2. Generic CPC
Assisted Conversions
Right idea, but flawed metric
Conversion £40Display Branded CPC Direct
Conversion £55Branded CPC
Generic CPC
Generic CPCGeneric CPC
...but 1 of 4 touches = 0.25
...but 3 of 4 touches = 0.75
@carmenmardiros
3. Generic CPC
Assisted Conversions
Right idea, but flawed metric
Conversion £40Display Branded CPC Direct
Conversion £55Branded CPC
Assisted Conversion = 1
Assisted Conversion = 1
Generic CPC
Generic CPCGeneric CPC
Not equally assisted
by Generic CPC
...but 1 of 4 touches = 0.25
...but 3 of 4 touches = 0.75
@carmenmardiros
4. All good (and bad) ideas start
with What If...
Frequency matters
more than position
First/Last click are essentially Recency-based
models. Why are they better than Frequency-
based ones?
High frequency is
less likely to be
random
Probability for a channel to appear 3 times on the
same conversion path significantly smaller than
appearing just once.
Frequency means
higher spend
Marketing optimisation problem if channels appear
frequently and don’t pull their weight.
@carmenmardiros
5. How to calculate
path coverage
Conversion £40Display Branded CPCGeneric CPC Direct
Conversion £55Branded CPC Generic CPCGeneric CPC Generic CPC{
Universe of conversion
paths with Generic CPC
=2
Universe of touches
=8 Generic CPC
global path coverage
4/8 = 0.5Generic CPC touches
=4
@carmenmardiros
7. Conversion £40Display Branded CPCGeneric CPC Direct
Conversion £55Branded CPC Generic CPCGeneric CPC Generic CPC{
Use it to calculate revenue
contribution?
0.25 * £40 = £10
0.75 * £55 = £41
Just another
attribution model...
@carmenmardiros
8. Clear dominant channel.
Primary heavy-lifter for customer segment 4?
Total paths with Generic CPC
for customer segment 4
Total touches
on these paths
path coverage
@carmenmardiros
9. Conversion £40Display Branded CPCGeneric CPC Direct
Conversion £55Branded CPC Generic CPCGeneric CPC Generic CPC{
Universe of conversion
paths with Generic CPC
=2
Universe of touches
becomes
=5
New Generic CPC
global path coverage
4/8 = 0.5
Generic CPC touches
=4
Filter “tipping point” touches
@carmenmardiros
10. Total non-direct, non-branded
touches on these paths
path coverage
(after direct / branded search excluded)
path coverage
(all channels included)
Total paths with
Generic CPC
@carmenmardiros
11. Convert path data so that
you can ask specific
questions
SQL and PowerPivot
@carmenmardiros
13. Run it through SQL
Assign
unique number
Split into
touches &
custom
channel
groupings
Calculate
path length
Calculate
position on path
Aggregate using DAX
and PowerPivot to work out contribution
@carmenmardiros
14. Not sold? How about asking
specific questions.
@carmenmardiros
15. ConversionDisplay Branded CPCOrganic Search Direct
ConversionBranded CPC DisplayOrganic Search Generic CPC
Two types of conversion
paths
Direct
Display
End of path is branded search or direct
End of path is upper-funnel channels
@carmenmardiros
16. Once here, is visitor nearly sold?
Is this what moved people to brand-preference stage?
Common patterns?
Differences between customer segments?
ConversionDisplay Branded CPCOrganic Search Direct Direct
End of path is branded search or direct
Last non-direct, non-branded
touches
@carmenmardiros
17. All paths
All paths where last non-direct,
non-branded touch was Email
Last non-direct, non-branded
touches
... then correlate with path coverage
@carmenmardiros
18. End of path nurturers
Closed the sale
Observed role: Moved people to conversion stage
(nurtured end of conversion path)
Common patterns “second last + last“ ?
Differences between customer segments?
ConversionBranded CPC DisplayOrganic Search Generic CPCDisplay
@carmenmardiros
19. All paths where last 2 touches were Email.
Which campaigns support each other?
Which campaigns don’t fulfill their function?
Patterns in “Second Last + Last”
touches?
@carmenmardiros
21. 40% global path
coverage
Total touches on paths with
Generic CPC
Generic CPC touches only
Missing touches
(multi-device, cookies...)
How many touches would
we have to be missing
for analysis outcome to be
different?
@carmenmardiros