Total 625 words used.
In daily operations, managers get bombarded with information which is usually provided for an audience of many rather than an audience of one.
DHL Express senior management team got this treatment with additional complexity by getting key insights in various separated reporting streams in Operations, Finance, and Sales. Let alone that those reports we generated manually overnight and error prone.
In addition, said data sets the way ‘traditional BI’ requires decision makers to
- calculate numbers and interpret visualizations like diagrams and graphs,
- have expert knowledge about what to look for, and
- triangulate static or dated data sets
to design, align, and agree a subsequent course of action which is not self-explanatory.
We tried to overcome these fundamental flaws of ‘traditional BI’ by leveraging AI enabled data-driven storytelling to turn information consumption into reading a newspaper which points you to exceptions in need of attention.
The ultimate vision is that each decision maker in the 220 countries and terretories can draw from a repository of all performance indicators at DHL Express and drag and drop the ones key for his or her role into a newspaper for an audience of one.
This is possible by using natural language generation AI to providing quick and easy natural language information on data in a newspaper look and feel with AI generated near real-time.
These data stories provide objective insights grounded in data that everyone can understand, leaving no room for interpretation.
Each narrative was personalized for each role’s information needs that a drill down in underlying data becomes an exception.
Our approach had 5 distinct phases:
- Explore key performance indicator usage. In this phase we answered the following questions for each senior management team member:
- Which key performance indicators are important for you?
- Why is confirmed a key performance indicator for a role?
- How is it used and why?
- What does it mean for a stakeholder if a key performance indicator goes up/down?
- What is a good narrative if a key performance indicator goes up/down?
- Build Senior Management Team personas to design the overall dashboard
- Confirm KPI key performance indicator calculation
It is key to ensure that the AI uses correct data inputs to explore the facts, assess them, and generate the narrative summarizing the facts. Without conducting this step with rigor, decisions are not sufficiently grounded in data, which is a must.
- Design narratives
- Implement narratives
Blend narratives into the top-level newspaper look and feel as well as blending them into complementing drill downs.
The output of this approach was a dynamic dashboard in which all data stories are AI generated in real-time. In case one of the deviations contains an outlier in need of management attention, a manager can now select a key performance indication and read why it is an outlier.
Based on how personas approach a key performance indicator and its underlying specifics it is presented in preferred diagram styles and narratives. Each specific allows for a second drill down layer explaining each specific’s underlying data.
Any additional issue related drill downs are then handed over to data analysts for a full deep dive.
Data-driven storytelling was confirmed to add value by
- freeing data scientists from report compilation duties who can now focus on data drill downs for problem solving purposes,
- freeing decision makers from
- data interpretations and triangulations across data sources,
- screening tons of data which are not in need of attention at a given point in time
who can now focus on creative problem solving to serve customers,
- turning numbers into easy-to-read and easy-to-understand stories that help us turn insights into action and can be easily forwarded to whoever needs to address an issue operatively,
- focusing on exception handling,
- making complex interdependencies easier to understand, and
- enabling cross-functional, collaborative senior management team-level decision-making.
Nomination Title: Overcoming the flaws of traditional Business Intelligence by leveraging AI enabled data-driven storytelling
