Data, Analytics & Knowledge
Know what you'll sell before you order it
Most Saudi businesses reorder by feel or by last year's number. We build a model trained on your actual sales, seasonality and lead times, so purchasing decisions stop being a guess.
3 to 6
weeks to a working first model
2
sales channels connected minimum
12
months of history needed to start
The direct answer
Demand forecasting is a model that predicts how much of each product you will sell over the coming weeks or months, by SKU and by location. It is for retailers, distributors and manufacturers who currently reorder from a spreadsheet gut check and end up either overstocked or out of stock on their best sellers.
What this removes.
Stockouts on your best sellers
Today
Your top products run out mid-month because reordering is based on last month, not on the trend.
With the system
The model flags the products trending up two to three weeks before you would have noticed the gap.
Cash tied up in dead stock
Today
Slow-moving inventory sits on shelves and in warehouses because nobody flagged the drop-off in time.
With the system
Slow movers get flagged automatically so purchasing can cut orders before the next shipment lands.
Ramadan and season swings catch you off guard
Today
Demand spikes and drops around Ramadan, Hajj and school terms, and every year the buying team relearns it by hand.
With the system
Seasonal patterns are built into the model from your own history, so the spike is planned, not reacted to.
No visibility across branches
Today
Head office sees total sales but not which branch is short on which SKU until a manager calls.
With the system
A branch-level view shows exactly where stock needs to move before a customer complains.
What lands in your hands.
Forecasting model
Trained on your sales, returns and stock history at SKU and branch level.
Data pipeline
Automated pulls from your POS, ERP or spreadsheets, refreshed on a set schedule.
Reorder point logic
Thresholds calculated per SKU from lead time, variability and target service level.
Dashboard
A live view of forecast versus actual, by product, branch and week.
Exception alerts
Flags when a SKU is trending outside its forecast range, sent to your team.
Handover documentation
Model logic and refresh process documented in plain language, in Arabic and English.
Systems and platforms we work with
- OpenAI
- Anthropic
- Google Gemini
- Meta
Systems and platforms we work with
- React
- Next.js
- TypeScript
- Node.js
- Python
- Flutter
- PostgreSQL
- Supabase
- Tailwind CSS
- Docker
- GitHub
- Google Cloud
- Figma
Five stages. You sign off every one.
Read each stage as a small contract: what we need from you, what lands in your hands, and the sentence that has to be true before we move on.
- Data audit3 to 5 days
- Model build2 to 3 weeks
- Dashboard and alerts1 to 2 weeks
- Pilot on live orders2 to 4 weeks
- Live operation and supportongoing
Data audit
3 to 5 days
We look at what sales and inventory data you actually have before promising anything.
- Grant read access to POS or ERP exports
- Share at least 12 months of sales history
- A written data-readiness assessment
- A list of gaps that need fixing first
We move on when we agree the data is clean enough to model, or we agree what needs fixing first.
Model build
2 to 3 weeks
We train a forecasting model on your history and test it against real past periods you can check.
- Confirm which SKUs and branches matter most
- Review backtest results with us
- A working forecast model
- Backtest accuracy report against your own past data
We move on when the backtest accuracy holds up against a period you already know the real answer for.
Dashboard and alerts
1 to 2 weeks
The forecast gets wrapped in a dashboard your purchasing team can actually read and act on.
- Tell us who needs to see what
- Sign off on alert thresholds
- A live dashboard, forecast versus actual
- Configured exception alerts by SKU
We move on when your purchasing lead can read the dashboard and act on it without us in the room.
Pilot on live orders
2 to 4 weeks
The model runs alongside your current reorder process so you can compare before switching over fully.
- Run one purchasing cycle using the forecast
- Flag any calls that look wrong
- Side-by-side comparison of forecast versus manual reorder
- Adjustments to the model based on what you flag
We move on when one full purchasing cycle has run and the forecast outperformed or matched the manual process.
Live operation and support
ongoing
The model refreshes on schedule, alerts keep flowing, and we stay on for tuning as your product mix changes.
- Use the dashboard for reorder decisions
- Flag new products or discontinued lines
- Scheduled model refreshes
- Ongoing support and retuning as your data changes
We move on when we are past handover; this stage runs for as long as you keep the service active.
Asked before signing.
How is this priced?
Pricing is based on the number of SKUs and sales channels we model and how much data cleanup is needed upfront. We quote a fixed price after the data audit in stage one, not before, because a business with clean POS exports costs less to model than one with paper records to digitize first.
Do you need our data to leave our systems?
No. We connect to read-only exports or a read-only database connection wherever possible. Where data has to move for processing, it is handled under PDPL, stored in Saudi-region infrastructure where available, and never used to train models for other clients.
What if we don't have 12 months of clean sales history?
We can start with less, usually six months minimum, but accuracy will be lower until a full seasonal cycle is captured. For new product lines with no history, we use category-level patterns as a starting point and refine as real sales come in.
Does this work with our existing ERP?
We have connected to common Saudi retail and distribution setups including Odoo, SAP and standard POS exports. If your system exports to Excel or CSV on a schedule, that is usually enough to start. We confirm the integration path during the data audit before quoting.
Stop reordering by gut feel
Send us a sample of your sales data and we will tell you within a week whether a forecasting model is worth building for your business.