COMMUNICATING DATA

According to the Myth, Cassandra was astonishingly beautiful and blessed with the gift of foreseeing the future.

Her curse was that no one believed her…

Purpose

why is the message being communicated?

There are four purposes for communicating data science results:
to increase knowledge, to instruct, to facilitate informed decision-making, and to persuade. Which of these applies to the messages you are sending?

Strategy

what is your approach for gaining attention? 

Communicators have both active and passive strategies available to them. Active strategy are synchronous, i.e. they demand the audience's attention. E.g. a pitch provided to management. Passive strategies are asynchronous, they can reach their audiences at their convenience - if at all.  E.g. sending out an infographic in the internal newsletter and relying on information-seeking audiences to read it. The “push-pull” model combines both strategies by sending messages to audiences (push), while also making information and materials available to interested parties (pull).

Context

what factors influence receipt and intepretation

Contextual factors include other sources of information, personal experience, and competing priorities and are often outside the control of those sending messages. These factors can have influence at various points during the communication process and can even prevent effective communication.

Storylines

what outcome do you wish to communicate?

Once storylines are determined, messages must be developed. Messages – chunks of information that support the storyline – should be based on understanding built up through the application of data science methods. Each message should be able to stand alone by communicating a single idea, but, collectively, the messages should provide rationale for the larger theme (i.e., the storyline).

Sources

how will you incorporate various sources

Sources are differentiated based on the intimacy of contact, with interpersonal sources involving one-on-one interaction and mediated sources involving one-to-many interactions. Communication often involves a mix of both interpersonal and mediated sources, such as when data received from another (e.g., census data) becomes part of interpersonal communication (e.g., presentation to a senior manager).

Channels

by which means will your story be delivered?

 To have a better chance of reaching the intended audience, data scientists should consider the following factors:

• Availability, or whether audiences can access certain sources or channels (e.g., slide decks, dashboards, reports).
• Preference, or where and how audiences obtain information, which is closely related to availability.
• Credibility, or how believable a source is, based on perceived trustworthiness and expertise.

How to address the

three fundamental

audience expectations

why should they believe

rationale for conclusions

what to do now?

How will you use the tools at your disposal?

numbers

Probability

• fractional odds , e.g. '5-to-2' or '5/2' or '5:2',
• decimal odds, e.g. '2.5',
• percentages, e.g. '29%',
• set membership, e.g. 2 in 5

Suppose that 18 out of 20 patients (90 percent probability, odds of 9:1) in an experiment lost weight while using diet A, while 16 out of 20 (80 percent, odds of 4:1) lost weight using diet B. The relative risk of losing weight by choosing diet A over diet B is 1.125, while the odds ratio is about 2.25.

Absolute / Relative

• indexes, e.g. from 100 to 118
• units, e.g. from $112 to $189

Higher is Better

• lower is better, e.g. weight loss, debt, 
• ranking

Averages

• central tendency, e.g. mean, median, mode 
• spread, e.g. std, variance, quartiles
• interpretability, i.e. what to expect

words

Jargon

• technical, e.g. overfitting
• polysemous, e.g. training, testing
• difficult, e.g. conditional probability

Buzzwords

• bullshit bingo, e.g. cloud, smart data
• trade-off, e.g. familiarity, triteness 
• specificity, e.g. insight, 20% less users

figures