It depends on your till system. Shopify shops should look at Prediko (from $49 a month under $100,000 annual sales) or Inventory Planner Essentials ($119.99 a month); Square shops get stock tools with Square Plus ($49 a month per location). Stocky closed on 31 August 2026. Many small shops need reorder points before they need forecasting.
That last point saves money. A forecasting tool is only as good as the sales history and stock counts behind it, and for weighed or perishable goods the standard apps fit badly. Alongside the comparison you'll find what the AI part of each tool really does, and a way to test any forecast against your own last season before paying for it.
Stocky has closed: what that changes for Shopify retailers
Shopify's Stocky app, which many small retailers used for purchase orders and stock planning, was delisted from the app store earlier in 2026 and stopped working on 31 August 2026. According to Shopify's guide to moving off Stocky, you keep read-only access to export your data for at least 90 days after that date, and historical purchase orders can't be imported into Shopify. Export your purchase order history now if you haven't; it's the record of supplier lead times and order quantities that any replacement will want.
The export is worth more than an archive. Each purchase order carries the date you raised it and the date the stock was received, and the gap between the two is your real lead time, which is rarely the one the supplier quotes. Take an illustrative ceramics supplier who quotes 14 days. Its last eight orders took 16, 22, 18, 15, 25, 21, 19 and 24 days: 160 days in total, an average of 20, with a worst case of 25. Any tool you set up with the quoted 14 will run you short on almost every order. Before the read-only access ends, turn the export into a one-page supplier sheet. Filled in for a small homeware shop, it might read:
| Supplier | Quoted lead time | Actual average | Worst in last 8 orders | Minimum order | Order day |
|---|---|---|---|---|---|
| Ceramics (mugs, bowls) | 14 days | 20 days | 25 days | 48 units per glaze | Monday |
| Textiles (tea towels, throws) | 7 days | 10 days | 13 days | $300 per order | Wednesday |
| Candles | 5 days | 7 days | 9 days | 24 per scent | Any |
Every tool below asks for these numbers in some form, and Sidekick needs them pasted in because it doesn't store them.
Purchase orders now live in the Shopify admin itself. Shopify points merchants to its built-in inventory tools in the admin and point of sale, plus Sidekick, its AI assistant, which can answer questions such as "What should I reorder?" and suggest how to rebalance stock between locations. That is advice on request rather than a planning system with purchase-order calendars, which is why dedicated apps exist.
The shortlist, compared
List prices in USD from vendor pages or app listings, checked 27 September 2026.
| Tool | Works with | Price | What the AI part does | Best for |
|---|---|---|---|---|
| Shopify admin with Sidekick | Shopify | Included with Shopify plans | Answers reorder questions and suggests rebalancing from your sales data | Small ranges where you want a second opinion, not a plan |
| Prediko | Shopify | $49 a month under $100,000 annual sales; higher bands cost more; raw-material and bill-of-materials add-on $20 | Revenue and inventory forecasts, a 12-month purchase-order calendar, restocking recommendations | Growing Shopify brands with suppliers on long lead times |
| Inventory Planner Essentials (Sage) | Shopify, single warehouse | $119.99 a month | Monthly forecasts by variant, reorder suggestions, low-stock alerts, purchase orders | One-stockroom Shopify shops wanting an established planner |
| Square Plus | Square point of sale | $49 a month per location | Stock counts, vendor and purchase-order management, sell-through and ageing reports; Square AI chat (in beta) for questions about your data | Delis, butchers and gift shops already on Square |
| General AI assistant on your sales export | Any system that exports CSV | $0 to about $20 a month | Analysis you ask for; you still place the orders | Weighed and perishable goods, one-off reviews |
If you use a different point of sale, check its inventory module for reorder points and suggested orders before adding anything. An extra app means another sync to go wrong, as anyone who has tried to keep stock in step across several sales channels will know.
