Adaptive Reporting:The Treasure in the Data Lake

Executive Summary
Full article published in issue 6/2020 of REthinking: Finance

Companies’ investments into data / analytics infrastructure and processes are justified by the underlying assumption that business data is a valuable raw material, and that processing this data for decisioning results in a competitive advantage. In this context, optimal management reporting needs to address four key aspects:

1. Content — What is being reported, i.e., which KPIs and which analyses?
2. Access — How is the reporting process structured, incl. following which organizational and technical setup?
3. Timeliness — When should reporting take place, e.g., at which frequency or dependent on which events?
4. Audience — Who is being reported to, incl. potentially multiple orthogonal target audiences?

Innovations over the past decade have led to key advances in answering these questions for the three aspects of access, timeliness and audience, but the very first aspect and arguably most important aspect has been lacking: The content of management reports has yet to see similar qualitative leaps as the other three aspects recently have.

The predominant way to select the content for management reports is a top-down process, following historical and/or industry-specific reporting conventions, at times supplemented by a strategic, hypothesis-driven approach to add supplementary KPIs. Crucially, business data itself is only used to populate a given reporting structure, but hardly ever to decide upon the content of the reports. This leads to a psychological bias in current management reporting called the ‘streetlight effect’ in which stakeholders focus primarily on information that is readily available, to the detriment of situational awareness and decision quality.

The alternative is a bottom-up, data-driven approach that we call adaptive reporting. Adaptive reporting enhances existing approaches by using the day-to-day stream of changing business data itself, i.e., the changing patterns found in this data stream, to decide what to include in a report for the target audience to consume and base their decisions upon. The key advantage of adaptive reporting is that it allows stakeholders to actively address the unknown unknowns based on the traces they leave in a company’s business data. For example…

• A leading global bank used the patterns surfaced by adaptive reporting to identify a previously unknown network of cybercriminals and fraudsters
• A leading German bank found a previously unknown process gap in their credit risk assessment / management process that increased their exposure to the failure of high-risk B2B credits
• A leading German industrial insurance uncovered counterproductive internal competition between their regional business units in client acquisition and key account management activities

Three steps of the bottom-up analysis and decisioning process
that drives adaptive reporting

Adaptive reports like these examples are usually derived in a three-step process that frontloads data analysis and pattern detection, then proceeds to validate and prioritize the analytical insights, and finally pushes for decisions based on highest priority results. This process — and the resulting increases in situational awareness and decision quality — can however not be implemented using traditional means of data analysis and report generation, because a team staffed with human analysts will not have the capacity to constantly mine through continuously changing flows of business data, no matter the size of the team.

Instead, the data analysis process needs to be automated end-to-end, from data ingestion to the application of appropriate analytical methods, up to the visualization and prioritization of results. Artificial Intelligence (AI) provides suitable methods to achieve the required level of automation to make truly adaptive reporting possible. For example, the Inspirient Automated Analytics Engine is able to autonomously slice & dice any given structured dataset; perform time series analyses and create forecasts; conduct regression analysis, clustering and outlier detection; apply natural language and network analysis; and describes the results as free form natural language texts or prioritized management presentation slides.

These conceptual and technical advances allow a glimpse into the future of reporting: With fully automated analytics and adaptive reports, the analytical bottleneck that currently keeps companies from realizing the full value of their data will be a thing of the past. Instead, the ever-increasing streams of business data accessible to a company will be processed on the fly, and a feedback loop will ensure that reports are adapted continuously to their target audience. Furthermore, with AI being able to take care of most day-to-day analytical and data science tasks, the current distinction between BI and data science will blur and eventually fade away.

The full article has been published in issue 6/2020 of REthinking: Finance (December 2020) available at rethinking-finance.com/archiv/ausgabe-6–2020-reporting-analytics (in German).

More information about the Inspirient Automated Analytics Engine is available at www.inspirient.com.