This document discusses attribution and measurement strategy. It notes that attribution needs to align with incrementality and considers whether to focus on new customers or revenue, long term customer value, channels that drive transactions, and mobile investment. It also considers what to attribute, including touch points, users, winner takes all vs multitouch attribution, lookback windows, channels, and precision. The goal is to encourage longer term consideration, quickly measure performance marketing channels, and encourage cross-channel collaboration. Examples provided focus on these goals through various attribution methods.
5. Purchase
on desktop
Day 0 Day 10
Impression on
Instagram
on mobile app
Click on
marketing email
on desktop
Click on Google
PLA on desktop
Click on
Facebook
retargeting ad
on mobile app
9. Purchase
on desktop
Day 0 Day 10
Impression on
Instagram
on mobile app
Click on
marketing email
on desktop
Click on Google
PLA on desktop
Click on
Facebook
retargeting ad
on mobile app
Last touch 100%
Uniform fractional 20% 20% 20% 20% 20%
U shaped 35% 10% 10% 10% 35%
Decayed 30% 25% 20% 15% 10%
First touch 100%
Illustrative
13. I don’t know if I
should be driving
new customers or
revenue metrics.
We need to focus on the
long term customer value.
Some channels acquire
better long term customers.
Consideration is our
biggest problem as a
company.
We need to be able to
quickly allocate dollars
to channels that directly
drive transactions.
Attribution needs
to align with
incrementality.
Why is it so
complicated?
Mobile is so
important, but we
don’t invest
enough there.
No one
understands
attribution.
We should have
one model for
the whole
company.
Attribution drives the wrong
behaviors - if you retarget
existing customers, you get
the most credit, but that’s not
good for the company.
There’s no incentive for
channels to work together -
we’re all just doing our own
thing, practically competing
against each other.
They should only get
credit if the ad was
for the precise item
that was purchase.
14. ❏ Who is a user? (cross device? cookies?)
❏ Winner takes all or multitouch approach?
❏ If multitouch, relative share methodology?
❏ Look back window duration?
❏ Channels to include? (emails? organic?)
❏ Touches to include? (impressions? opens?)
❏ Precision between marketing and purchase item?
❏ Variation by channel/user segment? (look back windows? weights?)
❏ Multipliers to adjust for incrementality?
❏ Include customer lifetime value?
❏ Others?
So you got back some successful results from an A/B test on your growth hacking initiative
Time to celebrate!
You need a way to continuously monitor without the expense and effort of an A/B test which also recognizes the robust ecosystem of broader initiatives and marketing within your company
Realistically, your initiative exists amongst many others in your company. How do you continue to measure and optimize? This is an example of a touch stream where a user sees multiple ads across different websites and devices, seeing impressions on some and clicking through on others, and ultimately purchases 10 days later. If your initiative was one of these ads, how do you know that it continues to work after the A/B test is done and it’s just rolled into the ecosystem of multiple initiatives at your company?
Attribution is defined as… basically, it’s the way of giving credit (or assigning value) to certain touchpoints (clicks, impressions, etc.) involved in a transaction or signup etc (desired outcome)
Attribution is defined as… basically, it’s the way of giving credit (or assigning value) to certain touchpoints (clicks, impressions, etc.) involved in a transaction or signup etc (desired outcome)
There are many different methodologies for attribution - how do you decide with so many choices? Spend time. Define what some of these are
As some examples of possible credit… and this is before any more advanced modeling methods but just simple rule based options
So how do we get to attribution? And what is the relationship with strategy? In order to get to the right attribution for your aims as a company, you need strategy to drive attribution.
So how do we get to attribution? And what is the relationship with strategy? In order to get to the right attribution for your aims as a company, you need strategy to drive attribution.
So how do we get to attribution? And what is the relationship with strategy? In order to get to the right attribution for your aims as a company, you need strategy to drive attribution.
There will be many different perspectives - what focus are you trying to provide with attribution? Wall of words… but that’s what happens when you start conversations about how people feel about attribution and what should be done. Lots of different perspectives and opinions on what is needed. And there are obvious biases - a display team will care about including impressions, a paid search team will care about last clicks, a finance team will want it to be accurate and reflect incrementality or cause and effect. (more detail on what types of decisions exist)
Need to decide major strategic elements to figure out direction of MTA. While surely not comprehensive, these are the pieces that I’ve found we’ve had to think about internally when making the strategic decision of how to build attribution. Some of these will be easy and obvious, some will be controversial and difficult to get to a decision. And some will be easy technically, some may require investment to get to what you want (for example, cross device).As you think about these, need to consider operational as well. A long look back window will result in delay of an answer. A complex model will take longer to process and not be able to be real time.
So how do you do it? Talk to all involved, understand the strategy you want to guide dollars to and how to use attribution
Then create a system and figure out how to work within it (multipliers, add LTV, different efficiency goals in different channels, etc)
Attribution is tough - if everyone is happy, you probably did it wrong. Your goal is a system that meets your strategic aims that people can mostly live with and work within.