How it works
Inventory forecasting, explained without the hand-waving
Forecasting turns your past orders into a sales rate, then projects that rate against the stock you hold now. The rate gives you days of cover. Compare that with your supplier's lead time and you know whether today is the day to order.
The order quantity is the demand you expect while you wait, plus a safety cushion, minus what you already have. Written out: reorder point = (daily sales × lead time) + safety stock Everything below is that one idea, taken slowly, with the numbers filled in.
What forecasting really means here
"Demand forecasting" sounds like it should involve a model predicting the future. For a Shopify store with a few hundred to a few thousand SKUs, that is not what earns its keep. What earns its keep is answering one narrow question well: will this product run out before a new order can arrive?
That question has four inputs, all of which you either already have or can find out in an afternoon. Nothing here is beyond arithmetic, and being able to follow it is the point. If you can follow it, you can spot when a recommendation is wrong, which is the only way a planning tool ever earns trust.
Cedar Wick Candle · 8oz
Projected stock
1. Sales velocity: how fast it actually moves
Sales velocity is average units sold per day. The naive version is total units divided by days. The useful version weights recent weeks more heavily than old ones, because what a product did in March matters less than what it did last week.
sales velocity = units sold ÷ days in the window, weighted toward recent weeks
Say a candle sold 540 units over the last 90 days. That is 6 a day. If the last 30 days were noticeably busier than the 60 before them, a weighted rate lands above 6 and the plan tracks the trend instead of lagging behind it.
Two things ruin this number if you let them. Refunds should come back out of the count, or you will plan for sales that were reversed. And a single freak week, the viral post or the one wholesale order, should not be allowed to redefine normal.
2. Days of cover: how long today's stock lasts
days of cover = stock on hand ÷ sales velocity
With 320 candles on the shelf and 6 selling a day, you have about 53 days of cover. This is the single most useful number in inventory planning, because it is directly comparable to the one number your supplier gives you: how long they take.
3. The reorder point: the line that matters
The reorder point is the stock level at which you have to place an order to avoid running out. It is not a preference, it is arithmetic.
reorder point = (sales velocity × lead time) + safety stock
The candle sells 6 a day and the supplier takes 30 days. You will sell 180 while you wait, so 180 is the floor. Add a 9-day cushion, which is another 54 units, and the reorder point is 234.
This is why the alert everyone sets on a round number is a trap. "Tell me when it drops below 50" is fine for something that sells one a day and arrives in a week. For a product that sells 6 a day and takes a month, 50 units means you ran out three weeks ago and did not know it.
Note that lead time is the input Shopify has never stored. Your catalogue, your stock, your costs, your vendors, and your order history all live in Shopify. How long your supplier takes lives in your head or in a spreadsheet. It is the one thing any planning tool has to ask you for.
4. Safety stock: the cushion, sized honestly
Safety stock covers the difference between the average case and a bad week. Two things go wrong at once in practice: demand runs hotter than usual, and the supplier runs later than promised. The cushion absorbs both.
The wrong way to size it is a flat percentage across every product. A steady seller that ships 6 a day, every day, needs almost no cushion. A product that sells 0 some weeks and 40 in others needs a lot. Sizing them the same means you are simultaneously over-buying the predictable one and still running out of the spiky one.
The right way is to size the cushion against how much that particular product's sales actually vary, and against how confident you want to be. That confidence level is called a service level. Aiming to cover 95% of weeks costs more stock than aiming for 85%, and that trade is yours to make, product by product if you want.
5. How much to order
Once you have crossed the reorder point, the quantity is the shortfall between where you are and where you need to be, adjusted for how your supplier actually sells to you.
order quantity = reorder point − stock on hand − stock already on order
With a reorder point of 234 and 45 on hand, the shortfall is 189. Then reality applies: your supplier ships in cases of 12, so 189 rounds up to 192. If they have a minimum order of 240 units, you order 240. If you already have 100 arriving on a purchase order, those come off the top first, and this is where a lot of spreadsheets quietly double-order.
