Skip to main content
Dec 28 2018

Making Products from Data

The influence of the AI wave is being felt everywhere. Unimaginable challenges are becoming feasible to address and efficiencies are already being achieved in many areas from chatbots to product market analysis. On the other hand, while many enterprises have great data potential, they don’t know exactly how to carve out the benefits and are under pressure to find a way forward.

Data product is a software application that realizes the multi-aspect potential of data. Single-purpose applications, which have been around for a long time, use only the one aspect of the data that is strictly relevant to the application. But increasingly, applications are deriving multi-aspect value from data that happens to become available as a side outcome of another application. In the digital era, applications are connected to a networked system where they work together to solve complex challenges by sharing their own data with each other. Even traditional software can contribute to the network of applications by using systematic innovation to create data products that follow the sequence of data to insights to actions.

Check our infographic about Transforming telecoms with Artificial Intelligence.

Think, for example, of telecommunications mediation where data records are collected from network elements and converted into a suitable format, usually for billing purposes.

First, the data that flows through the billing mediation platform can be used for other purposes. For instance, subscriber, device and location profiles can provide great insights for marketing actions targeting specific customer segments.

Second, the functioning of the mediation platform can be monitored, referred to as ‘audit logging’. For instance, an anomaly detection application can gain critical insights from the audit logs such as noting a sudden increase of data volume in a processing stream potentially leading to storage fill-up and stream failure. This can help the operations team by giving it a head start implementing corrective actions before the situation escalates.

Third, every click of the mediation platform user interface generates useful data that carries insights into how the product is being used. This, in turn, can generate feature requests to product management or upsell opportunities to sales.

In most situations there is no scarcity of data volume or the number of insights that can be generated. The question is what information matters the most. The data product must be broadly aware of what is happening and proactively bring human attention to important issues. It customizes content to make it relevant for specific user roles and provides guidance as to why an issue happened. 

Things can be complex under the hood, but the data product provides a user-friendly way to interact with the complexity of the data and suggests recommended actions. The recommendations get better over time when the system learns from the outcomes of taken actions as well as user feedback on the validity of the recommendations. Within complex domains, such as telecommunications, it is not reasonable to build a comprehensive AI system fully automating all decision-making. AI is most powerful when working together with humans, augmenting their capabilities and bridging the hardest gap in automation — the transition from insights to action.

In a data-driven enterprise with many domains of business, the data product is not a standalone point solution. It is based on a platform that exposes AI services. Existing domain-specific intelligence can be turned into a service and made available for other domains. There are also common lower-level AI capabilities that are useful irrespective of the domain, which are more beneficial to share rather than developing and operating multiple similar implementations.

Implementing data as a platform breaks organizational silos and makes AI available for everyone, even developers without much experience in machine learning. The developers can build data products independent of each other leveraging the same services without worrying about availability or other operational concerns. The services layer hides the underlying complexity, provides efficient access to intelligence and, therefore, a head start on data product releases.

To challenge yourself and accelerate the transformation to a data-driven enterprise, consider these three goals.

  1. Make every product a data product
  2. Invent new data products by combining existing ones
  3. Become the leading data innovator in your line of business!

 

Share your thoughts on this topic by joining the Twitter discussion with @nokia and @nokianetworks using #AI #IoT #innovation

About Timo Similä

Timo is an all-round data scientist who likes to be in between business and technology. He started as a machine learning research assistant in 2001, received his PhD in 2007, and worked in various practical data science and leadership roles at Xtract, Comptel and Nokia. He is currently the Head of Data Science in Nokia Software, Digital Intelligence Product Unit.

Tweet me @timo_simila