What each tool is like to live with
Sidekick: good questions, no memory of your suppliers
Sidekick is free with your plan and reads your sales and stock. Ask it which products are selling faster than usual, what has sat unsold for 90 days, or where to move stock between two locations. It won't hold supplier lead times, minimum order quantities or a purchase calendar for you.
That gap shows up in the first answer. An illustrative exchange in a homeware shop:
Prompt: Which products should I reorder this week?
Reply: Based on the last 30 days, these are selling faster than
their stock will last:
- Speckled stoneware mug (sage): 41 in stock, about 3 a day,
roughly 14 days of cover
- Linen tea towels, set of 2: 18 in stock, about 1 a day
- Fig soy candle: 9 in stock, about 0.8 a day
Consider reordering these soon."Soon" is the problem. The mug's supplier takes 20 days, so 14 days of cover means the shop is already six days late, and the reorder point worked out further down (75 mugs) was passed about eleven days ago. The fix is to give Sidekick what it doesn't hold: "Our lead times are: mugs 20 days, tea towels 10 days, candles 7 days. Which of these will run out before a new order could arrive?" Keep that sentence in a note and paste it in each time. For ideas on making the most of it, see using Shopify Sidekick to run your store faster.
Rebalancing between locations is where it's most useful to a two-site shop, and where a quick sanity check pays. Asked "Should I move any stock between the main shop and the second branch?", an illustrative reply might suggest moving 12 sage mugs from the main shop (36 in stock, 2 a day) to the branch (4 in stock, 1 a day) and 6 wool throws the other way. Before anyone packs a box, check two things. First, the counts: if the branch's four mugs are really zero because two broke and two sit on a display nobody sells from, the suggestion is too small. Second, the trip: moving 12 mugs is worth it only if someone is driving between the sites anyway, so batch transfers into one weekly run rather than acting on each suggestion the day it appears.
Prediko: forecasting priced for small brands
Pricing follows your last 12 months' sales, which makes the entry band cheap for a small shop. Every plan includes unlimited users, products and purchase orders. The raw-materials add-on suits brands that make some of what they sell, such as a homeware brand that finishes its own candles or cushions. There's a 14-day free trial, long enough to run the backtest described below.
The purchase-order calendar is the part that earns the fee for brands on long lead times. Take an illustrative candle brand with under $100,000 a year in sales, so on the $49 band. Its forecast for its best-selling scent across November and December is 1,100 units. The supplier needs 10 weeks to make them and freight takes another two, so to have them on the shelf by 1 November the order has to go in by about 9 August. Working back like that for 40 products by hand is where most small brands miss Christmas; a calendar that shows "order by 9 Aug: 1,100 units" in July is the thing to check the trial actually produces for your range.
Inventory Planner Essentials: the simpler version of a mature planner
Essentials is the cut-down tier of Sage's planning software for single-warehouse Shopify merchants. It syncs sales and stock from Shopify, forecasts monthly by variant and suggests purchase quantities. Its higher tiers handle more complex set-ups, and their prices aren't published on the app listing. At $119.99 it costs more than Prediko's entry band, so compare both on your own data.
Forecasting by variant is where it pays for itself, because shops tend to reorder products evenly across colours and sizes. An illustrative case: a gift shop orders 400 of a mug, 100 in each of four glazes. After eight weeks the sage glaze has sold out and 70 mustard mugs are still on the shelf, because sales split roughly 45% sage, 30% white, 15% blue and 10% mustard. A variant-level forecast would suggest the next 400 as about 180, 120, 60 and 40. Same spend, far fewer lost sales, and less stock sitting in the back room.
Square Plus: stock control first, AI second
Square's free plan tracks basic stock. Plus adds the tools a shop actually reorders with: counting on a phone, vendor profiles, purchase orders, bulk intake by scanning, and reports on what's ageing on the shelf. Square AI, still in beta when I checked, answers questions about your sales. For many food shops this is enough, and our tutorial on faster stocktakes with a phone scanner pairs well with it.
