- Home
- Capabilities
- Supply Chain Optimisation
Capability
Forecasting and allocation under real constraints.
A forecast that is accurate on average and wrong in exactly the weeks that matter — festival demand, a promotion, a monsoon — has not helped anyone.
01 / The problem
Where this usually goes wrong.
Supply chain forecasting is routinely evaluated with a single aggregate error figure, which is close to useless. The error that matters is concentrated in specific SKUs, specific locations and specific weeks, and it is asymmetric: a stockout on a fast-moving line and excess stock on a slow one are not equivalent mistakes.
The deeper issue is that a forecast is not a decision. Knowing demand will be 40% higher does not tell you what to order given lead times, minimum order quantities, shelf life, warehouse capacity and a supplier who is unreliable in December. The optimisation is the actual product.
02 / What we build
The parts of the system.
- Hierarchical demand forecasting
- Forecasts at SKU, location and time granularity that reconcile to the aggregate, with promotions, holidays, seasonality and weather included as explicit drivers rather than absorbed as noise.
- Prediction intervals, not point estimates
- Inventory decisions depend on the range of plausible demand, not the midpoint. A single number discards precisely the information the planner needs.
- Constrained allocation
- Optimisation that respects lead times, minimum order quantities, shelf life, capacity, budget and supplier reliability — producing an order plan rather than a demand curve.
- Planner-facing tooling
- Recommendations with the reasoning visible, and the ability to override with the override recorded. Systems that cannot be overridden get worked around, and then you have neither the system nor the plan.
03 / How it is measured
What we agree to be judged on.
Set before the build starts, against a measured baseline, and reported honestly afterwards — including where the numbers are disappointing.
- Weighted by value and velocity
- Accuracy on your A-class SKUs matters more than the long tail, and an unweighted average will hide a failure there.
- Service level and stockouts
- Availability against target, and stockout days on the lines that actually drive revenue.
- Inventory turns and working capital
- The other half of the trade. Availability bought with a warehouse full of capital is not an improvement.
- Against the planner
- Experienced planners are a strong baseline and know things absent from the data. Beating them is the bar; assisting them is often the better product.
04 / Honest limits
What this will not do.
Forecasts cannot anticipate genuine discontinuities — a competitor exiting, a new regulation, a supply shock. What a good system can do is widen its intervals when the recent past stops resembling the training period, and say so.
Optimisation is also only as good as its constraints. If real lead times differ from the ones recorded in the ERP, the plan will be confidently wrong, and auditing those constraints is part of the build.
05 / When it applies
You probably need this if:
- Stockouts and excess inventory happening at the same time
- Planning runs on spreadsheets and individual experience
- Forecast accuracy is reported but never acted on
- Seasonal and promotional demand is consistently misjudged
Recognising two or more of these is a reasonable trigger for a diagnostic. Recognising none of them is a reasonable trigger for not spending money here yet.
06 / Related
Often scoped together.
These capabilities share data, infrastructure or evaluation approach with supply chain optimisation, and are frequently part of the same engagement.
Next step
Is supply chain optimisation the right instrument for your problem?
The diagnostic exists to answer exactly that, including the possibility that the answer is no. It is time-boxed, fixed-fee, and the written assessment is yours.