1. The need for people-based measurement
A white-paper about why mighty marketers
need to speak to people, not cookies
2. ExecutiveSummary
2
The largest study of its kind globally
which shows how crucial accurate
people-based measurement and
attribution really are! In this growing
multi-device world, it is now more
important than ever to move on from
cookie-based tracking methods
which risk undervaluing almost 40%
of touchpoints.
Christian Bartens, CEO, Datalicious
insights@datalicious.com datalicious.com/optimahub
3. This study found that each person has on average 3 unique cookie identifiers. In effect, a
person working on a home laptop, work laptop, and a mobile would be seen as 3 unique
people through the lens of cookie-only measurement. The speed of mobile adoption is only
increasing the gaps in our measurement.
Despite this rapid shift, without visibility of the impact of mobile, advertisers have been
reluctant to invest significantly in mobile, often relying on their intuition to justify investments.
By taking the view of a person across devices, advertisers gain a better view of the various
touch points in a consumer’s journey towards a sale.
Broken lens - Cookies causing fragmentation
ExecutiveSummary
3
A mobile-first world requiring better data
The shift to mobile should not be underestimated. In total, there are 31.15 Million mobile
subscriptions in Australia, greater than the entire population of the country. While desktop
share of web traffic is down 19% YOY, there is a 36% YOY growth of share of web traffic on
mobile phones, with an average time spent on the internet at 1h 36m daily on mobile devices.
The majority of current tracking methods cannot accurately reconcile all these touch points to
the person heading towards the sale. Until recently the cookie has been the best means of
tracking someone through their digital journey, but with each device and browser requiring a
different cookie, there are a lot of gaps in this type of tracking. In fact, we found that 38% of all
paid digital touch points were missing when using only cookies to build the path to purchase.
Gone are the days of mass broadcast
Deciding where to allocate advertising investment has only become more complex over time.
Today’s consumers are scattered and influenced by many different channels, making
investment decisions even harder for marketers.
The unique power of digital in connecting advertisers with consumers has meant that a
consumer can interact with a brand at any given time. As such, consumers move comfortably
between devices, like researching on a laptop or scrolling through advertising on their mobile,
before finally making their purchase on whatever device they see fit for the occasion.
insights@datalicious.com datalicious.com/optimahub
4. ExecutiveSummary
4
People are the key to better media attribution
A people-based approach to measurement combined with advanced marketing attribution
models from companies like Datalicious provide marketers with an accurate view of
cross-device performance and audiences. The model has enriched customer journey data with
more digital touch points and reveals the true value of mobile advertising channels.
In addition, this research looks to understand the impact of moving from a traditional
cookie-only-based measurement methodology, to a people-based measurement methodology
using custom attribution. The research was carried out by Datalicious using Facebook's Atlas
data for 3 of Australia's most recognised brands.
The figures presented in this study are based on >1 billion media touch points across >80
million purchase paths and >1 million conversions across the three brands. The study seeks to
understand the true opportunity for marketers to drive efficiencies and grow their businesses
based on customer’s behaviour across digital channels and devices.
The inaccuracy of mobile measurement was one of the
biggest challenges we faced using cookies, which
prompted us to look at alternative approaches. The most
valuable insights from this program was the 2x increase in
credit given to mobile advertising, which was previously
masked due to last click, cookie attribution. It was really
important to understand the value of our digital media
plans against our wider business goals, and to confidently
increase and investment in mobile moving forward.
Kate Marshall, Digital Media Manager, Foxtel
insights@datalicious.com datalicious.com/optimahub
5. There is a need for better data in a mobile
first world to make better decisions
6. Insight#1
6
It starts with a data problem. An attribution model can only be as good as the data going in. To
build the most effective model, we must address the data problem.
Sound recommendations on where media investments should be placed require the impact
that particular channel or tactic is having to be quantified. With every new device and/or
browser creating a different proxy view of a person, the potential to draw the wrong conclusion
on its impact is a major possibility. Whilst this is not new news to digital marketers, this research
has given us the opportunity to quantify just how many different proxy identifiers we see for
each real person.
