Horizon 3 · Advanced Analytics

Advanced analytics

Modelling what has not happened yet, and choosing the best move before you commit to it. Forecasts, elasticity, simulation, and optimisation, run on your numbers.

Past tense to future tense

The first two horizons look backward: what happened, and why. Advanced analytics turns to face the other way. What will sales do next quarter? What happens to volume if we lift price four percent? Which route holds the promise at the lowest cost?

None of these have a clean answer in the history alone. They need a model: of demand, of how customers respond, of the constraints you operate under. Three live demos below, on sample data, so you can see the kind of question this horizon answers.

Example one

Forecasting, seasonality and all.

Two years of history show the rhythm: the festive peaks, the summer lull, the climb underneath. A real forecast carries that seasonality forward with a band around it, and shifts with the assumptions. Switch the scenario and watch the projection and its range move.

Monthly sales forecastUnits in thousands · 24 months actual, 12 projected · sample data
Year on year
+12.0%
▲ Year 2 vs Year 1
Festive peak
204k units
October (Diwali), Year 2
Model accuracy
94%
back-tested on held-out months
Actual Forecast Likely range
The band is the likely range, widening the further out we look. Accuracy is the model back-tested against months it had not seen. Sample data; the method is real.
Example two

Price elasticity, as a live what-if.

Lift the price and volume falls, but by how much, and what does that do to revenue and to profit? They do not peak at the same price. Drag the price and watch all of it move at once.

Price simulatorOne product, modelled demand · sample data
Price per unit
₹100
₹60₹170
Volume
Revenue
Profit
Unit margin
Modelled demand, cost ₹55/unit. Revenue peaks near ₹88, profit near ₹116, the gap is the point. Sample data.
Example three

Optimisation: hold the promise, cut the cost.

Pickups and deliveries planned by hand wander all over the map and still miss windows. We model the stops, the drive times, and the SLA, and solve for the route that keeps every promise on the least distance. Toggle between the hand-drawn plan and the optimised one.

Last-mile route · one shiftEight stops, 180-minute SLA · sample data
Distance
Drive time
On time vs SLA
Map is live (OpenStreetMap, CARTO). Stops, the depot and the SLA are illustrative sample data; the optimisation under a time constraint is the real method.
The top of the climb

If you can model it, you can decide it before you spend on it.

Talk to us →