What forecasting can't do
Being straight about this matters more than the sales pitch, because every one of these has burned someone.
- It cannot predict a launch. A brand-new product has no history. Nothing can forecast it. Set a manual threshold, watch it, and let the forecast take over once there is a pattern.
- It cannot know about your marketing. If you are running a promotion next month, the forecast has no idea. That is what the override is for.
- It cannot fix a wrong lead time. Every number downstream depends on it. If your supplier says 30 days and reliably takes 45, the plan is wrong by 90 units before it starts.
- It cannot see a supply shock coming. A factory closure, a shipping delay, a raw material shortage. Forecasts read your demand, not the world.
A good planning tool is honest about all four. It should tell you when it does not have enough data instead of dressing up a guess, and it should make overriding easy rather than treating your judgement as an error.
How Replinish runs it
Replinish reads up to 365 days of your Shopify order history on the day you install, then keeps up in the background as new orders come in. It computes velocity, days of cover, reorder point, safety stock, and suggested quantity for every variant, and recomputes overnight.
Every product page shows those five numbers with their inputs next to them, in a sentence rather than a formula: "We suggest 240. You sell about 6 a day, your supplier takes 30 days, we keep a 9-day cushion, and you've got 45 on hand." Change any input and watch the answer move.
Where there isn't enough history to forecast a product honestly, it says so and offers a manual threshold instead. That is a deliberate choice: a number you cannot trust is worse than no number, because you will act on it once and then never trust the tool again.
Next: how a forecast becomes a purchase order, including lead times, minimums, and case sizes.
Questions about forecasting
How does inventory forecasting work for a Shopify store?
Inventory forecasting turns your past orders into a sales rate per product, then projects that rate forward against your current stock. The rate tells you how many days of cover you have left. Comparing that to your supplier's lead time tells you whether you need to order now. The order quantity is the demand expected over the lead time, plus a safety cushion, minus the stock you already hold.
What is the reorder point formula?
Reorder point equals average daily sales multiplied by lead time in days, plus safety stock. For a product selling 6 a day with a 30-day supplier lead time and a 9-day cushion, that is (6 × 30) + (6 × 9) = 234 units. When stock falls to 234, it is time to order.
How much sales history do you need to forecast well?
Enough to see the product's pattern rather than its noise. A few weeks of steady sales gives a usable rate. Products with seasonal swings need a year to be read properly, which is why Replinish reads up to 365 days of Shopify order history on the first day. Below the threshold where a forecast would be honest, Replinish says so rather than producing a confident-looking number.
Is AI needed for inventory forecasting?
Not for most Shopify stores. The arithmetic that decides the great majority of reorder decisions is a weighted average sales rate, a lead time, and a cushion sized to volatility. That is transparent, checkable, and hard to get badly wrong. Complexity in this category usually adds opacity faster than it adds accuracy.
How does forecasting handle seasonal products?
By weighting recent sales more heavily than old ones, so the rate tracks a season as it turns rather than lagging behind a yearly average. A full year of history means last winter's shape is visible when this winter starts. For genuinely spiky products the safety cushion widens automatically, because the cushion is sized from how much that product's sales actually vary.
What if a product had a one-off spike, like a viral post?
A single abnormal week can drag an average upward and make you over-order. Replinish weights recent weeks more heavily but does not treat one outlier as the new normal, and it flags a suggested quantity that is far outside your usual order size before you commit to it, so an unusual number gets a second look rather than a purchase order.
Can I override the forecast?
Yes, and you should when you know something the data does not. Every input is editable: the sales rate, the lead time, the cushion, and the final quantity. If you have a promotion planned or a supplier has warned you about a delay, change the input and the recommendation moves with it.
See it run on your own products
Replinish does this arithmetic for every variant in your store, every night, and shows its working on every single one.
Launching on the Shopify App Store shortly. Tell us your store and we'll let you know when you can install it. Plans are billed by Shopify, and you pick one in your Shopify admin after installing.