Database Reference
In-Depth Information
lyst working in a corporation. That they would communicate data differently shouldn't be
surprising, and may be entirely appropriate.
The important part is articulating your goal—actually writing out the answers to the three
questions just listed. If you're not certain about the answer to any one of these questions,
don't go any further until you're sure. (And it's OK if your sole purpose is to make someone
laugh. You don't have to be trying to achieve world peace with every data message.)
Principle #2: Use the Right Data
As the saying goes, sometimes less is more. One of the most impactful examples of commu-
nicating data that I've ever seen involved the presentation of a single number: 14. That was
the single data point shared with a group of managers assembled to discuss customer service
within an organization. The group of managers came to learn that this number represented
the number of times a particular customer had been transferred between departments during a
single call to a helpline. It motivated an entire organization to revamp the customer experien-
ce.
Sometimes less is really less, though. While driving in the car, I heard a report on the radio in
which a number of cities were compared based on the percentage of fish packages that were
mislabeled. Digging into the data myself later that day, I found that the sample sizes were too
small to infer much of anything about the relative mislabeling rates in the cities. A whole
host of listeners were misled by the story at least as much as by the fish labels.
And more is often less. It's possible, and actually quite typical, to overwhelm the audience
with data. It's easy to see why this happens: you worked hard to gather the data, and it feels
like that data increases the weight of your message and lends additional credibility. But all
that extra data only serves to drown out the message. Shannon and Weaver identified this
problem: “if you overcrowd the capacity of the audience, you force a general and inescapable
error and confusion.” In other words, if a data point doesn't add to your message, then it de-
tracts from it.
The last and most important point about selecting data is that your message must be both eth-
ical and based on sound epistemology. In other words: don't lie with statistics—we have
enough of that to contend with already. Don't fall prey to the many and various forms of stat-
istical and logical fallacies, such as mistaking correlation for causation, taking unreasonable
inductive leaps, applying the Gaussian when it doesn't apply, inferring more than the sample
size allows, and so on. These are just a few of the many icebergs to avoid (in this topic, I
hope to show you how to avoid some of them when you use Tableau).
