The most accurate, actionable and user-friendly AI-econometrics powered tool to optimise pricing strategies in the Consumer-Packaged Goods sector

Project Information

ATHENA 2.0

Grant agreement ID: 849543

DOI

10.3030/849543

Project closed

EC signature date 28 February 2019

Start date 1 March 2019

End date 31 August 2021

Funded under

• INDUSTRIAL LEADERSHIP - Innovation In SMEs

Total cost

€ 2 155 750,00

EU contribution

€ 1 509 025,00

Coordinated by

WISE ATHENA SL
Spain

Periodic Reporting for period 2 - ATHENA 2.0 (The most accurate, actionable and user-friendly AI-econometrics powered tool to optimise pricing strategies in the Consumer-Packaged Goods sector)

Reporting period: 2020-03-01 to 2021-08-31

Summary of the context and overall objectives of the project

Product pricing is the most critical element for FMCGs and CPGs companies, but it is very complex to optimise them due to the high number of factors involved and their interrelations.
When defining prices for their products (SKUs) CPGs need to consider price elasticity, cross-elasticity, cannibalisation, competitors' pricing strategy, etc. Additional complexity comes from customers’ micro-segmentation, type of retailer and channels. In addition, prices move on a daily or weekly basis. As consequence, fixing the right price for products has become a major challenge for FMCGs/CPGs companies. The problem is so complex that current state-of-the-art reaches accuracies around 60%-70% in demand prediction.

The combination of Artificial Intelligence with Econometrics for the CPGs sector is determining an inflection point for revenue management and pricing departments in this industry. During more than 4 years we have been developing the algorithms, models, AI modules, machine learning and big data that is able to find the optimal prices, creating an accurate, affordable and easy-to-use tool that redefines the future of products’ pricing. The technology applied so far is offering to the customer a stable and friendly software: Athena Console. Our main goals to do so are (i) to automate our Data Processing and Data Science processes that currently need to be executed or supervised by a data scientist, so we can scale the product in the market and (ii) create an Intelligent Agent for the Board, integrating a natural language processing and generation tool to allow more natural interaction between Athena Console and our customers.

The project objectives are:
Objective 1: Improve scalability and efficiency through processes automation
Objective 2: Enhanced natural language interface and self-learning capabilities for an outstanding user experience
Objective 3: ATHENA 2.0 validation and platform fine-tuning
Objective 4: Enhancing market positioning

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The Reporting Period 2 was mainly focused on WP3 as technical work and WP1-WP4 as transversal tasks.
The main objective of WP3 was to demonstrate accurate performance and usability of the enhanced platform under real world conditions. Pilots were focus on demonstrating the 2 key extensions developed:
a) The Intelligent Agent was preliminary tested by 3 real clients and then was progressively introduced to the rest of our actual clients.
b) The automation of Data Science workflow for each customer was tested by our Data Scientists and IT teams during the pilots’ set-up and re-calibration iterations after every promotional cycle (every 4 months).
Beta testers were recruited from existing or prospective customers under real business conditions. Training, technical, helpdesk and customer success support was offered to beta testers to effectively set-up and use new Athena’s features, and they were asked to provide feedback on usability, effectiveness, and overall user experience.
The enhanced beta version of Athena incorporated an intelligent agent as a recommendation engine and visualization of the future impacts of the decisions made in the present. Once the new developments of Athena 2.0 were incorporated, pilot testing of the recommendation engine of the intelligent Agent was a success, demonstrating that this feature provides value to the users.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

We will analyse the existing workflows at the Data Science and Data Processing teams, where we will develop automations. In both areas the workflows are long, sometimes tedious, and implies too much manual execution and supervision. First, data needs to be cleaned up to make sure the model gets good figures so it can show results with 90% accuracy. Nowadays, 90% of this workflow is done manually. Second, data science team needs to find the right model after training thousands of them, taking in consideration price elasticity, price cross-elasticity, cannibalization, competitors pricing strategy, special promotions, seasonality and pass through prices to final customer, among others. Currently 60% of the Data Science team workflow is done automatically.

In this project we have identified potential automation improvements and define a development and integration roadmap to reach Athena 2.0 beta version, including the technical validation methodology. We explained the process for the automation of execution of the current scripts used to generate the pricing and trade promotions optimizations. This made automatic 80% of the current manual tasks for Data Processing team and 85%-90% for Data Science. Also, we have defined and develop automate deep quality assurance modules (Semantic QA), that has allowed to early detect any problem based on previous iterations.

In addition to automations, we have built a recommendation engine to find the most relevant, actionable insights for the customer and predict their interests using Reinforcement Learning (AI) techniques.

equipo.png