12. The app was used primarily by
Boston techies, who have a lot
more money than the majority
of the population
13. And so the data suggested
that the potholes in the rich
parts of the city were causing
the most problems
14. GARY KING, HARVARD UNIVERSITY
“This was an example of really bad analytics
and it’s even worse because it’s the kind of
thing that feels like it should work and does
work a little bit.”
15. What they think they’re solving for
The most problematic potholes
30. When you use last-touch
attribution, this is how you
end up valuing the different
parts of your website…
31. Product Support Blog Pricing Contact
Pretty
good!
Meh?
Rubbish
Great!
Amazing!!!
32. MATTHEW ZOOK, UNIVERSITY OF KENTUCKY
“Once something gets measured and put into
a model then everything starts optimising
around that metric.”
33.
34. No-one in their right minds would read this
blog post and then buy a car insurance
policy, but when you value your content by
how it contributes to conversions… then you
end up trying to make that happen.
35. A slightly better way to
measure value is a linear or
time decay model
47. Even with linear or even first
touch attribution, this looks like
incredibly poor return
48.
49. But… maybe we can measure
the value of this TV spot instead
as if it were a bit of TV coverage
50.
51. The way TV was historically
measured was via impressions
and how it contributed to brand
salience (as determined by
recall surveys)
52. This was an ok way to measure
TV 30 years ago, when there
wasn’t many other marketing or
advertising channels available
53. But today… there’s just too
much noise to make such data
meaningful. Brands who still
measure in this way
dramatically over value TV
54. AN ALL TOO COMMON RESPONSE TO THIS PROBLEM
“Lets temporarily stop all other brand activities so
we can properly measure PR…”
55. But we know that PR, or advertising,
or other brand building activity never
works in isolation. Stopping
everything necessarily reduces the
efficacy of the activity you’re trying
to measure.
97. What you should be doing
Creating content on channels appropriate for
the goal and audience
98. Goal
Audience
Channel
Content
Distribution
Metrics
KPI
What are you trying to achieve?
Who are you trying to reach?
Where can you reach them?
How can you meaningfully engage on this channel?
How can you get your content to the right people?
What can you accurately measure?
What’s a good proxy to measure the goal?
Strategy Questionnaire
99. Here’s an example of how we
recently applied this plan to
Facebook video
110. This lead to a 30% increase in
overall page likes over the duration
of the campaign, at a cost of about
$2 per like.
111. • Bad Data < No Data
• Measure everything, but always optimise for just one KPI
• Only use ROI for direct response & conversion activities
• Ensure alignment of content with channel and audience
• Tie everything back to the goal
Takeaways