Forecasting

How to forecast a new product with only three months of sales

Most forecasting advice assumes a year of history. A new product has weeks of it, and it is usually the one you are about to commit a container to.

Marcus VolstedCo-founder, Restocio Published 11 September 2026Updated 11 September 20269 min read

How to forecast a new product with only three months of sales

With three months of sales you can measure how fast a product sells and roughly which way it is heading, but you cannot see its season, and nobody honestly can. So measure the rate over the days the product was actually on sale, and size the first reorder to last a shorter stretch than usual, so the next order is placed on more evidence.

The timing is the cruel part. A new product is where an importer commits hardest, to a supplier's minimum and a 70-day boat, on a few weeks of data. The formulas live in the forecasting formulas guide. This page is about feeding them a quarter of a year.

What can three months of sales actually tell you?

Two things, with care. The first is a sales rate, as long as you divide by the days the product could actually be bought. The second is a faint hint of direction. What it cannot give you is seasonality, because you have not seen one season yet, let alone the same one twice.

The direction is weaker than it looks. Drop the part-months at either end, since a launch month with nine days of sales is not a slow month, and three months often leaves two whole ones. Two points and a ruler always give you a line. Do not bet a container on it.

Seasonality is where to be blunt. Eurostat's guidelines on seasonal adjustment, 2024 edition, read on 11 September 2026, say it outright. "Very short series (less than 3 years) cannot be seasonally adjusted using either moving average or model-based methods." The UK's Office for National Statistics, read the same day, adds that after a seasonal pattern changes "it can take between three and five years" to pin the new one down. If the agencies whose job this is want three years, anyone reading a season out of your three months is guessing.

Why is the divisor so much more dangerous on a new product?

Because the same stockout is a far bigger share of a short history. Fourteen empty days in a year of sales is under 4 percent of the record. Fourteen empty days in a product's first eight weeks is a quarter of everything you know about it, and dividing by calendar days treats that quarter as zero demand.

Put numbers on it. A product launched eight weeks ago has sold 280 units. It sold out in its second week and sat empty for 14 days. Divide by a standard 60-day window, four days of which came before the product existed, and you get 4.7 a day. Divide by its 56 days of life and you get 5. Divide by the 42 days it was on sale and you get 6.7, about 43 percent above the 60-day figure, on the product that just proved it sells out.

That divisor is the whole point of the reorder point calculator. On a mature product it is a refinement. On a new one it is most of the answer.

How big should the first reorder be?

Smaller in one place. Keep the days that cover the voyage and the cushion for a late boat, because those protect the shelf until the goods land. Shorten the stretch the order is meant to last afterwards. That is the part you can trim without touching that protection, and it means more of your buying waits for more evidence.

The days of stock guide calls these pipeline, cushion and cycle days, and only the cycle is a free choice. Here is the product above at 6.7 a day, on a 70-day boat with a 60-day cushion, in hand arithmetic.

	Order lasts 90 days after landing	Order lasts 30 days after landing
Days the stock position is sized for	220 (70 at sea, 60 cushion, 90 selling)	160 (70 at sea, 60 cushion, 30 selling)
Target for stock on hand plus on order	About 1 470 units	About 1 070 units
Bought too many if the true rate is half	About 730 units	About 530 units
Of those, caused by the selling days	About 300	About 100
Orders a year at this pace	About 4	About 12
Likely to clear a supplier minimum	More often	Less often

Read the fourth row first. Of the 730 units the long cycle overbuys at half the rate, only 300 come from the cycle. The other 430 are voyage and cushion, identical in both columns, because nobody gets to plan on a shorter boat. Cutting the cycle is a smaller lever than it looks, and the only one that leaves your late-boat protection intact.

The short column is not free. The 400 units you skipped are the ones you would sell if demand runs stronger than you measured, and more orders means paying more often for whatever your forwarder charges per booking. What you get back is three smaller chances to be wrong instead of one large one.

What if the MOQ takes the choice away?

It often does, and then the short cycle is gone before you get to choose it. If the minimum is above your short-cycle quantity, you are buying the long column whether you meant to or not. The useful question stops being how much to order and becomes whether this product has earned that minimum yet.

