Manufacturing
In annual savings from reduced changeover losses
Reduction in changeover losses in first quarter
About
A FMCG manufacturer scheduling production lines manually in Excel. We replaced it with a natural-language AI planner that generates optimised schedules from live constraints — and that planners can interrogate and adjust without writing a query.
Industry
Manufacturing
Company size
500 – 1,000 employees
Founded
2003
The Company
A tight working capital model with a forecasting problem at its centre
Apex Consumer Brands is a mid-sized FMCG enterprise with manufacturing operations in two states and a distribution network spanning 15 regional markets. The company operates on tight working capital, with a business model that requires significant upfront raw material procurement against payment terms from distributors and modern trade partners that frequently extend to 60 or 90 days.
This structural tension between payables and receivables made accurate cash flow forecasting not a nice-to-have, but a core operational requirement. The business needed to know — with confidence and advance notice — when inflows would arrive, where shortfalls were developing, and what actions the treasury team needed to take before problems became crises.
The challenge
A manual process producing forecasts that consistently missed
The finance team was managing cash flow using a weekly manual process — pulling accounts receivable data from the ERP, reconciling it against expected payment schedules, and producing a 4-week cash flow forecast in Excel. The process was labour-intensive, error-prone, and consistently produced forecasts that were too optimistic about the timing of inflows.
Payments that the forecast showed arriving in week three routinely slipped to week five or six — either because distributors were themselves cash-constrained, or because invoice disputes or delivery discrepancies created delays the forecasting model had no way to anticipate. These surprises triggered reactive short-term borrowing at unfavourable rates and strained supplier relationships when payables had to be extended without notice.
Two analysts were spending three days per week producing a forecast that was obsolete almost as soon as it was complete. The CFO's core ask was direct: a cash flow model that could look forward 90 days with enough accuracy to allow proactive treasury management.
The Solution
A machine learning-driven rolling cash flow forecasting system
Seven Billion built a machine learning-driven rolling cash flow forecasting system that replaced the manual Excel process entirely. The model drew on four primary data sources: historical payment behaviour by customer, analysing actual payment timing of each distributor against invoice due dates over a 24-month period; current accounts receivable aging covering all outstanding invoices stratified by customer, age, and dispute status; a proprietary credit risk scoring model built on payment history and external signals; and sales pipeline data covering upcoming orders and expected shipment dates.
The model was built using TensorFlow on Google Cloud Platform, with BigQuery as the data warehouse and Vertex AI managing model training and inference. A rolling 90-day forecast was regenerated daily, with automated alerts triggering when projected cash balances fell below defined thresholds at any point in the forecast horizon.
Alongside the forecasting model, a collections prioritisation tool was deployed that ranked outstanding receivables by expected delay risk — allowing the collections team to direct outreach toward the accounts most likely to slip, rather than following a uniform aging-based process.
The Results
Rs. 3.2 Cr recovered and a treasury function transformed
Rs. 3.2 Cr in additional working capital was identified within the first three months — a combination of accelerated collections from high-risk accounts flagged by the prioritisation tool and reduced short-term borrowing as the more accurate forecast eliminated reactive liquidity management.
Days Sales Outstanding reduced by 19% — a structural improvement driven by the shift from uniform aging-based collections to risk-weighted prioritised outreach. Rolling 30-day cash flow forecast accuracy reached 94% within two months of deployment, compared to an estimated accuracy of under 60% for the previous manual process.
The finance team's weekly cash flow preparation process, previously requiring two analysts for three days, was replaced by a fully automated daily update. Two supplier payment disputes were avoided in the first quarter because the early warning system flagged projected shortfalls in time to initiate conversations proactively.
Two supplier disputes were avoided in the first quarter because we saw the shortfall coming three weeks early. That is the kind of impact that is easy to quantify — and impossible to achieve without the right system.
CFO, Apex Consumer Brands
KEEP READING
Explore more studies on AI, analytics, and enterprise intelligence.
