Too Much or Too Little Data - How To Optimize Digital Performance

Melania Calinescu, Head of Data Science

Delve into the history of data analytics to better understand what drove the AI revolution here at

Reading the news on your phone, streaming a favorite TV show, sharing photos of your most recent holiday, video calling your mom – every digital action we take adds data in the highly dynamic and exponentially expanding digital ecosystem. Businesses struggle to enhance their analytics capabilities in order to adapt to this new, data-heavy reality and keep up with their customers’ needs. 

Before I dive into examples of advanced predictive analytics technologies for optimal decision making, let’s trace what brought the industry to this point (see the figure below). Every breakthrough in the history of data analytics facilitated deeper understanding of the customer and the market in order to make better business decisions. 

Starting from consumer panel analytics that offered market benchmarks through statistical inference, the next advancement came in the form of decision-support systems to answer the demand for scalability and automation in data collection and analysis. With accumulating data and continuous advancements in computers, data mining and the need to understand patterns in large data sets brought cloud analytics. From here on, answering questions around what items to recommend to customers, what customer behavior characteristics inform future business decisions and what market patterns can be predicted from limited business trends ignited the application of large-scale machine-learning algorithms. We find ourselves today in search for broader and more sophisticated artificial intelligence technologies as intelligent business development is key to staying competitive in today’s digital world.

How can businesses know that their digital performance improvements are good enough to keep them ahead of the competition? Adequately answering this question is what drove the AI revolution here at We started this journey with small steps in 2017, with our first ML-powered product aiming to measure how people use their mobile devices, what apps keep them more engaged and what apps fall out of the market. With access to only sparse data on mobile usage patterns in the population, creative application of robust multidimensional cluster regression algorithms still achieved quality market trend benchmarks across the entire mobile ecosystem. 

As we continued to increase focus on AI/ML systems as the foundation to product development, we expanded our portfolio of algorithms, adding multiadaptive regression splines, scalable support vector machines and robust hierarchical multilabel classification algorithms. While mobile market performance has several numerical performance indicators, we broadened our attention to also address text, image and audio data in order to provide reliable market analyses on app user reviews, app screenshots, image and video ads. In 2021, we helped app publishers both make better sense of their own data through anomaly detection algorithms as well as more effectively understand drivers of significant market shifts through event attribution algorithms and dissect the app economy at deeper class and genre granularity through ensemble algorithms involving deep learning language models.     

For businesses committed to data-driven decision making, navigating their own large digital datasets while having too little data on the competition may reduce their decision effectiveness. can enable such companies by providing the performance goalposts for benchmarking, addressing the too little data problem. can also augment and enrich existing first-party datasets to provide additional granularity and filtering, and solve the too much data problem. All supported by a rich AI technology portfolio! Stay tuned for our next posts where we dive deeper into some of our new products and the underlying AI/ML systems!

March 18, 2022 News

Related blog posts