Innodata has developed an intelligent pricing system for large businesses

«Иннодата» разработала Систему интеллектуального ценообразования для крупного бизнеса
Press release
13 February 2018
Innodata presents the Intelligent Pricing System designed for automated and accurate forecasting and balancing of prices and rates. The solution will be in demand among developers in construction, retail, transport and logistics companies and large service organizations with a constant range of services.

The high probability of errors due to the "human factor" in manual forecasting of supply and demand makes the pricing process more complex, requiring long-term, laborious and expensive research. In order to ensure efficient sales, many factors must be taken into account to set the best price. The innovative Intelligent Pricing System developed by Innodata will help do this relatively quickly and efficiently.

How the system works

Using technologies based on BigData and neural networks, the company’s specialists have developed a qualitatively new approach to post-processing of incoming data that provides the maximum efficiency of the mathematical models built, reduce the errors, and increase the interpretability of the results. This system:

  • Creates a basic model for predicting the pricing dynamics, identifying the main visible and hidden factors affecting the variation dynamics;
  • builds, optimizes and monitors business models;
  • performs fine tuning of parameters and variables that affect its operation;
  • enriches the model with additional data.

The process of generating recommendations and comments on pricing is automated.

The system works as follows: three blocks of information are generated on a daily basis for system users. The statistics block provides an interactive report that includes indicators related to sales dynamics, price levels, customer activity, etc. It also provides generating reports of various degrees of aggregation, ranging from the summary indicators of the company to the level of a specific property item. The forecast block provides the sale probability for a property item in the next month, which is updated daily. The forecast results can be aggregated down to the apartment type level and even down to the level of stacks in a particular section of a project.

The block of recommendations includes daily updated values of the price change rate for real estate objects, types of apartments, and stacks. At the same time, the recommendations can be customized by the user and changed dynamically depending on whether it is possible to change prices which are calculated for a particular real estate object based on the analysis of the input system data.

Business indicators

The main business tasks solved by the Intelligent Pricing System are as follows:

  • Maximizing revenues
  • increasing sales without increasing costs
  • prompt response to events affecting pricing in a highly competitive market
  • forecasting the pricing dynamics
  • accounting for a number of affecting factors,
  • minimization of the human factor.

The system allows forecasting sales and the best period for price changes, reducing human efforts for the price formation process due to optimization of the business process, and provides online support. Unique solution capabilities include: assessment of transaction probability, calculation of a daily forecast for each transaction, grouping of results, and price management based on the actual demand for the object. If the forecast for actual demand exceeds the planned value, then there is an opportunity for more frequent price increases.

The solution model is balanced and provides for about 200 variables, including seasonal factors. It efficiently uses both internal and external determining factors, such as, for example, fluctuations in currency exchange rates.

Implementation effect

The results obtained by the system are achieved using modern self-learning algorithms for a mathematical model (for example, using XGBoost). An analytical model is built on the basis of several developed methods. The model takes historical data into account and uses it efficiently. The final model training is performed in real time. 90% of accurate transactions falls exactly in the period that is covered in the model. Provided that 85% or more of data completeness is provided, the model correctly predicts the statistics of expected transactions.

“The main effect of using the Smart Pricing System in the company's business architecture is to achieve the main goal of maximizing revenue without increasing costs,” comments Maxim Sytnikov, Product Owner of the solution, Innodata. “The business effect of using the system is difficult to overestimate: first of all, it is revenue maximization without increasing costs, increasing the competitiveness level, stimulating demand, increasing revenue, fine-tuning value variations by predicting a future transaction, checking the expediency of recommendations and elasticity of demand in real time, increasing additional profit due to a flexible approach to data. And as a nice bonus, you’ll get an optimization of labor costs, for example, of the analytical department that performs pricing manually, as well as an acceleration of decision-making from a day to a few minutes.”