The verdict in three sentences
A distributor living with both 8 % stockouts and 90 days of inventory cover does not have a volume problem, it has a SKU-level forecasting problem. AI models trained on 3 years of ERP history, with Ramadan, Eid, Dubai Shopping Festival and back-to-school seasonality, generate purchase order proposals that buyers approve. As a 2026 ballpark in Dubai, budget 90,000 to 200,000 AED (25,000 to 55,000 USD) and 10 to 14 weeks, targeting 3 % stockouts and 15 to 25 % less stock.
Why spreadsheets break at 4,500 SKUs
Rashid, supply chain director of an FMCG distributor based in Jebel Ali, supplies 1,500 supermarkets, groceries and wholesalers across the UAE and Oman. Buyers place supplier orders using a 3-month moving average in Excel. Before Ramadan, dates, juices and baking lines run out, while slow movers sit 5 to 6 months in the warehouse. Imports with 6 to 10 weeks of lead time from Asia and Europe amplify every error.
AI forecasting treats each SKU with its own signals: sales history, Hijri calendar seasonality that moves about 11 days earlier each year, promotions, pricing, supplier lead times, summer slowdown when residents travel, and past stockouts that hide true demand.
| Indicator | Current | 12-month target | Main lever |
|---|---|---|---|
| Stockout rate (order lines) | 8 % | 3 % | Forecast per SKU and per customer |
| Average inventory cover | 90 days | 68 to 75 days | Safety stock calculated per item |
| Monthly forecast error (MAPE) | 45 to 55 % | 25 to 35 % | Seasonal models + Hijri calendar |
| Buyer time per supplier order | 3 to 4 hours | 1 hour | Pre-filled order proposals |
| Dead stock (over 180 days) | 12 % | 6 % | Alerts and targeted clearance |
| Ramadan peak covered without stockout | 70 % of SKUs | 95 % of SKUs | 10 to 12 weeks of anticipation |
Options, budget and timeline in 2026
Figures below are 2026 ballparks for the UAE market. The decisive factor is data quality: 3 years of sales per SKU and per customer in the ERP, identifiable stockouts and stable item codes.
| Option | Upfront cost | Recurring cost | Limit |
|---|---|---|---|
| ERP forecasting module (SAP B1, Odoo, Dynamics) | 15,000 to 45,000 AED setup | Existing licences | Simple models, Hijri seasonality poorly handled |
| Inventory planning SaaS | 10,000 to 30,000 AED | 2,200 to 7,500 AED per month | Generic, limited regional calendars |
| Enterprise planning suite | 550,000 AED and above | High subscription | Oversized below 400 million AED revenue |
| Custom AI, forecasting only | 90,000 to 130,000 AED | 1,100 to 2,200 AED per month | Orders still keyed manually |
| Full custom AI with order proposals | 150,000 to 200,000 AED | 1,800 to 3,700 AED per month | Needs 3 years of clean history |
A typical plan: 2 to 3 weeks of ERP data extraction and cleaning, 4 to 5 weeks of modelling, 2 to 3 weeks of buyer interface and integration, then 2 to 3 weeks running in parallel with the old method. At Kolonell, this sits within our custom web applications and AI practice: the amount is set in the quote based on SKU count and ERP.
Before launching, audit data quality on a sample of 200 SKUs: check that item codes did not change over 3 years, that stockout days can be identified, and that promotions are flagged. Where stockouts are not recorded, the model underestimates true demand by 5 to 15 %. Fixing those gaps during the first 2 to 3 weeks is the single best predictor of forecast accuracy later.
Mini case study
Rashid's company turns over 220 million AED with 165 million AED of cost of goods sold. At 90 days of cover, inventory is worth 165,000,000 / 365 x 90 = 40.7 million AED. A 20 % cut frees 8.1 million AED; at an 18 % annual holding cost, that saves 1.47 million AED a year. On stockouts, lost sales fall from 4 % to 1.5 % of revenue: 5.5 million AED of sales recovered at a 22 % gross margin, or 1.21 million AED. Total annual gain: 2.68 million AED, for a 160,000 AED project and 30,000 AED a year to run. Payback in under 2 months.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
FAQ
How much history does the AI need?
Three years per SKU captures three Ramadans and three back-to-school seasons. With 18 months the model still works, but seasonal accuracy drops by 10 to 15 points.
How does it handle Ramadan moving every year?
The model uses the Hijri calendar as a variable, so the peak shifts about 11 days earlier each year. That is the main gap with tools built on the Gregorian calendar only.
Do buyers lose control?
No. The tool proposes quantities, buyers approve or adjust, and each adjustment feeds the model. Time per order falls from 3 to 4 hours to about 1 hour.
Do we need a new ERP?
No. SAP Business One, Odoo, Microsoft Dynamics or a regional ERP work if they export sales, stock and orders. Integration is 15 to 20 % of the budget.
When do we see results?
First order proposals arrive around week 10. The stock reduction shows after 3 to 6 months, once existing overstock has been sold through.
Let's scope your project. Share your SKU count, ERP and depth of history, and we will define the models and a 10 to 14 week plan. 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.
