Store sales forecast
At least 60 days of history and 100 orders, plus a passing forecast confidence check.
Example: A store with 100 orders but less than 60 days of history has met only the order requirement.
Results
When will I see my results?
Use recorded figures and computable early estimates from the start. Estimates come with a range, a certainty label, their basis and a next step. The requirements below govern confident results and actions; they are not promised dates.
What you can use first
Early estimates appear as soon as the available data supports their calculation and range. From day one, once the available data has synced: sales, orders, products, recorded carts, current stock and funnel counts. Visits and cart activity need to be captured first. Product gross profit becomes computable where you have entered the relevant costs; missing costs are not zero.
Example: A synced order can contribute to your sales total immediately. Enter that product’s cost to make its gross profit computable.
With limited data, it is too early for a confident conclusion yet. Read each estimate with its range, certainty label and basis; more relevant orders can narrow uncertainty. Reorder plans, recommendations and offer winners still need their own checks. If validation contradicts a prediction, we hide it and show observed figures instead. An inconclusive check does not establish a contradiction.
Past orders imported from Salla count from installation once synced. You do not need to collect that history again. Connect the available order history; new visits supply new browsing and cart data.
Requirements for confidence and actions
Category sales forecast
At least 60 days of history and 60 orders in that category, plus a passing confidence check.
Example: Orders in another category do not fill this category’s 60-order requirement.
A product’s own forecast
First, 40 orders on 25 different sale days. Then the product’s daily history must also pass the 60-day / 60-order check and the forecast confidence check.
Example: 40 orders spread over 25 sale days clear the first stage only. A rarely sold product can have a simpler demand estimate from 5 orders when its history is usable; that is not a full forecast.
Payday-week and Ramadan patterns
At least 365 days of history. Each pattern still needs relevant observations. A Ramadan effect also needs at least 2 observed Ramadans and a supported pattern; a single year does not establish it.
Example: Imported past orders count toward the 365 days. Connect that history before waiting for a new year of sales.
Stock checks against orders
At least 30 matched orders in a 7–28-day window, with 2 days allowed for records to settle. The platform must supply a supported stock ledger, and the stock movements must agree with the orders.
Example: 30 orders without matching stock movements do not confirm stock removed outside orders. Current recorded stock is a separate fact.
Offer results: the Lab add-on
A winner needs at least 14 days, 100 buying shoppers in each group and more than 95% probability of being best. A recommendation needs 30 days, 50 per group and more than 90%. Offers: Coming soon.
Example: Compare checkout value between shoppers who receive an offer and shoppers who receive no offer. Enough buyers alone do not establish a winner. Checks on the comparison and the risk of choosing wrongly must pass too: expected relative loss below 0.5% for a winner and 2% for a recommendation. While a price trial is running, you see activity only: days left, products, orders, units and sales so far per group. The qualified profit effect and its range appear only after the single fixed final analysis, if the checks pass, and remain available to view afterwards.
Customer purchase predictions
At least 100 customers to learn from and 100 to check against. Each set needs 20 who did and 20 who did not have the outcome being predicted, plus a passing accuracy check.
Example: For a purchase prediction, both buyers and non-buyers are needed in each set. Allow roughly 6 months of usable order history as a planning guide, not a guaranteed date; the observation periods and returning customers matter.
Shopper groups and store comparisons
A shopper group needs at least 10 distinct shoppers to be shown. Store comparisons need at least 5 similar stores and show a sector-cohort aggregate, never an individual store’s data. These are privacy floors.
Example: Repeat orders by the same shopper do not create extra distinct shoppers. More orders in your store alone cannot supply the other stores needed for comparison.
What this means for your store
Approximate planning examples for reaching the confidence and winner minimums from a fresh start, not promised dates for results. Synced facts and computable early estimates are available before these minimums. A small store can start with the main report; the Lab needs buying shoppers in every group.
Assumptions: a month of 30 days, steady orders, no imported history, one order per distinct participating buyer, and every order eligible for a single offer split equally with a no-offer group. More groups, repeat buyers or fewer eligible orders mean longer. History already imported can shorten the path to a confident forecast. Quality and winner checks must still pass.
| Orders per month | Example pace | Confident store forecast minimums | Lab winner minimums |
|---|---|---|---|
| <100 | 25 orders/month | About 120 days | About 240 days |
| 100–300 | 200 orders/month | About 60 days | About 30 days |
| 300–1,000 | 600 orders/month | About 60 days | About 14 days |
| 1,000+ | 1,200 orders/month | About 60 days | About 14 days |
At the shared band boundaries, use your actual order pace. Product/category demand and customer counts cannot be inferred from total monthly orders. Payday patterns still need 365 history days; privacy floors still apply.
Not enough evidence yet?
Bring relevant visitors and orders, connect past orders and enter product costs. A result becomes confident when enough relevant data supports it. If the message names missing costs or unavailable inputs, fix that input; waiting alone will not help.
Keep reading
- FAQ
Plain answers about gross profit, offers, messages, stock, setup and your data.
- Sales forecasts
Daily and monthly sales forecasts for up to 90 days, set each week beside what actually sold.
- OffersComing soon
Spend thresholds, credits, tiered savings, gifts and clearance offers, each inside the limits you set and judged by the checkout value it adds.
- Get started
Salla: Available now. Shopify: Coming soon. Zid: Coming soon. WooCommerce: Coming soon. Install Qarari from your platform’s app store, add your product costs and choose what Qarari does for you.