We found that for every 1 person, there was on average 3 different cookies. If one of these
cookie paths leads to a conversion, the other two would not get any credit for its influence on
the event. Models and decisions should be made on as near to the complete path of paid
touchpoints that a consumer sees as possible. To do that, we need to stitch these separate
views together using a person as the common identifier.
Insight #1 - 38% of all touch points lost
1 Person
3 Cookies
Each participant had on average 3 distinct cookies on their path to conversion,
indicating 3 different devices or browser combinations
= Unique Device or Browser
Each cookie will have a different amount of paid touchpoints associated with it, depending on
the person and device, so it’s only when this data is reconciled to a person that we can start to
explore the volume of touchpoints actually missing. The research uncovered that on average
there were 22 paid touchpoints prior to a conversion when looking at people based
measurement. Cookie-only measurement missed 38% of unique touchpoints per person.
These would previously have been classified as non converting and down weighted in any
attribution model.
insights@datalicious.com datalicious.com/optimahub
7. Insight#1
7
The key driver is the disparity is mobile, greater than two thirds (72.5%) of all the missing
touchpoints happened involved a mobile device.
more digital touchpoints observed
per unique study participant
On average
38%
More touchpoints per unique person
16Average number of unique
touchpoints per participant
Average number of unique
touchpoints per participant
22
The data problem is misleading marketers and creating a misguided view of the true impact of
their digital advertising. A more complete view of a path a consumer takes which includes
mobile is key to optimising digital investment decisions.
Close to 40% of all touchpoints in journey to conversion lost through cookie-only tracking;
Mobile-only touchpoints constituted almost 29% of all missed touchpoints.
Cookie
World
VS
People
World
Cookie measurement can cause inaccuracies
Cookie-Based
Measurement
People-Based
Measurement
insights@datalicious.com datalicious.com/optimahub
8. Insight#2
8
With the data problem of cross-device measurement fixed and a people based view of the
media exposure, we can now start to better understand the true impact that mobile delivers
on conversions.
When we combined the data from the 3 advertisers in our study, we saw that 45% of all
customer journeys involved had exposure to advertisements that were across two or more
devices. In a cookie-only measurement approach, the cross-device impact of mobile would be
zero, due to the inability of using cookies to track journeys between devices.
When people-based measurement was integrated, we saw the proportion of conversions
attributable to cross-device impact greatly increase to 45%. Combined with the 7% of
conversions attributable to taking a mobile-only path, it means over half of all conversions
(52%) were influenced by mobile advertising.
In a cookie-only-based approach, we saw only 29% of conversions are attributed to
mobile-only journeys. When people-based measurement was accounted for, the combined
proportional impact on conversions attributable to cross-device (desktop + mobile) and
mobile-only was 52% attributed to mobile devices in the people based measurement. That's
on average nearly a 1.8x net increase in the overall share of conversion credit given to mobile.
Insight #2 - Mobile advertising now drives 52% of
all conversions
7%
29%
0%
45%
Cross-device journeys quantified
71%
48%
Not only does mobile play a significant role in purchases conversions, but it offers a very cost
effective channel to advertise. The combined proportion of budget investment in mobile from
among all our study participants accounted for 27% of the total advertising budget, but
the return on average was over 1.9x that investment, with mobile contributing an average
52% of all conversions.
Cookie-Based
Measurement
People-Based
Measurement
insights@datalicious.com datalicious.com/optimahub
9. Cookie-based measurement caused too much conflicting
data in our reports. With all touch points in a single view of
the customer, we can focus our marketing efforts and
invest in mobile.
Rene Rached, Head of Acquisition Marketing, Medibank
Insight#2
9
But it’s not just the presence of mobile in the journey and its cost effectiveness that’s of interest
to marketers. Mobile advertising can play different roles in driving people through the purchase
funnel. Using the people based purchase path and applying the Datalicious proprietary HMM
(Hidden Markov Modelling) machine learning attribution approach allows us to not only see
what touch points help drive the final conversion, but also how effective certain touch points
are at moving people along the conversion funnel.
