You Are Ordering Stock Based on Gut Feel. Your Competitors Are Using Data.
Most small businesses plan their inventory by looking at what sold last month and adding a rough percentage. Maybe you check last year’s numbers if you have been around that long. Maybe you ask your supplier what they think. Maybe you just guess and hope the order arrives before the demand does.
That approach works when demand is stable. It falls apart the moment something shifts. A trend spikes. A supplier delays. A competitor runs out of stock and their customers come to you. Suddenly you are either drowning in unsold inventory or turning away orders because you ran out of the one thing everyone wants.
AI demand forecasting does not replace your judgement. It gives you a starting point that is grounded in data instead of instinct. You still decide. But you decide with numbers behind you.
What AI Demand Forecasting Actually Does
AI demand forecasting tools analyse your historical sales data, seasonality patterns, and external signals (holidays, weather, social trends) to predict how much of each product you will sell in a given time period.
It is not a crystal ball. It is a statistical model that gets more accurate the more data you feed it. The core process looks like this:
- Historical sales data tells the model your baseline demand for each SKU
- Seasonality patterns tell it when demand naturally rises and falls
- Trend signals (social media mentions, search volume, local events) tell it when something unusual is happening
- The model outputs a predicted demand range for each product, not a single number
The range matters more than the point estimate. If the model says you will sell between 80 and 120 units, you order enough to cover the high end if the cost of stockouts is worse than the cost of overstock. If carrying costs are high, you order closer to the low end. The forecast gives you the range. You make the call.
Where the Data Comes From
You do not need a data warehouse to start. Most forecasting tools pull from sources you already have:
- Your e-commerce platform (Shopify, WooCommerce, BigCommerce) provides order history and SKU-level sales data
- Google Analytics shows traffic trends and conversion rate patterns
- Google Trends (free) shows search interest for product-related keywords
- Your advertising platform shows when you are driving extra demand through promotions
- Weather APIs (available in some tools) adjust for seasonal weather shifts that affect buying behaviour
The best tools ingest this data automatically and update forecasts weekly or daily. You do not need to export spreadsheets or manually normalise anything.
Tools You Can Actually Use
You do not need enterprise software. Here are options that work for businesses processing under 10,000 orders a month:
- Shopify’s built-in demand forecasting (available on Shopify Advanced and Plus) shows predicted demand curves directly in your inventory dashboard. It uses your store’s historical data. Free if you are on those plans.
- Inventory Planner by Sage connects to Shopify, WooCommerce, and other platforms. It builds demand forecasts, suggests reorder quantities, and flags overstock. Plans start around $200/month for smaller catalogues.
- Cogsy focuses on demand forecasting and replenishment for direct-to-consumer brands. It pulls from your store data and ad spend to forecast demand. Plans start around $250/month.
- ChatGPT or Claude (the DIY version) can analyse a CSV of your sales data, identify seasonal patterns, and project next quarter’s demand. It is not as automated as a dedicated tool, but it costs nothing and gives you a starting forecast in 10 minutes.
Pick the option that matches your order volume and budget. A free CSV analysis is better than pure guessing. A dedicated tool is better than a CSV analysis. The point is to start somewhere.
How to Build Your First Forecast (With or Without a Tool)
If you want to test this today without spending anything, here is a 10-minute process using a spreadsheet or ChatGPT:
- Export 12 months of sales data by SKU. From Shopify: Analytics > Reports > Sales by Product. Export as CSV.
- Group SKUs into categories. You do not need a forecast for every single variant. Group similar products (e.g. “candles” rather than “lavender candle 200g” and “vanilla candle 200g” separately).
- Identify seasonal patterns. Look at monthly totals for each category. Which months spike? Which dip? Note the percentage above or below average for each month.
- Calculate your baseline monthly demand. Take the median of the last 6 months (not the average, since outliers skew averages). That is your baseline.
- Apply seasonality multipliers. If December is typically 2x your baseline, multiply the baseline by 2 for December. If February is 0.7x, multiply by 0.7.
- Add a safety buffer. If stockouts cost you more than overstock, add 15 to 20% to the forecast. If carrying costs are high, add 5 to 10%.
This gives you a workable demand forecast for each product category. It will not be perfect, but it will be better than ordering based on what “feels right.”
If you paste your CSV into ChatGPT or Claude and ask for a seasonal demand analysis, it will do steps 3 through 6 for you in about 30 seconds.
What to Do With the Forecast
A forecast is useless if it sits in a spreadsheet. Here is how to act on it:
- Set reorder points. If your forecast says you will sell 100 units next month and your supplier takes 2 weeks to deliver, place the order when inventory hits 50. Never wait until you run out.
- Adjust for promotions. If you are running a 20% off sale next month, the forecast will underpredict. Manually increase the predicted demand by your expected lift (typically 20 to 40% for a decent sale).
- Flag slow movers early. If a product is consistently selling below forecast, it is time to mark it down and move it out. The forecast catches the trend before it becomes dead stock.
- Plan cash flow around peak months. If your forecast shows a 3x spike in November, you know you need to order inventory in October. That means you need cash available in October. Plan accordingly.
Start Simple, Get Smarter Over Time
Your first forecast does not need to be sophisticated. A seasonal baseline with a safety buffer beats gut feel every time. As you collect more data and plug in a dedicated tool, the forecasts get tighter and the surprises get smaller.
The businesses that survive seasonal swings are not the ones that predict the future perfectly. They are the ones that have a range instead of a guess, a reorder point instead of a panic, and a plan instead of a prayer.
Want more like this? Join MOKU Club for free. Weekly resources, early access to new guides, and occasional templates you can actually use.