The ageing report is the one to open first. In an illustrative gift shop on Square, it might show 22 lines with no sale in 120 days, holding about $1,900 of stock at cost, most of it bought for last year's Christmas. Nothing clever is needed to act on it: stop those lines from appearing on the next purchase order, move them to a table near the till, and bundle the slow candles with a matching holder. Forecasting helps you buy the right things next time; the ageing report frees up the cash tied up in the wrong things this time.
Reorder points: the step most shops should take first
A reorder point is the stock level at which you place the next order, set so the new delivery arrives before you run out. Every tool above either uses one or calculates one, and you can set them by hand today:
Reorder point = average units sold per day x supplier lead time in days
+ safety stock
Example (illustrative): a stoneware mug sells 3 a day, the supplier
takes 20 days to deliver, and you keep 5 days' cover as safety stock.
Reorder point = 3 x 20 + (3 x 5) = 75 mugs.The five days of safety stock isn't a guess. It's the gap between the supplier's worst delivery and its average one from the supplier sheet: 25 days minus 20, times 3 mugs a day, gives 15 mugs. Suppliers who deliver like clockwork need little safety stock; erratic ones need more, and the sheet tells you which is which.
The mistake to avoid is setting the point from the quoted lead time. With 14 days in the sum, the reorder point comes out at 3 x 14 + 15 = 57 mugs. Fifty-seven mugs last 19 days, the average delivery takes 20, so the shop runs out for a day on a typical order and for six days when the supplier has a slow month. It shows up as the same best seller missing from the shelf every few weeks, which owners usually blame on demand rather than on the number in the sum.
Set reorder points for your top 30 products by revenue and review them each quarter. If that alone ends most stockouts, you may not need a forecasting app yet. If you still get caught out by seasonal swings, launches or long lead times, that's the point where forecasting earns its monthly fee.
Weighed and perishable stock: where forecasting apps struggle
Forecasting apps are built for products sold in units with shelf lives measured in months. A butcher or a delicatessen breaks that model in three ways:
- Weight, not units. You sell 340g of a cheese, not "one". Many tools can store weight-based items, but forecasts in units of "1 kg" rarely match how you buy.
- Days, not months. A soft cheese or fresh sausage has days of life. A monthly forecast is useless; you need a par level by day of the week.
- What you buy isn't what you sell. A butcher buys a side or a whole animal and sells a dozen cuts from it. The ordering question is "how many sides this week?", which depends on the mix of cuts you expect to sell and the yield of each.
The butcher's version of a reorder calculation is a yield sheet. Illustrative figures, from one shop's own cutting records for a 40 kg half pig: loin 7 kg, belly 5 kg, leg 11 kg, shoulder 9 kg, sausage meat and trim 5 kg, bone and waste 3 kg. If next week's expected sales are loin 12 kg, belly 6 kg, leg 10 kg, shoulder 8 kg and sausages 9 kg, divide each by its yield: loin needs 1.7 halves, sausages 1.8, belly 1.2, and leg and shoulder under one each. So the order is two halves, set by loin and sausages, and the planning job is what to do with the spare 12 kg of leg and 10 kg of shoulder: a roasting-joint offer, or more of it minced into sausages. No inventory app asks that question, but an assistant given the yield sheet and the sales forecast can work it through each week.
For these shops, a general AI assistant working on your own sales export often does better than an app. Export 12 weeks of sales by product and day, remove customer details, and use a prompt like this:
You are helping a small delicatessen set daily par levels.
Attached: 12 weeks of sales by product, weight sold (kg) and date.
1. For each of the top 25 products by weight, give average kg sold
for each day of the week, and the highest day in the period.
2. Suggest a par level per day = average + a buffer, where buffer is
half the gap between average and highest. Show the working.
3. Flag products whose sales changed by more than 25% between the
first 6 weeks and the last 6 weeks.