The model revealed that mobile advertising was a dominant influence in shifting consumers
from passive to aware and aware to intent. Desktop was proven to play a crucial role in the
final stages when consumers really do need a bigger screen to go through an in-depth
conversion process. The same modelling approach using cookie-only data showed desktop as
the only dominant influence in each stage of the funnel.
48%
73%
52%
27%
MobileDesktop
% spend % credit
insights@datalicious.com datalicious.com/optimahub
10. Insight#2
10
Passive / Disinterest
Desktop Mobile
Desktop
Desktop Desktop
Desktop Desktop
Mobile
Awareness
Intention
Desire
Action
The optimal device to drive prospects to convert
For these 3 advertisers in our study, the findings showed the following insights on
cross-device behaviour -
• Mobile is the best way to start a user's journey and move them to awareness
• Mobile is the best way to turn passive prospects into active ones with intent
• Desktop is the best way to push prospects towards the final purchase stage
• Desktop is the best way to close the customer journey
Whilst this is an aggregate finding, each advertiser saw mobile playing different roles in the
purchase funnel, demonstrating the need for every brand to identify and measure their unique
mobile strategy.
Cookie-Based
Measurement
People-Based
Measurement
insights@datalicious.com datalicious.com/optimahub
11. Insight#3
11
With better data powering our models we can more accurately understand the impact of paid
media and assign the appropriate amount of credit to each device, tactic and channel that
played a role.
When it came to which channels were driving the conversions, the people based attribution
model gave a very different perspective than cookie-only. To help quantify the differences we
compared this to a traditional cookie-only last click model and also a cookie-only custom
attribution. The results are displayed below.
Insight #3 - Cookie-Based Last-Click Attribution greatly
undervalues paid media
39%
3%
49%
28%
7%
47%
18%
25%
17%
42%
16%
-11% -3%
+4% +10%
+38%
-5%
-31% -2%
Credit is being misallocated due to inaccurate measurement
People-based attribution improves upon existing cookie-only-based approach and refines
the mix; Facebook + Instagram (and other paid display channels) get more deserved credit
for their upper and mid funnel activities.
Paid Search Facebook, Instagram Display
Other Display Direct, Organic Search, Referrals, etc
9%
Cookie-Based
Last-Click Attribution
Cookie-Based
Custom Attribution
People-Based
Custom Attribution
insights@datalicious.com datalicious.com/optimahub
12. Insight#3
12
The left-hand column of cookie-only-based measurement with last-click attribution is arguably
the most inaccurate approach. The middle column of cookie-only-based measurement with a
custom attribution model improves the accuracy of the results but cannot solve the
cross-device challenge. The result using this approach is that some channels are still
undervalued that tend to be more popular on mobile devices. Only the people-based
measurement model gives a more accurate view of what channels contribute to driving
conversions in today’s cross-device world.
With People-Based Custom Attribution, display channels, including Facebook, show a
significant increase in the credit assigned compared to a cookie-only model. This is due to
both the inclusion of the mobile touchpoint and the custom attribution model recognising the
impact of upper funnel advertising.
Paid search is still a dominant and influential component, but the biggest difference comes from
direct. By combining multiple cookie paths at a person level, what would have previously been
seen as direct or organic conversion can be linked to a previous paid touchpoint. With the
study, we found that direct only accounted for 16% of conversions vs 49% in a cookie last click.
When we further break the channels down by device around mobile. The people based
custom attribution revealed that all paid mobile channel credit was significantly higher
compared to last click cookie only measurement.
Cookie-Based
Last-Click Attribution
Cookie-Based
Custom Attribution
People-Based
Custom Attribution
37%
6%
33%
16%
18%
10%
6%
30%
17%
10%
8%
16%
7%
10%
16%
26%
9%
7%
People-based Attribution results in greater accuracy of channel effectiveness
Paid Search Desktop Facebook, Instagram Display Desktop
Other Display Desktop Direct, Organic Search, Referrals, etc Desktop
Paid Search Mobile Facebook, Instagram Display Mobile
Other Display Mobile Direct, Organic Search, Referrals, etc Mobile
3%
1%
2%
3%
9%
insights@datalicious.com datalicious.com/optimahub
13. Insight#3
13
Whilst custom cookie only attribution is able to give a good estimate of desktop contribution,
mobile display ('Facebook' and 'Other display') credit shifts closer to the truth with each model.
