Ordering
Your best seller is sold out on Friday. The slow seller fills the stockroom.
Orders based on last year miss in both directions. sktime forecasts sales per product and store, so orders follow expected demand.
For retailers
Less cash tied up in stock.Fewer empty shelves.
Know how much to order for every product and store, instead of guessing. Free, open-source forecasting software, set up for you by the sktime developers.
Book your free 60-minute call
No obligation. We look at your numbers together.
Example forecast
How many units should I order for the weekend?
How many loaves should the store bake on Saturday?
How much stock does the promotion need?
How many checkouts should be open this afternoon?
Past salessktime forecast
- sktime forecasts
- Between 440 and 520 units on Saturday
- You decide
- Order 520 units for the weekend, so the shelf stays full.
- sktime forecasts
- Between 290 and 350 loaves on Saturday
- You decide
- Bake 340 loaves: fewer sell-outs on Saturday, less waste on Tuesday.
- sktime forecasts
- Between 2.1× and 2.7× normal sales in the promotion week
- You decide
- Order 3 pallets and plan for the dip afterwards.
- sktime forecasts
- Between 130 and 150 customers at 5 pm
- You decide
- Plan for 150 customers: open 6 checkouts at 5 pm, 3 at midday.
Sound familiar?
Where retailers lose money, and where sktime helps.
Out-of-stocks and overstock cost retailers around 6.5% of their sales worldwide. The cause is almost always planning on guesses. It shows up in five places: (IHL Group, 2025)
Fresh food
Your fresh bread is sold out by 2 pm on Saturday and thrown away on Tuesday.
With a shelf life of a few days there is no room for error. sktime forecasts per product, store, and weekday, including weather and holidays, so you order what will sell.
Promotions
The promotion worked. Now three pallets are left over.
sktime includes prices and promotions in the forecast, so you know how much stock a campaign needs and how demand drops afterwards.
Seasons
The first warm weekend empties the barbecue aisle. The reorder arrives a week late.
Seasonal peaks return every year, but never on the same date. sktime learns the pattern from your sales history, so you build up stock in time.
Staffing
Long queues at the checkout after work, idle staff at midday.
The same methods forecast customer traffic per store and hour, so shift plans follow actual demand.
What the numbers say
What modern forecasting delivers in retail.
Studies and practice reports show what forecasts based on machine learning achieve in retail. sktime brings these methods together in one open-source toolkit, so you can use them without building everything from scratch.
25%
fewer stock shortages in fresh produce
Reported on average by food retailers using AI-based planning systems, along with at least 10% fewer write-offs and up to 30% lower inventory planning costs.
McKinsey & Company, 201930%
fewer returns in bakeries
A machine-learning forecasting service cut bakery returns by 30% on average in 2022, mostly bread and rolls, according to sales reports.
Hübner et al., Journal of Industrial Ecology, 202420%
less surplus stock at Otto
Otto uses AI forecasts to steer part of its purchasing and reduced its surplus stock by a fifth.
The Economist, 2017These figures come from studies and practice reports on machine-learning forecasting in retail, not from sktime projects. We work out what is realistic for your business in a first conversation, using your data.
sktime is open source: no license fees per user or product.
Your numbers
What could this be worth for your business?
A back-of-the-envelope estimate. Enter your annual revenue and gross margin and see a rough range in euros. Everything else is a typical retail value from public sources.
Roughly estimated annual savings
–/ year
– of your revenue
CautiousExpectedOptimistic
––
- Lower cost of holding stock
- – – –
- Extra profit from fewer empty shelves
- – – –
- Cash released from stock (one-time)
- – – –
- Money that no longer sits on your shelves and in your warehouse.
Not included: the cost of setup and support. After a first conversation we give you a fixed price for a trial project.
How we calculate this
Two effects, three steps. Revenue and gross margin come from you. The percentages are fixed assumptions from the sources below.
1. Surplus stock
revenue × 8% stock on hand × 35% surplus share
Stock on hand is about 8% of annual revenue (roughly 6% in grocery to 10% in fashion). Around 35% of it is surplus: slow-moving, out of season, or more than needed. At €50M revenue that is €1.4M.
2. Lower cost of holding stock
surplus × 10 to 20% less surplus × 25% holding cost
Better forecasts mean you order less excess. Otto reported 20% less surplus stock, so we use 10% (cautious) to 20% (optimistic). Holding stock costs about 25% of its value per year: capital, storage, markdowns, write-offs. The stock you no longer hold is the one-time cash released. It does not count toward the yearly savings.
