This document discusses how an agency built a bespoke marketing attribution solution for a client. It provides an overview of attribution modeling and different approaches like last touch, first touch, and multi-touch attribution. It then describes how the agency cleaned and processed the client's raw marketing data to perform attribution analysis. This included flagging first touch points, conversions, and assigning attribution windows. The analysis evaluated conversion rates and the roles of different marketing channels in assisting and driving conversions. It concluded by discussing challenges, next steps, and how the attribution solution provides insights on campaign performance.
6. The Attribution Puzzle
Months (900+ Interactions)
Social Display PPC Email
Test
Drive
https://www.thinkwithgoogle.com/articles/consumer-car-buying-process-reveals-auto-marketing-opportunities.html
7.
8. What is attribution?
• Moving beyond giving all credit in the conversion to the final marketing
channel.
• An attribution model is the set of rules that determines how credit for
conversions is assigned to touchpoints in conversion paths.
• Are we undervaluing any channel that initiates conversion paths?
• What about the channels that nurture our potential leads?
• How do different channels play different roles in the conversion funnel?
Social Display PPC Email
Test
Drive
9. What are we hoping to achieve?
• How many conversions were multi-touch?
• Where do marketing channels fit in the user journey?
• Each channel contributes in different ways to the conversion funnel.
• Varied marketing channel portfolio is necessary for multi-touch user
journeys.
• Look for things to investigate further.
10. The Fallacy of Last Touch
• This is the standard approach
to attribution.
• 100% credit to the converting
channel.
• Useful for businesses with a
sales cycle that does not
involve a consideration phase.
Social Display PPC Email Convert
0 0 0 1 1
100%
11. First Touch
• 100% credit to the first
channel.
• Useful when comparing with
Last-touch or other models.
Social Display PPC Email Convert
1 0 0 0 1
100%
12. Assisted Conversions
• Credit evenly distributed to
every channel EXCEPT the
last touch.
• Useful when comparing with
Last-touch.
Social Display PPC Email Convert
1/3 1/3 1/3 0 1
25% 33% 33% 33%
13. Linear (Even)
• 100% credit is distributed
evenly to all touchpoints.
• If you consider each touchpoint
equally important during the
consideration process.
Social Display PPC Email Convert
1/4 1/4 1/4 1/4 1
25% 25% 25% 25%
14. First and Last
• 40% credit is assigned to each
the first and last touchpoint,
and the remaining 20% credit
is distributed evenly to the
middle touchpoints.
• If you most value the first and
the final touchpoints.
Social Display PPC Email Convert
4/10 1/10 1/10 4/10 1
10% 10%
40%40%
15. Spatial
• The touch point closest (in
terms of position) to conversion
gets most of the credit, and
touch points prior to that will
get less credit.
Social Display PPC Email Convert
1/10 2/10 3/10 4/10 1
1st 2nd 3rd 4th
40%
30%
20%
10%
16. Temporal
• The touch point closest (in
terms of time) to conversion
gets most of the credit, and
touch points to that will get less
credit.
Social Display PPC Email Convert
1/40 3/40 16/40 20/40 1
Day 1 Day 3 Day 16 Day 20
40%
30%
20%
10%
20. Getting the Raw Data Feed
Because of file size you’ll probably
need to get it delivered to an FTP
You can ask for the full
data feed, this file is
delivered hourly and
contains all data, this file is
Huge, you can get this
delivered to D3 on the
amazon cloud, which is
nice
22. Process this Data File in Python (4 Steps)
(Did this whole thing in 140 lines of code)
Step 1: Clean file (remove all page views where page views equals zero),
flag fist touch point in visit, count page views in visits, create sort key.
Step 2: Group by tracking ID, and sort by time (need to sort by the sort key),
flag conversion event (Only one conversion Event per Visitor)
Step 3: Read in file backwards, create attribution window, count touch points
from conversion, write conversion time to the same row as the conversion
touchpoint.
Step 4: Re-order file and step three reversed the process.
23. Multi-Touch Attribution
• .
• In this scenario, it makes sense to distribute credit for the conversion beyond
the last touch.
24. Analysis – Conversion Rate
• A Very Small Amount of visits resulted in a conversion.
29. Next Steps
• Combine with marketing spend to
calculate cost per conversion.
• Do further analysis on users who
didn’t convert.
• Use the data to learn about how the
channels perform in different
campaigns, with different objectives
and target audiences.
30. Conclusions
• Attribution modelling is about assigning credit to channels involved in a
user’s journey to conversion.
• Last Touch is useful but flawed when viewed in isolation.
• SEO and Social are important in assisting conversions.
• PPC and Direct Traffic are important as last touches.
• Use a suite of tools and models to analyse the complexity of the user
journey.
• Provide analysis of specific campaigns to better understand the role of
various channels.