Case studies
What the work looked like,
and what it changed
8 engagements described with the trade-offs left in. Every number below is one we measured, not one we estimated.
- Meridian ExchangeRetail & E-commerce2024
Rebuilding Meridian Exchange's bidding engine around per-lot single-writer partitions
Meridian was losing bids to lock contention in the closing seconds of every auction, and disputes over bid order had to be settled manually. We rebuilt the engine around per-lot single-writer partitions and an append-only log, trading cross-lot transactions for provable ordering at 1,800 bids per second.
- 1,800 bids/sec
- Sustained throughput at auction close
- 94 ms
- p99 bid acknowledgement, down from 6.2 s
- 0 disputes
- Contested bid orderings in 14 months live
- Kestrel IndustrialManufacturing2023
Moving Kestrel Industrial off a 19-year-old shop-floor system without stopping a line
Four plants, 300 machines and a VB6 application nobody could test safely. We replaced it one production cell at a time behind a routing gateway, keeping Oracle as the book of record until the final cell was live. Total unplanned stoppage across the entire programme came to 11 minutes.
- 11 minutes
- Total unplanned stoppage across 4 plants
- 24 h to 90 s
- OEE reporting lag after cutover
- 2/year to 2/week
- Production release cadence
- Northwind CommoditiesEnergy & Utilities2025
Replacing Northwind's seven-hour valuation batch with an incremental dependency graph
A seven-hour overnight batch left Northwind's traders working from yesterday's P&L whenever a curve was corrected. We replaced it with an incremental graph over versioned inputs, cutting the full run to 34 minutes and a curve correction to 90 seconds, with 12 months of valuations replayable to the input.
- 34 minutes
- Full end-of-day run, down from 7 hours
- 90 seconds
- Revaluation after a corrected forward curve
- 12 months
- Valuations replayable to their exact inputs
- Corvus LogisticsLogistics & Supply Chain2024
Route planning and ETAs for 1,400 Corvus vehicles that planners actually accept
Corvus planned 220 daily routes by hand and quoted ETAs that were often an hour wrong. We paired a solver seeded with the previous day's plan with a gradient-boosted ETA model, and gave drivers an app that queues work offline. Planning fell from four hours to 22 minutes a day.
- 9 minutes
- p90 ETA error, down from 70 minutes
- 18% to 11%
- Empty running as a share of fleet kilometres
- 22 minutes
- Daily planning time, down from 4 hours
- Ardent ChemicalsHealthcare & Life Sciences2025
Validated electronic batch records for Ardent Chemicals, with requalification as code
Paper batch records took 11 days to review and every software change meant six weeks of manual requalification. We built a validated electronic batch record system with requirements traced to 2,900 automated tests, cutting requalification to three days and batch review to 34 hours.
- 34 hours
- Batch review time, down from 11 days
- 3 days
- Requalification per release, down from 6 weeks
- 22 to 6
- Deviations raised per 100 batches
- Brightmoor FoodsAgritech2026
Forecasting fresh produce for Brightmoor against waste cost, not average error
Four days of shelf life and a median of three units per store-SKU per day, which is too thin for per-store models. We pooled into category models trained on the newsvendor quantile instead of average error, taking waste from 6.1 to 2.9 per cent while availability rose to 98.7 per cent.
- 6.1% to 2.9%
- Units written off as waste
- 98.7%
- On-shelf availability, up from 96.1%
- 40 minutes
- Weekly retraining across 5.6m series
- Orbit SupplyPublic Sector2026
Extracting tender line items for Orbit Supply with citations a buyer can check
Orbit's buyers rekeyed 40,000 pages of tender documents a month, and the errors only showed up at award. We built schema-constrained extraction with page-level citations and deterministic validators, so buyers verify highlighted figures. Intake dropped from 26 minutes to four per tender.
- 4 minutes
- Intake per tender, down from 26 minutes
- 98.2%
- Field accuracy on a 1,200-document gold set
- GBP 0.04
- Model cost per document processed
- Cordage BankBanking & Financial Services2023
Migrating Cordage Bank to ISO 20022 while MT stayed live on the same rails
Moving 500,000 daily payments to ISO 20022 with MT still live on the same rails, without losing data on round trips or settling anything twice. We shadow-ran the new path for 11 weeks across 3.2 million payments, found 41 mismatches, then cut over and took p99 authorisation from 610 ms to 180 ms.
- 41 mismatches
- From 3.2m payments shadow-compared over 11 weeks
- 180 ms
- p99 authorisation latency, down from 610 ms
- 0 duplicates
- Duplicate settlements in 14 months post-cutover