How a Picture Postcard Can Help You Develop a Data-Driven Analytics Culture
Are your meetings fruitless? Thinking about a picture postcard from a friend's vacation can be the model you need to move your enterprise to a more data-driven culture.
- By Mikael Berndtsson
- October 21, 2019
Postcards you receive from vacationing friends typically have a nice picture on the front, maybe a hand-drawn arrow and some text (such as "our beach/hotel"). Along with your address, the back of the postcard may feature a message that describes what your friends have experienced at the location. You'll look at the picture and the message and decide whether you would like to visit the same place.
Receiving a postcard with just a nice picture on the front but no personal message simply does not give you the same thrill.
Now consider your typical business meeting. Someone presents a nice graph that describes a trend for a given metric moving in a positive direction. Everyone at the meeting agrees that it is nice to see it trending in the right direction. You move on to the next agenda item. A data scientist shows a graph of the results of text mining customer reviews. You "get" the big picture of the visualization; you can interpret the different sizes of 3-D bubbles in pretty colors, but it is difficult to see a clear pattern or trend. Once again, you nod and say "interesting," and you move on to the next discussion item.
Sound familiar? This is not a data-driven culture, and it will not generate any business insights. Changing how participants use data and graphs at a meeting to be more data-driven is one way to implement a data-driven culture.
To get there, think of that picture postcard. At your next meeting, use the elements of a postcard described below to coach yourself and your colleagues toward more data-driven insights.
The Picture
When you're presented with a plain postcard -- a picture that shows something that is either too simplistic or too complex to understand -- speak up.
When you see a simple graph showing only overall figures, there are several questions you can ask to help drive the discussion in the right direction. First, ask for more details and drill down into the data. For example, are the positive numbers true for all product categories or are there product categories that don't have the same positive trend?
Ask for a definition of what is actually shown in the graph. If the graph is showing figures for customers, ask how customer is defined. Your organization probably has different types of customers. Are all these types of customers included in the graph?
Perhaps the picture on the postcard is a snapshot that was valid when the picture was taken ten years ago, but the small town featured in the picture is now completely replaced by bigger houses. When you see a graph showing a simple snapshot of business data (e.g., we currently have 125 registered VIP customers), ask for details. When was the data collected? What has happened to the data since the graph was generated? Can we access the current numbers right now at the meeting?
Furthermore, ask how the snapshot picture is related to target values, categories of values, or how it compares to the same period last year or month. Isolated snapshots of business data will not show if the metric is moving in the right direction. Similarly, when the graph contains too many low-level details, ask the speaker to explain the overall trend(s).
A senior data scientist told us "I have roughly five minutes to explain the findings from the predictive analytics. If the business person does not understand the explanation and findings within that time frame, the findings and associated graphs are simply ignored." If you can't follow the message of the visualization or animation, ask for an explanation or another type of visualization/animation.
The Personal Message
Does the presentation have a personal message just like that engaging postcard does? If not, then don't be afraid to ask for an explanation: what is the business problem we're trying to solve or the question we're trying to answer?
It is tempting for a presenter to display an impressive graph or table based on complex calculations and a huge amount of data without any personal message to the audience. In his book Web Analytics 2.0, Avinash Kaushik refers to this as "puke out data." On more than one occasion, we have seen a presenter withhold his/her own interpretation of what the graph is showing. Ask the speaker to interpret the data and offer potential options or recommend actions to take. Oftentimes the data analyst knows more about overall patterns and trends in customer data than business people working more closely with customers.
Ask yourself, why am I seeing this presentation? Is the presentation related to a key performance indicator, a business problem, or a hypothesis, or is it just an arbitrary metric? If it is just an arbitrary metric, ask why the graph was developed.
The Address
See that address on the postcard? Are you and your colleagues addressing your presentation to the right audience?
A side effect of introducing self-service analytics (and a more data-driven culture) is that the decision power moves closer to the front-end business users. Thus, if you are a presenter, reconsider the default approach to first make a presentation to the management group. Instead, consider presenting it to the people actually responsible for the product or service. They will give you more details and valuable feedback because they are closer to the data. Then you can make a second presentation to the management group.
Similarly, if you are listening to a presentation, ask about feedback from the business people closer to the data or if future meetings have been planned with them.
The Sender and the Stamp
Is the postcard from a friend? Did your friend send the postcard without a stamp, letting you pay for the shipment? Do you trust the presenter and analytics used?
Nowadays, many organizations have adopted self-service business intelligence, which allows more users to develop and perform their own data analysis. To enforce quality assurance, the entire chain -- from raw data to finished graph -- should come with a stamp of quality certification. Reports and graphs displaying official organizational figures have the highest level of quality certification and can be trusted. When there is no quality certification present, ask the presenter how the graph was developed, what data was used, and what type of analytics tools were used.
The Decision
Just as the combination of the postcard's picture and personal message helps you decide whether you would like to visit the same place, the information in a presentation should be adequate to drive business decisions.
At the end of a presentation, you will be in one of three situations.
- You can draw insights and take action. Congratulations. This is the desired outcome. Your organization is moving toward a data-driven culture.
- You cannot draw any insights. A lack of insights is a sign that the analytics may be of no (future) interest or that the presentation format is not suitable. Ask the presenter to return with answers to your questions and revised analysis.
- You can recognize the insights but do not take action. If the insight drawn is "we are moving in the right direction," consider launching a small experiment with a new approach that could potentially lead to even better results. In his book, Turn the Ship Around!, David Marquet discusses different types of questions leaders can ask. Following that ladder of questions, ask meeting participants what insights can be drawn and what they would like to do or intend to do. Merely asking "What do you think about this?" can result in a random set of opinions, with few suggestions for how to implement the insights.
A Final Word
The perspective we've suggested in this article is not rocket science, but these types of questions and considerations are not used as frequently as they should be. To move your organization to a data-driven culture, start by taking notes at your next business meeting and use the elements of the postcard as a template. Consider what types of questions are already being raised, then focus on asking the right ones.