In this guide · 8 sections
- 01What 'AI demand forecasting' means in small-shop software
- 02What Shopify, Zoho Inventory and Excel offer today
- 03How much sales data a forecast needs
- 04Seasonal peaks and events in Singapore
- 05Run your own 30-minute test against a spreadsheet baseline
- 06When a custom forecast or dashboard is worth it
- 07Do it yourself or get help?
- 08Frequently asked questions
Short answer
Most small-shop 'AI forecasting' is reorder suggestions based on recent sales, not a model that understands your market. It needs at least 12 months of clean sales data to catch seasonal peaks. Test it against a simple spreadsheet baseline such as Excel's Forecast Sheet before relying on it.
What 'AI demand forecasting' means in small-shop software
Demand forecasting estimates how much of each product you will sell in a coming period, so you can order the right amount at the right time. In enterprise software this can involve models that take in promotions, weather and pricing. In tools aimed at small shops, it usually means one of three simpler things.
| Feature | What it does | Where you see it |
|---|---|---|
| Reorder points | Alert when stock drops below a level you set | Zoho Inventory, Zoho Books, many POS systems |
| Sales-based reorder suggestions | Look at recent sales velocity and current stock to suggest quantities | Shopify Sidekick, inventory add-ons |
| Time-series forecast | Projects past sales forward, including trend and seasonality | Excel Forecast Sheet, dedicated planning apps |
| AI assistant over your data | Answers questions such as 'what is low?' and can trigger tasks | Shopify Sidekick, Zoho's Zia and MCP agents |
None of these knows about a competitor's sale, a supplier delay or a TikTok video that sends one product viral. They project patterns from your own history, so the quality of that history decides the quality of the forecast.
What Shopify, Zoho Inventory and Excel offer today
Shopify (admin + Sidekick)
- Sidekick can suggest what to reorder and rebalance across locations based on sales data
- Can draft purchase orders for you to review
- ABC analysis report ranks products by revenue
- Built into Shopify plans
Best for: Shops selling mainly through Shopify and POS.
Zoho Inventory
- Reorder levels per item
- Zia assistant answers stock questions in plain English
- MCP lets you build replenishment or low-stock agents
- No ready-made forecast listed on its AI page
Best for: Multi-channel sellers comfortable configuring agents.
Excel Forecast Sheet
- Exponential triple smoothing (ETS) forecast
- Seasonality detected automatically or set manually
- Can show a 95% confidence interval around each forecast
- Works on any sales export
Best for: A free, transparent baseline for any shop.
Shopify's Stocky transition guide describes Sidekick helping to forecast demand, identify products to replenish and create purchase orders, after the Stocky app closed on 31 August 2026. Zoho's AI in Zoho Inventory page lists Ask Zia, invoice insights, a blueprint generator and MCP-based agents such as replenishment agents. Microsoft's Forecast Sheet guide explains Excel's method.
Our inventory app comparison covers the plans and prices of these tools.
How much sales data a forecast needs
A forecast can only learn a seasonal pattern it has seen. If December sales are always double, the model needs at least one December to know; with two or more, it can tell a real pattern from a one-off.
Clean data checklist
Excel's guide notes that the timeline needs consistent intervals and can cope with up to 30% missing points by interpolation. New products have no history at all; use a similar product's pattern or a manual estimate until you have a season of data.
Seasonal peaks and events in Singapore
Retail demand in Singapore moves with predictable events: Chinese New Year, Hari Raya, Deepavali, the Great Singapore Sale period, 11.11 and 12.12 marketplace sales, and year-end holidays. Dates for lunar and religious festivals shift each year, which simple monthly models can miss.
- Tag each event in your sales history, with its actual dates for that year.
- For festivals that move between January and February, compare by weeks before the festival rather than by calendar month.
- Plan supplier lead times backwards from the event, including factory closures around Chinese New Year.
- After each event, record what sold out and what was left over; this is better input than any model.
Run your own 30-minute test against a spreadsheet baseline
Before trusting any AI suggestion, compare it with a simple method on a period you already know the answer to. This is 'hindcasting': forecast a past period using only the data before it.
Export sales by week or month
Two years if you have them, for your top 20 products by revenue.
Hide the last 3 months
Use only the earlier data to make each forecast.
Make a baseline
In Excel, select the dates and values and use Data, Forecast Sheet. Also try a plain 'same period last year' figure.
Get the AI suggestion
Ask your tool what it would have ordered for the hidden period, or record its live suggestions for a month.
Compare with what actually sold
Note the error for each product. Keep whichever method is closer more often, and watch for big misses on high-value items.
When a custom forecast or dashboard is worth it
Off-the-shelf suggestions are enough for most small shops. A custom approach makes sense when stock is spread across Shopify, marketplaces and a warehouse system that don't share data, when supplier lead times vary a lot, or when cash tied up in stock is the main constraint.
Our guide to omnichannel POS and inventory covers getting all channels onto one stock count first, which every forecast depends on.
Do it yourself or get help?
Start with clean data and the Excel baseline; it costs nothing and shows whether a paid tool adds value. Most of the work is tidying history, not choosing an algorithm.
Our Custom Business App (from S$4,900) can bring orders, stock and purchase orders from several channels into one system with reorder rules built around your suppliers. The Finance Dashboard (S$790) shows cash, receivables, payables and sales against expenses by month from your accounting software, which helps when deciding how much cash to commit to stock; it does not prepare forecasts for you. IT Sorted is a Singapore-based team of developers with over 10 years of experience; you pay upfront, send the brief in writing and follow progress on your order page.
Frequently asked questions
Shopify's help centre describes using Sidekick, its built-in AI assistant, to forecast demand, identify products to replenish and draft purchase orders based on your sales data. It also offers an ABC analysis report. Treat suggestions as a starting point and check them against your own knowledge of upcoming events.
At least 12 months of sales per product to see seasonal peaks once, and ideally two years or more so the pattern is reliable. The data should be in consistent periods, with stock-outs, promotions and one-off bulk orders flagged so they don't distort the forecast.
Yes. Excel's Forecast Sheet, under the Data tab, uses exponential triple smoothing to project a time series, detects seasonality automatically and can show a 95% confidence interval. It needs consistent time intervals and handles up to 30% missing points by interpolation. It makes a good free baseline.
It can save time on routine reorders, but it is rarely magic. Most tools project your own past sales, so a shop with patchy data or frequent stock-outs will get weak results. Test any tool against a simple spreadsheet forecast on a past period before relying on it.
Monthly models struggle because the festival falls in January or February in different years. Tag each year's festival dates in your sales history and compare sales by weeks before the festival instead of by calendar month. Plan orders backwards from the date, allowing for supplier closures.