, so a quote is never a black box - the model's number sits right next to the evidence behind it.
- 04Deployment
A secure web app on Azure
We deployed an internal web interface on Azure Container Apps. An account manager enters a device model, pulls its enriched specs, selects a validated peer group, sets the margin, and gets a price in seconds.
Results
Sarlin moved from manual, intuition-based spreadsheets to objective, repeatable pricing - delivered as a working tool, not a slide deck. The same logic now applies consistently across more than 1,500 devices.
More than 1,500 industrial devices, once scattered across inconsistent records, were standardized into a single governed register that the pricing engine draws on for every quote.
Around 312,000 historical ERP transaction records were cleaned and analyzed to train the models - turning years of dormant data into a pricing asset.
Separate models for material, labor and travel costs feed an integrated total, so each quote breaks down into drivers a manager can see and defend.
Pricing no longer depends on who is in the room. The engine replaces manual Excel guesswork with repeatable logic - and shows the cost history behind every number, so teams can trust and explain it.
Beyond the numbers
- Standardized 27 device categories and 51 manufacturers from inconsistent records, with 20+ technical specifications enriched automatically
- A clear path to scale: from the pricing tool today toward deeper integration with Sarlin's ERP environment
