Statista_

AI prototype for a statistics portal: Reduced operational costs by 65% and increased data quality by 140% in just four weeks, improving customer experience and efficiency.
CLIENT
Statista
YEAR
2023
INDUSTRY
Market Research
ProjeCt CategorY
C2 - Innovation

Executive Summary

AI prototype for statistics portal

The statistics portal Statista is developing a new AI-based data research service that combines the vast capabilities of modern LLMs with the exclusive facts from the Statista data pool. The goal of the collaboration between AI experts from the TalentFormation Network and Statista's in-house team was to increase the speed of data retrieval and response generation, improve response quality, and significantly reduce operational costs.

The project team's innovative, data-driven approach led to a comprehensive redesign of the prototype using state-of-the-art LLM technologies. Through customization and over 100 targeted experiments, the performance of the system was significantly improved. The impressive results of this project - a significant reduction in response times while reducing costs and improving quality - demonstrate the impact of external expertise and innovative collaboration.

PERFORMED
experiments

100

reduced
latency time

10%

Decreased
operating costs

65%

improved
data quality

140%

The Starting Point

Statistics portal uses AI-based service to improve customer experience

Statista, one of the world's leading statistics portals, provides users with access to an extensive database of statistics, forecasts and infographics. To further improve the user experience, Statista developed a prototype for AI-powered research that needed to be optimized in terms of data quality, speed, and cost. With rapid developments in the field of artificial intelligence and limited internal resources, Statista sought a solution to make the prototype more efficient and perform better.

IN THEIR WORDS

"Matthias and his team were a stroke of luck for us: the combination of AI expertise in breadth and depth, the very systematic evidence-based approach and the immediate fit with our internal team on a human level is rare. Without exception, we immediately implemented all findings from the collaboration, are still benefiting from the valuable impulses, and will certainly be going into further rounds with Matthias and his team!"

Dr. Ingo Schellhammer

Chief Technology Officer

The Goal

Faster and more accurate search results

In collaboration with TalentFormation, Statista set out to improve the performance of the prototype to provide users with relevant statistics for their specific questions quickly and accurately. The focus was on accelerating data retrieval, improving response quality and speed, and significantly reducing operating costs to increase competitiveness. The project also aimed to build internal AI expertise.

Project goals:
- Improved data quality
- Increased speed
- Reduce costs

Ai for your Business_

Want to harness the full potential of artificial intelligence for your business?

Our Approach

Collaborate effectively with external experts and internal teams

Heureka Labs' experienced AI experts from the TalentFormation Network used a data-driven and experimental approach to define clear performance indicators for quality, speed and cost. Through close collaboration between the TalentFormation team and Statista's internal development team, we were able to implement significant measures within four weeks:

Analysis and New Development
The existing code was extensively analyzed and rebuilt using the latest frameworks and technologies. Special attention was paid to improving the architecture and using more efficient tools.
Optimize Data Management
By adapting the database indexes and data storage, we have significantly improved the speed and quality of search results.
Perform Experiments
To find the best solution for Statista's needs, we tested different database indexes, different embeddings and LLM configurations, reranking and filtering algorithms, and advanced retrieval techniques such as multi-query retrieval and HyDE in over 100 experiments.

The Outcome

Impressive results in record time

The results speak for themselves: in just 4 weeks of the project, we achieved an average reduction in response times of 10%, with peaks of up to 65%. At the same time, operating costs were reduced by an impressive 65% and response quality increased by 140%. These results demonstrate not only the effectiveness of our approach, but also the potential that lies in combining expert knowledge and collaborative working.

100
Performed Experiments
10%
Reduced Latency Time
65%
Decreased Operating Costs
140%
Improved Data Quality

The TalentFormation Factor

This project highlights the value of external expertise and a collaborative mindset. The experience of the experts from the TalentFormation network not only led to significant improvements in the prototype, but also laid the foundation for the further development of Statista's internal AI capabilities and resulting technological competitiveness.

Ai for your Business_

Want to harness the full potential of artificial intelligence for your business?