4. List anything that looks like a data error (negative sales,
impossible weights) instead of using it.
Output a table I can paste into a spreadsheet. Do not guess missing data.An illustrative extract of what comes back for a delicatessen open Tuesday to Saturday:
| Product | Tue avg (kg) | Fri avg (kg) | Sat avg (kg) | Sat highest | Sat par |
|---|---|---|---|---|---|
| Mature cheddar | 2.1 | 4.4 | 6.2 | 8.0 | 7.1 |
| Prosciutto, sliced | 0.6 | 1.3 | 1.9 | 2.7 | 2.3 |
| Soft goat's cheese | 0.4 | 0.9 | 1.1 | 1.3 | 1.2 |
Flagged: 14 rows dated Monday (shop closed). Possible date error;
not used in the averages.
Flagged: prosciutto sales up 38% in the last 6 weeks.The working is visible (cheddar: 6.2 average plus half the 1.8 gap to 8.0 gives 7.1), so you can check it. The Monday rows turned out to be Saturday takings the till posted after midnight, which means Saturday was understated: move them back and rerun before trusting any Saturday par. The prosciutto jump was a new sandwich on the café menu, so that par should rise and stay up, not be averaged away.
Check the output against what you know. If it says you sell more pork belly on Mondays and you're closed on Mondays, the export has a date problem. A tidy export matters more than the choice of assistant; the steps in cleaning messy data apply directly to till exports.
How much sales history a forecast needs
A tool can only learn seasonality it has seen. With less than a year of clean sales history, a forecast can't know about the December rush or the summer barbecue weeks, so it will under-order for them. In your first year, set seasonal adjustments by hand and treat the tool's suggestions as a baseline.
New products have no history at all. Good tools let you link a new product to a similar existing one; check how each candidate handles that, because a homeware brand launching 40 new lines a season lives or dies on it. And stockouts poison history: if a product sold nothing for three weeks because it was out of stock, the tool may read that as low demand. Ask how it treats out-of-stock periods.
The size of that error is easy to work out. An illustrative candle sold 8, 9, 7, 8, 0, 0, 0, 9, 8, 7, 8 and 8 over twelve weeks, and the three zeros were weeks when it was out of stock. Averaged over all twelve weeks, that's 72 divided by 12, or 6 a week. Averaged over the nine weeks it was on the shelf, it's 72 divided by 9, or 8 a week. A tool that takes the zeros at face value orders a quarter too few, which sets up the next stockout, which drags the average down again. During a trial, find one product you know ran out last year and look at what the tool forecasts for it. If the forecast dips for the weeks after the gap, the tool is learning from your stockouts.
For new lines, the setting to look for links the new product to an existing one, and it matters which stretch of the old product's history you borrow. A new glaze of the stoneware mug should borrow the sage mug's first eight weeks after launch, when it sold about 5 a day on novelty, not its current 3 a day. Borrow the settled rate and the new glaze sells out in its launch month.
Backtest a forecast before you pay for it
You can test any forecasting tool against last year during its free trial:
- Pick a period you know well, such as last October to December.
- Connect the tool, but ask it (or set its dates) to forecast that period using only the sales before it. Some tools make this easy; if one can't, ask its support team how customers test accuracy.
- Compare the forecast with what actually sold for your top 20 products by revenue.
- Work out the error for each: the difference between forecast and actual, divided by actual. Average them.
- Compare with your own ordering. If what you ordered last year was closer to actual sales than the tool's forecast, the tool isn't adding much yet.