The MOQ guide has the surplus arithmetic, and says a product with no history cannot run it. Three months changes that a little. You have a rate, just not one to lean on, so run the sum twice, at the rate you measured and at half of it. Half is not a statistical bound, just a deliberately gloomy number. Pick your own.

If the surplus at the gloomy rate would outlive the product's season or tie up cash your proven lines need, the minimum is bigger than your evidence. Negotiate it down, or let an unproven product run short for a while. Running short costs sales. Overbuying on a guess leaves stock you may be discounting next year.

How does Restocio handle a product with only weeks of history?

It measures the rate over the product's own life, counted from the day the variant was created in Shopify, until the product outgrows the default 60-day window. Days the app recorded it as sold out come out of the divisor. For its first 14 days, the card's calculation details note that the product is new and that its rate will swing.

That note moves no quantity. It tells you the number underneath is thin. Seasonal handling will not rescue a new product either: a product marked seasonal is sized on last year's same months, and a new one has no last year, so it falls back to its recent rate. The Demand page forecast does not help here: it covers the whole shop rather than one product, changes no order, and needs two full years before it fits a seasonal pattern, while the page reads at most one.

On a boat, Restocio sizes an order so stock on hand plus on order covers the voyage, the cushion and one more voyage: 200 days of sales on a 70-day boat with a 60-day cushion, with rail or air added if the shelf will empty before the boat lands. Days of stock only changes that when it is longer than that span, which on a 70-day boat it cannot be. To give one new product a shorter cycle, type its quantities on the card. A quantity you type on a freight line replaces the app's number for that line, supplier minimum included. Lines you leave alone keep the app's numbers. The full loop is in the Shopify replenishment guide.

Where can this still go wrong?

Mostly in the days before you connected the app. It only counts a day as sold out if it recorded the stock that day, and recording starts at connection. A new product that ran out earlier has those empty days counted as ordinary selling days, so its rate comes out too low. Check that one by hand.

Two more worth knowing. Age comes from the day the variant was created in Shopify, and a move from another platform cannot carry that date across. For the first 60 days after a move a rate can be wrong either way: too high if your old orders were imported, because up to two months of sales get divided by a few weeks of age, and too low if the products sat in Shopify before the shop went live. A product you delete and re-create under the same code can run high the same way. Check those by hand until they are 60 days old. And nothing in the app picks a comparable product or decides what a successor inherits from its predecessor. That call stays with whoever places the order.

Common questions

Can you forecast a product with only 3 months of sales?

Partly. Three months is enough to estimate how fast a product sells, provided you divide units sold by the days it was actually in stock, and to get a weak sense of whether it is speeding up or slowing down. It is not enough to see a season. Treat the result as provisional and size the first reorder to match.

Can you work out seasonality from three months of data?

No. Three months has not covered a single year, so there is no repeat to measure. Eurostat's guidelines for official statistics say a series shorter than three years cannot be seasonally adjusted with the standard methods. A seasonal factor built from three months is a guess, and it multiplies straight into your order quantity.

How much should I order on a new product's first reorder?

Enough to cover the voyage and a cushion for a late delivery, plus a shorter selling stretch after landing than a proven product gets. The voyage and cushion protect the shelf. The selling stretch is where overbuying piles up if your rate is too high, so trimming it moves more of your buying to when you know more.

My new product sold out in its first weeks. Is its sales rate wrong?

If it was calculated over calendar days, almost certainly, and too low. Divide units sold by the days it was actually available. On a short history a fortnight of empty shelf can be a quarter of the record, so the correction is large. In the example on this page it lifts the rate from 4.7 to 6.7 a day.

Should I accept the MOQ on a product with three months of history?

Work out the surplus at your measured rate and again at half of it. If the surplus at the lower rate would outlast the product's selling season or tie up cash your proven products need, the minimum is bigger than your evidence supports, and it is worth negotiating before you sign.

Can I use a similar product's sales to forecast a new one?

Often it is the best starting point you have, especially when the new product replaces an older one. Treat it as a judgement rather than data. Decide how much of the older product's demand you expect to carry over, write the number down, and check it against the new product's own sales as they arrive.

Marcus Volsted

Co-founder, Restocio

Marcus co-founded Restocio and works on it daily with a Swedish importer who plans their purchasing in it every working day. Restocio is built in Sweden by two founders, one Swedish and one Danish, and Marcus is the Danish one. Why we are building it.