The verdict in three sentences
When you import 1,800 SKUs with 60 to 90 days between purchase order and arrival at your Greater Toronto Area warehouse, every forecasting error costs twice: stockouts on fast movers and containers of slow movers. A demand forecasting model connected to an ordering dashboard costs, as a 2026 market range, CAD 55,000 to 125,000 over 3 to 5 months, provided you have 3 years of sales history. The expected result is forecast error cut from 40% to 20% and 10 to 15% less tied-up inventory.
Why spreadsheet forecasting no longer holds
In a Canadian distributor of imported specialty foods (rice, oils, canned goods, spices, beverages for ethnic and mainstream grocers), replenishment is often run from a spreadsheet: the supply chain director looks at the last three months, adds a safety buffer and orders from suppliers in Asia, Europe or Latin America. This ignores seasonality (back to school, Diwali, Lunar New Year, Ramadan, holiday season), promotions with grocery chains, CAD/USD exchange moves and past stockouts that artificially depressed sales.
A forecasting model combines these signals by SKU and by warehouse, then calculates a suggested order quantity that accounts for lead time, stock in transit, supplier minimums and target service level.
| Metric | Spreadsheet method | With forecasting model |
|---|---|---|
| Average forecast error (MAPE) | 40% | 20% |
| Stockout rate on top 200 SKUs | 12% | 5% |
| Average days of inventory | 110 days | 90 to 95 days |
| Tied-up inventory (on CAD 8M) | CAD 8M | CAD 6.8M to 7.2M |
| Order preparation time | 3 days a month | 4 hours a month |
| Slow movers detected | At stock count | Every week |
What a forecasting project costs in 2026
Budget depends mostly on data condition. Clean history in the ERP sharply reduces preparation, while data spread across several files can take up to 40% of the project.
| Item | 2026 range | Comment |
|---|---|---|
| Audit and clean-up of 3 years of history | CAD 11,000 to 27,000 | Sales, stockouts, promotions, returns |
| Forecasting model by SKU and warehouse | CAD 16,000 to 34,000 | Seasonality, trends, events |
| Order quantity engine | CAD 11,000 to 25,000 | 60 to 90 day lead times, in-transit stock, supplier minimums |
| Ordering dashboard and alerts | CAD 10,000 to 21,000 | Buyer approval before sending |
| ERP integration (NetSuite, Business Central, Acumatica) | CAD 4,000 to 11,000 | Stock reads and PO creation |
| Support and tuning over 2 cycles | CAD 3,000 to 7,000 | Adjustments after first arrivals |
| Monthly hosting and compute | CAD 200 to 800 per month | Weekly recalculation |
A realistic timeline is 3 to 5 months: 4 to 6 weeks of data work, 6 to 8 weeks of modelling and interface, then two ordering cycles run in parallel with the old method.
At Kolonell, this kind of project falls under complex custom applications, from 7,000,000 to 15,000,000+ FCFA (10,670 to 22,870+ EUR). Exact pricing is set in the quote after the data audit.
Data you need before starting
The project succeeds or fails on data. Before signing, check that you can export, for each SKU and warehouse, at least 36 months of daily or weekly sales, the dates and depth of past promotions, stockout periods, supplier lead times actually observed (not the contractual ones) and minimum order quantities. Observed lead times often differ by 15 to 25 days from the contract, especially on ocean freight through Vancouver or Montreal, and that gap alone explains a large share of stockouts. A two-week data audit, before any modelling, tells you whether the 20% error target is realistic for your catalogue.
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
David, supply chain director of a food importer in Mississauga, carries CAD 8 million of average inventory. A 12% reduction frees CAD 960,000 of cash. With a carrying cost of 18% a year (financing, warehouse, shrink and expiry), the annual saving is about CAD 172,800. Meanwhile, cutting stockouts from 12% to 5% on SKUs generating CAD 12 million of annual sales recovers about 7% of lost sales, roughly CAD 840,000 of revenue, or CAD 210,000 of gross margin at 25%. A CAD 95,000 project pays back in under 3 months.
FAQ
Are three years of history really needed?
Yes, to capture at least three full seasons. With only 18 months, forecast error usually stays around 28 to 30% instead of 20%.
Does the AI place orders on its own?
No. It suggests quantities and the buyer approves or adjusts in one click. Those adjustments then improve the model each cycle.
How does the model handle exchange rate or price shocks?
The CAD/USD rate and supplier price increases are included as variables. Above a 10% shock, an alert asks for a manual review of quantities.
Do we need to change ERP?
No. The tool reads stock and sales from your current ERP and sends back approved purchase orders. Integration is 8 to 10% of the budget.
What service level should we target?
Typically 95% on the 200 SKUs that make 80% of sales, and 85 to 90% on the rest, to avoid tying up cash in slow movers.
Let's scope your project. Send us an extract of your sales history and supplier lead times, and we will scope the model and a 3 to 5 month pilot. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
