Effective data communication
Effective communication of findings into actionable insights is an essential skill. Without effective communication, smart data analyses may be interesting but not impactful. The importance of effective data communication is especially evident during the current covid-19 pandemic. The proliferation of statistics, dashboards, maps, and charts from the White House press room to news outlet, daily briefings in state capitals across the nation, and countless social media posts reinforce the need for effective data communication that accurately reports facts, highlights implications, and recommends action.
So how do we practice effective data communications? Here are three key principles that can guide us in communicating data to diverse audiences and stakeholders.
First, know your audience and engage them with what they know and can relate to. Too often, we assume that our audience is like us, with similar familiarity of statistical concepts and shared understanding of technical vocabulary. But that is rarely the case. Instead, we need to distill essential takeaways from our analyses and present it in terms that can be comprehended by our audience(s). That does not mean we do the most basic analysis possible, but that we explain our analyses and findings in laymen’s terms so that can be understood by audiences without our technical training.
The “flatten the curve” mantra and widely duplicated graphic of a gradual distribution and steep distribution is a good example of this. Without needing to explain the assumptions and statistics behind the epidemiological statistical models, audiences can understand the catastrophic potential outcomes that could happen justifying the implementation of social distancing and drastic economic shutdowns. It is easy to understand, can be easily shared as a hashtag on social media or highway signs, and provides a useful analogical platform by which further discussion and debate can occur.
Second, calibrate how much technical details to share. As part of knowing our audience, we need to calibrate how much we describe the statistical analyses and methodologies used to identify the insights we want to share. We should only include details necessary for our audience to trust and understand the empirical evidence we are discussing. Share too much technical details and a general public audience may miss the main takeaways and disengage; but share too little detail to a specialized audience and they may be skeptical of your analyses and be unconvinced by findings shared.
Take for example Andrew Cuomo’s presentation of competing prediction models on April 10. Speaking to his constituency of New York state residents (of which I am one), it would be a distraction to discuss the assumptions and statistical methodologies that underpin different predictions of covid-19 cases. The only details provided were descriptors like “severe”, “moderate”, or the team that produced the model. The lack of technical detail is appropriate since his purpose was not to assess each model’s predictive performance, but to demonstrate how the state had escaped the dire outcomes predicted by various projections.
By contrast, in a video conference that same week led by the University of Rochester Medical Center, more technical details were provided about statistical analyses used. For an audience of medical and public health officials, details about exponential and fractional polynomial projections are appropriate and help inform the discussion. To this audience, technical details are not a distraction but a signal of statistical rigor and evidence that results can be trusted.
Third, contextualize the data. It can be difficult for audiences to understand the range and implications of coefficients, predicted effects, and change over time. By highlighting real-world significance of otherwise abstract findings, we translate and emphasize implications of our analyses. To explain exponential growth, reframe findings in terms of the rate at which incidents double from an initial baseline of every 10 days to 6 days and so forth. Or to explain why even a small effect matters, extrapolate how much that incidence rate would cost in dollar terms; for example, predicting total potential loss revenue from even a small reduction in productivity aggregated and annualized across the company. Analogies can also help audiences understand the magnitude of effects by providing a numerical frame of reference; a five percent increase in content reach is equivalent to an increase of about eight million views every quarter, or about the population of New York City.
Angela Markel’s explanation of the reproduction rate is an example of effective contextualized interpretation. By translating how a seemingly small increase of reproduction and transmission from 1.0 to 1.2 in terms of how many additional people a small group of covid-19 positive people can infect, she distilled the findings of complicated epidemiological modeling into easy to grasp implications and justification for further vigilance.
Similarly, to contextualize the scale of suffering and impact of covid-19 on New York state, Andrew Cuomo described how the number of deaths from covid-19 has surpassed that of the 9/11 terrorist attacks on New York City. While the analogy is imperfect, it is an effective framing of the scale of devastation in relation to a shared reference point.
Effective data communications, in the best of times, can be challenging. Examples of good data communication during the covid-19 pandemic above reveals how essential effective communications are. While the stakes of what we do for clients and stakeholders may not be as high, we can nonetheless take lessons from how public officials are communicating implications of complicated epidemiological modeling in our work with teams and clients. By knowing your audience, calibrating technical details to share, and contextualizing results in our data communications, we will be more effective in communicating data findings and translating results into actionable insights.
This post originally posted here on my Linkedin.