Here is the sum filled in for five of an illustrative homeware shop's top sellers, backtested over last October to December:
| Product | Tool's forecast | Actually sold | Error | What the owner ordered |
|---|---|---|---|---|
| Stoneware mug, sage | 260 | 300 | 13% | 240 |
| Fig soy candle | 410 | 350 | 17% | 500 |
| Linen tea towels, set of 2 | 180 | 150 | 20% | 150 |
| Advent calendar (new that year) | 90 | 160 | 44% | 120 |
| Wool throw | 120 | 115 | 4% | 150 |
The tool's average error is about 20%, right on the edge. Take out the advent calendar, which had no history to learn from, and it falls to about 14%, better than the owner's own orders on the mug, candle and throw. The lesson for this shop was that the tool is worth paying for on repeat lines, and new products need the "link to a similar product" setting or a manual figure.
A forecast within about a fifth of actual on your best sellers is useful for most small retailers. Much worse than that and you'll be overriding it constantly. For the mechanics of how these forecasts are built, read how AI inventory forecasting works for small businesses.
Picking by shop type
| Shop | Start with | Add later, if needed |
|---|---|---|
| Homeware brand selling online and at pop-ups, 400 products, suppliers on 8 to 12 week lead times | Prediko or Inventory Planner Essentials, after a backtest | Raw-materials add-on if you finish products yourself |
| Delicatessen on Square, 300 lines, many sold by weight | Square Plus stock tools plus daily par levels from an AI-assisted sales review | A forecasting app only if packaged lines dominate |
| Butcher buying whole or half carcasses | Yield spreadsheet and par levels by day; pre-order lists for holidays | Nothing, until the spreadsheet stops coping |
| Small Shopify gift shop, 150 products, local suppliers | Shopify's built-in tools and Sidekick | Prediko's entry band when reorders take more than an hour a week |
For the gift shop in the last row, the switch point is a quick sum. Say the owner spends three hours every Monday working out orders by hand, and a trial shows a planning app cutting that to one. That's about eight hours saved a month for $49, before counting any stockouts avoided. If the same owner only spends 40 minutes a week on reorders, the app saves perhaps two hours a month and the built-in tools with reorder points are the better deal. Time yourself on two ordinary ordering sessions before the trial starts, so the comparison uses real minutes rather than a feeling that ordering takes ages.
If you'd like help deciding which of these fits and setting it up with your suppliers' lead times, that's the kind of job my AI implementation consultation covers. For the wider first-year budget, see what AI costs a small retail shop in its first year.
Further reads
- Can AI Help a Small Shop Set Prices and Promotions? — Use the same sales data to set prices and promotions.
- How to Forecast Covers and Cut Food Waste With AI — Waste-cutting forecasts, useful for delis with a café.
- Taking a Local Shop Online With AI: Product Listings in a Weekend — Getting a physical shop's range online, fast.
- AI Vendor Lock-In: How to Keep Your Data and Prompts Portable — Keep purchase history portable when apps close.
- How to Evaluate an AI Software Vendor: A Small Business Scorecard — Judge any forecasting vendor's accuracy claims.
- How Small Shops Can Use AI to Forecast Stock and Reorder — Simpler AI-assisted reorder methods for very small shops.
- How Pet Shops Use AI for Stock, Subscriptions and Advice — Reorder rules for dated and seasonal lines, bag run-out maths for subscriptions, and an advice assistant that knows when to say 'ask your vet'.
- How Clothing Boutiques Use AI to Spot Trends Before Buying Stock — Combine free trend tools with AI analysis of your own sell-through, then buy trend pieces at test depth. Includes prompts, sample outputs and a buy plan.
- How to Use AI to Set Reorder Points and Prevent Stockouts — Reorder point maths made practical: clean the history, measure real supplier lead times, pick service levels per item and let AI do the sums for every line.
- Why AI Stock Forecasts Go Wrong: 8 Mistakes Small Businesses Make — The eight ways small businesses feed AI stock forecasts bad history or misread the output, each with a worked example and a fix you can apply this week.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Shopify Help Centre page on migrating from Stocky; Prediko pricing page; Inventory Planner Essentials listing on the Shopify App Store; Square point-of-sale pricing page. Checked 27 September 2026.