3. Extra profit from fewer empty shelves
revenue × 4.1% lost sales × 10 to 25% fewer × gross margin
Empty shelves cost retailers about 4.1% of sales. McKinsey reports 25% fewer shortages in fresh food, so we use 10% to 25%. Outside fresh food, expect less. Only the margin on a recovered sale counts as profit.
Annual savings = step 2 + step 3. The expected value is the midpoint of the cautious and optimistic case.
Simplifications: we use gross margin, which is a little more generous than margin after variable selling costs, and we treat both effects as independent.
This is an illustration, and we do not promise these results. The reduction rates come from reports on other companies, not from sktime projects, and lost sales from empty shelves vary a lot between businesses. In the call we will explore options for your own use case.
Before you ask
Questions retail executives often ask.
- We don't use any AI yet. Is it too early for us?
- No. Demand forecasting is a good first AI project because you can measure the result directly in stock levels and sales. A pilot on a slice of your data shows the effect before you roll forecasts out across the business.
- Is our data good enough?
- Usually, yes. Sales history per product and store from your ERP or checkout system is enough to start. Prices, promotions, and holidays improve the forecast and can be added later. In early pilots, data preparation and cleaning may be needed. We aim to make the results reusable and, ideally, to improve data quality and governance across the company.
- Does our data stay with us, and will it work with our systems?
- sktime is designed to run on your own servers or in your own cloud, so your data doesn't have to leave the company. It works with the sales data your ERP or checkout system already exports, and the forecasts go back into your ordering process as files or through an interface. Because sktime is open source and developed by a broad community, you can check what it does at any time and are not tied to a single vendor. The core framework has a permissive license forever, with no license fees and no subscription requirement.
- Are we in control of the software in our own system?
- Yes. The core framework is open source with a permissive license forever, guaranteed by the non-profit GC.OS. Custom code built for you does not run on proprietary systems elsewhere, and you are not locked out of it. You choose the maintenance model, and we always discuss it with you in advance: either the code is handed over to your team completely, or you pay for continued maintenance.
- Who do we sign with, and what if a developer leaves?
- Depending on the type of engagement, your contract partner is the German Center for Open Source AI (a non-profit) or sktime Enterprise Solutions GmbH in Ulm, Germany, which provides services and software around sktime. For support after setup, we offer service level agreements (SLA). The non-profit setup keeps maintenance of the framework layer secure in the long term. An SLA is a legally enforceable guarantee that enough developers are always available to help maintain your infrastructure, even if someone leaves.Meet the team
- Do we need to hire new people?
- No. But we believe in the sovereignty of our users, and that means helping you build a competent team instead of locking you into endless consulting contracts. Our work usually comes with training for your internal teams, for example on AI, forecasting, or Python, so you can operate the system yourselves quickly. If you prefer, we can also handle most requests with our own staff, forward deployed engineers, or freelancers. Your team then carries a minimal load, and your planners spend less time in spreadsheets and never have to write code.
- What does it cost?
- The sktime core framework is free, with no license fees per user or product, now or later, and it is governed by a non-profit. You pay for the setup and support you order. Pilots have a fixed price, always agreed in advance. Larger projects have defined deliverables, and we estimate and cap the price in a scoping phase before they start. Custom software or models that we maintain for you may incur costs through an SLA or licensing. We always discuss the contract and billing model with you in advance.
How we start
Three steps to forecasts you can trust.
- 01
First conversation and scoping
We look at your data and where planning costs you the most today. From that, we set the scope and a fixed price for the pilot.
- 02
Pilot on your data
Over two to six months, we build a first forecasting system on your own data. You get solid benchmarks, all the code, and a report, and decide from there how to continue.
- 03
Use in daily planning
The forecasts go into your regular planning process. Under a service agreement, sktime Enterprise Solutions maintains the system and stays your contact when questions come up.
Your first step
Let's look at your numbers together.
sktime is free and open source. In a free 60-minute call, the people behind sktime look at how you plan and order today, and what better forecasts could be worth for your business.
sktime is downloaded over a million times a month and built by more than 500 contributors worldwide. Its core team now sets it up for you.
Book your free 60-minute call
No obligation, no sales pressure.
What happens in the call
- 01
Your situation
How you order and plan today, and where it costs you the most.
- 02
Your data
What you already have, and what a first forecast would need.
- 03
Your numbers
A first, rough estimate of what it could be worth, and what a pilot would look like.