You are on page 1of 55

Data Visualization

A primer on best practices

Lazaro Gamio | @LazaroGamio

What were going to go over


1. Evaluating data
2. Finding stories in your data
3. Finding a shape for your story

Some context
Where Im coming from, and the type of stuff
we do at @PostGraphics.

Feel free to interrupt.

Evaluating data
Data needs to be found and vetted
the same way sources are.

Theres so much of it
It doesnt make the process any easier.

Gasp.
Thats also just one of many places where
data lives. Heres a random example:

I have no idea what this is.


And neither will my audience.

Looks cool

It was a 63 mb spreadsheet
384,499 rows of goodness.

Lots of data Easy data


Theres just more to edit and sift through.

You need to spend time


with your data.
Take it out to dinner and a movie.
Ask about its hopes, dreams and aspirations.

Who is your source?


Data is more than the final product.
1. Who collected the data?
2. Who inputted the data?
3. Who formatted the data?
4. How is the data being delivered?
5. Whos giving it to you?
6. Is your data recent?

Population of Azerbaijan
CIA World
Factbook

Wikipedia

World Bank

Population Reference Bureau

How many people?


Azerbaijans population is 9.5ish million.

Takeaways
1. Data is often very specific.
2. Data is really big.
3. Your editing process begins at selecting a
dataset.
4. You need to put in a lot of work to process
data into something digestible.

What does your data say?


What do you want to learn? Most
importantly, what do you want your readers
to learn?

Questions to ask
1.
2.
3.
4.
5.
6.

Whos the mostest?


Whos the leastest?
What is the middle?
Who are the outliers? Why?
Is there a trend over time?
Does this trend correlate with another
trend?

Investigate outliers
1. It may be a good story.
2. Your data may be flawed.
3. It may mean nothing.

California is
just big, not an
outlier.

Mean vs. Median


Sometimes large values distort your mean
and can skew your perspective of the data.
Dont add to the confusion!

Mean: $513,000
Median: $350,000

Beware empty data


N/A does is not Zero. Investigate
missing data and tread lightly.

Trends
Do you know why there
is a variation? You should.

Correlation Causation

Takeaways
1.
2.
3.
4.

Ask answerable questions of your data


Learn the limitations of your dataset
Once you have findings, double-check.
Then, Triple-check.

Give your story a shape


You found some interesting facts.
Now chart them.

Before you start


1.
2.
3.
4.
5.
6.

Do you really need a chart?


Are you charting discrete quantities?
Are you charting a time series?
What do you want to draw attention to?
Is your data geographical?
If so, does location make a difference?

Content is king
The most important thing is knowing exactly
what you want to tell. The previous two steps
are the most important.

Charting the mostest


Are you trying to show one important value
with a couple of other values for context?

The Mostest.

Chronological data
Do you have data that changes over time?

A simple trend

Resist the urge to map


While maps are often useful, they are often
not the best tool for your story.

People married to
someone they met
in high school

Noise levels

Gun deaths

Dont make
population maps
Do it for the children.

Two Charts
Have two datasets that make sense together
and help tell the same story?

Two trends.

Takeaways
1. Keep it simple. Your readers will
appreciate it.
2. Pick the chart that best proves your point.
3. Tell the reader what they should be looking
at.
4. If appropriate, tell the story in multiple
charts.

Tools
Excel, Google Docs, Tableau, Datawrapper, R,
Matplotlib, D3, Adobe Illustrator, etc.

Google Docs/Excel

Great options

D3.js

Questions?

Thanks!

Lazaro Gamio | @LazaroGamio

You might also like