Perne
Demand forecast: illustration 1
Data and AI analytics

Demand forecast

We work as a team. We build a demand forecast by product for several weeks or months ahead, based on sales history. We compare simple methods with models from Python libraries, such as statsmodels or Prophet, and pick the one with the lower error on a test period. The result is a forecast table by product and week, and a note on how far it can be trusted. The model does not replace the buyer; it helps plan stock.

from 60,000 ₸ (≈ $120) one-off

timeline: 7–14 working days

  • A forecast by product and by week or month for the chosen horizon
  • A comparison of 2-3 methods on a test period
  • An error estimate for each product or group

Order

Message on WhatsApp
TelegramCall

The service name is added to the message for you.

This is a starting price; we give the final amount before work starts. USD is approximate.

What is included

  • A forecast by product and by week or month for the chosen horizon
  • A comparison of 2-3 methods on a test period
  • An error estimate for each product or group
  • A purchasing table with recommended quantities by date
  • Export of the forecast to Excel or Google Sheets
  • A short report on what is forecast reliably and what is not
  • A script or notebook to recalculate the forecast on new data

Who it is for

For shops and companies with regular sales that buy stock in advance and want less frozen stock and fewer empty shelves. It fits if there is at least a year of sales history with clear dates. It is not a fit for products without history, one-off projects, or businesses with chaotic demand that never repeats.

Not included

To avoid surprises, here is what the base price does not cover and what is discussed separately:

  • A guarantee of forecast accuracy: the error depends on how stable demand is
  • Connecting the forecast to your warehouse system in real time
  • A permanent service or API with model retraining: this is a separate ML task
  • Purchasing decisions on your behalf: the forecast is a hint, not an order

FAQ

After payment you receive the forecast table, the script or notebook with the calculation, and the report, and you can use them in your company without limits. The code itself does not contain your data, and we do not publish it on your behalf. Third-party libraries are distributed under their own open licences, and we will list them for you.

The price includes one round of fixes: a different forecast horizon, excluding an abnormal period, or adding one factor such as promotions. Switching to a new, more complex method or forecasting other cuts is priced separately. Send comments as one list.

Usually 7 to 14 working days from the moment we receive the sales history and agree on the forecast horizon. The timing depends on data cleaning and the number of products. If the history is short or messy, the timing grows, and we will warn you in advance.

Sales history by week or day: date, product code, quantity and, where possible, price. A list of the products that matter most to forecast, and any promotion or delivery periods. The horizon you need the forecast for, and how you place purchase orders.

Accuracy depends on the history: for stable products the error is usually small, and for new or seasonal ones it is noticeably larger. We show the error on a period the model has not seen, so the number is honest. We cannot promise 100 percent accuracy and do not guarantee it.

No, for this service an export of sales and stock into a table is enough. The forecast is calculated from that export, and it can be refreshed weekly or monthly. If you need real-time work, that is a separate task and it is priced by its scope.

You may also need

Sales and customer base analysis

Sales and customer base analysis

A review of sales and the customer base: which products and clients bring money, where buyers drop off, and what to do first.

from 30,000 ₸ (≈ $60) one-off

Timeline: timeline: 4–8 working days

A custom ML model for your task

A custom ML model for your task

We train a machine learning model on your data so it solves a specific task: a forecast, a classification or anomaly detection.

from 150,000 ₸ (≈ $300) one-off

Timeline: timeline: 14–30 working days

Shall we discuss your project?

The product name and a link to this page are filled into the WhatsApp message for you.

Message on WhatsApp
TelegramCall

Or send a request

We will reply in the channel you choose: WhatsApp, Telegram, phone or email. Your contact is used only to reply.

Service
How to reply

from 60,000 ₸ (≈ $120) one-off

Message on WhatsAppTelegramCall