Your new approach in Demand forecasting with datrics.ai

Olena Petashchuk
Datrics
Published in
4 min readApr 3, 2020

How to build Demand Forecasting in Retail: Datrics use cases

Would you like to predict future results for your business? Using datrics.ai demand forecasting models with complex criteria you can make accurate data-driven decisions on inventory and replenishment management, manufacturing processes, product pricing and much more.

Sales forecast 2014–2026

Demand forecasting outcomes with datrics.ai

Datrics is a web-based end-to-end data science platform that allows people to extract significant insights and build advanced machine learning models without a single line of code. Our team has created template solutions for particular issues to provide our users with more benefits. It doesn’t matter whether you’re a top data scientist or have limited data science and machine learning background, you can leverage our platform for this as well as to automate some routine tasks. Datrics.ai has a fully managed environment to automate MLOps activities required. Spend less time preprocessing the data, be focused on outcomes.

Predictive dashboard at Datrics platform

Our team decided to create the first template for demand forecasting. The matter seems quite straightforward: you have the historical sales data and you would like to understand the future sales performance of the same product. You can take the “pipeline” which was created by our Data Science team, upload your data, do minor tweaks and have a working forecasting model with predictions for the next week, month or quarter. Based on these predictions you can: manage your inventory better and avoid overstocking or running out of goods, optimize the prices, calculate your financial and sales KPIs and have data-driven assumptions for decision making on SKU or category level.

Which model to choose for demand forecasting?

There are plenty of different models that can be used when trying to undertake demand forecasting issues. But the most intuitive ones use regression analysis to predict the future behavior of data, like Autoregressive moving average model (ARMA). ARMA and other generalizations and variations of it are popular when doing time series analysis and forecasting. In a couple of recent years, the Prophet algorithm had made time series forecasting easier. It handles missing values and outliers well and practically does not require knowledge of time series analysis or fine-tuning of the algorithm. So a person can easily create powerful forecasts without actual practical experience with the analysis of time series for forecasting demand in particular. That’s the reason why we recommend using Prophet for demand forecasting.

The Prophet algorithm was developed by Facebook and aimed at forecasting the behavior of the time series. The key thing here is that you don’t need to know under-the-hood dynamics, you can apply it in a few clicks, including advanced seasonality effects. It is resistant to outliers and missing data and behaves pretty well without extra fine-tuning required if you have enough historical data. To evaluate the model we’ll use Mean Absolute Percentage Error (MAPE). It shows you by how many percentages points your forecasts are off, on average. The numbers you received are useful when comparing different models and give you a general feeling of how the model can perform, so do not chase them. You should be focused on the benefits you get from demand forecasting — making accurate data-driven decisions, rather than based on the assumptions.

Historical and predicted sales 2013–2019

Demand forecasting has never been so easy

Different products or industries require different efforts to make a highly accurate assessment. Therefore, you should not chase the metrics numbers, and there are no rules of thumb when it comes to demand forecasting. Just focus on the more extensive issue you’re trying to solve and our platform will help you to get the best results.

Who can test the Alpha version? Will the background of business analysts be enough?

As we mentioned before you do not need to have any data science, ML or coding experience to use dartics.ai platform. You can easily import, preprocess the data and run the demand forecasting model with no coding if you know your business domain and understand which data can influence your demand.

Check out the demo recording of the Alpha version with the ready pipelines for the demand forecasting.

Datrics.ai will create models and pipelines for you, which are robust to outliers and able to catch trends, seasonality and different dependencies. Based on these models and pipelines you will be able to make an accurate decision when you have enough data. Users can still experiment with data processing by trying different models and analyzing metrics and results to select the most suitable model.

Try our alpha version on https://platform.app.datrics.ai/signup.

More info about the company on https://datrics.ai/.

--

--

Olena Petashchuk
Datrics
Editor for

I am one of the partners at Datrics — Data Science platform helps SMEs to leverage from own and public data without writing a single line of code.