Silurian Back to Silurian

Sample service / Forecast & inventory risk diagnostic

Turn historical demand into forward-looking supply decisions.

This interactive sample demonstrates how Silurian turns historical demand into a forward-looking view of service risk, inventory exposure and working-capital pressure. By combining AI time-series forecasting with supply logic and operational judgement, it focuses management attention on the products and decisions that require intervention before performance is affected.

Illustrative demonstration. The data below is synthetic and designed to demonstrate the service workflow. The planned production service will use TimesFM, Google Research's AI foundation model for time-series forecasting. Pre-trained on 100 billion real-world time-points, it can identify patterns across different industries, frequencies and demand profiles. We will test its forecasts against simple statistical baselines using the client's own historical data before making recommendations.
13-week demand
14,780
Forecast volume
Projected minimum stock
1,820
Units after receipts
Safety stock
1,500
Planning threshold
Portfolio action
Monitor
No immediate intervention

Demand and inventory projection

Historical demand is shown alongside a 13-week forward demand forecast and projected closing inventory.

Historical demandForecast baselineIllustrative rangeProjected inventorySafety stock

Portfolio exception view

SKUPattern13-week outlookProjected positionRisk
SLR-317Demand acceleration+18.4%Below safety stock in week 7Red
SLR-204Growth+9.1%Buffer narrowingAmber
SLR-101Stable+3.2%Within policyGreen
SLR-422Decline-11.7%Excess stock buildingExcess

How the service works

Forecast first. Challenge second. Recommend third.

The forecast is a baseline, not a substitute for planning judgement. Silurian tests model performance against actual history and simple benchmarks, then overlays the operational context that a mathematical model cannot see.

  1. #1Profile the data. Check completeness, outliers, stock-outs and structural breaks.
  2. #2Generate the baseline. Run TimesFM and comparison methods over the agreed horizon.
  3. #3Back-test performance. Measure forecast error and bias against held-out actuals.
  4. #4Overlay supply logic. Add inventory, receipts, safety stock and lead-time assumptions.
  5. #5Focus management attention. Rank exceptions and identify where human review is warranted.

A diagnostic can start with a limited historical dataset and a defined group of products, without committing to a wider systems implementation.

Test the approach against your own planning data.

Contact to discuss a forecasting diagnostic