When the mobile touch points were linked to a person, we saw a realignment of the credit
between the channels, with Facebook display Mobile being more influential than previously
visible in cookie only models.
Returning back to our purchase funnel derived from the HMM model, we can overlay channel
to see which channels played which role in driving the conversion. This information is crucial to
selecting KPIs and also creative per channel.
For these 3 advertisers in our study, the findings showed the following insights on
cross-device behaviour -
• Display is the best way to start a user's journey and move them to awareness
• Facebook is the best way to turn passive prospects into active ones with intent
• Facebook is the best way to push prospects towards the final purchase stage
• Search is the best way to close the customer journey
Passive / Disinterest
Display Display
Display
Search Facebook
Direct Search
Facebook
Awareness
Intention
Desire
Action
The optimal channel to drive prospects to convert
Cookie-Based
Measurement
People-Based
Measurement
As our study results are an average across multiple clients, we encourage advertisers to
implement their own people-based measurement to understand their path to conversion.
insights@datalicious.com datalicious.com/optimahub
14. Insight#3
14
Education across the business on people based
measurement could help have a significant shift in digital
first thinking and a shift in effort towards bespoke digital
creative. By rolling this out on-going it could really help test
our media effectiveness and future planning efforts.
Kate Marshall, Digital Media Manager, Foxtel
insights@datalicious.com datalicious.com/optimahub
15. ImportantTakeaways
15
Credit should go to all deserving ads - With the study we saw the average
CPA drops by 21% with better data and an additional 38% more paid touch
points being included.
Mobile deserves to be in your models and reporting - The study saw mobile
gained on average 79% more credit than previously seen with cookie only
attribution.
Advertisers should ideally work with partners that can track the impact of
mobile on influencing a conversion to help accurately attribute ROI and
optimise their marketing spend.
Realigning media investment closer to the attributed credit, offers a
potential increase in value at little or no additional cost.
Applying learnings from People-Based Custom MTA, through a test and learn
program, would enable a properly allocated budget, in other words, more
conversions for the same paid media cost.
Important Takeaways
1.
2.
3.
insights@datalicious.com datalicious.com/optimahub
16. AboutDatalicious,Methodology
16
Datalicious is a full service analytics agency and technology firm, providing the tools and
insights to help companies achieve more effective marketing outcomes. From our beginnings
as an analytics consulting agency in Australia, Datalicious has expanded globally through our
growing product and services divisions. Datalicious products include the SuperTag tag
manager, DataExchange user ID management tool and OptimaHub cross-channel marketing
analytics platform.
Datalicious is part of the Equifax Group.
About Datalicious
Uses temporal dynamics via the Markov Chain Monte Carlo approach to find more accurate
weights based on both the position of an ad and the eventual user activity.
Identifies hidden user intent via externally observable activity to better tailor attribution impact
to actual marketing strategies of targeting customers at each stage differently.
Follows a nonlinear real world Bayesian distribution which avoids the “assumption” problems
and “p-hacking” that are inherent in traditional linear regression techniques.
Logit / Logistic: This model does not include any time dynamics or heterogeneity amongst
consumers. Logistic model “assumes” the same impact of an ad across any and all time of a
user’s journey.
Latent-Class Models: Although this model accounts for differences in consumer behavior and
can identify the latent stage of a customer’s need, there are no temporal dynamics and
consumer behavior is “assumed” to not change over time.
Shapely: Makes linearity assumptions leading to multicollinearity issues and is heavily reliant
on p-value based validations. Also the game theory assumptions of distributing residuals can
be debated to be of a lesser entropy distribution.
Other Techniques
HMM (Hidden Markov Model)
Benefits of HMM over other techniques
insights@datalicious.com datalicious.com/optimahub
17. Citations
18
Datalicious and Facebook worked with three Australian brands, Telstra, Foxtel and Medibank
across a three month period from October - December 2016 with an aim to understand the
impact of using a people-based marketing attribution model compared to a more traditional
cookie-only-based approach.
The data was collected through Atlas tracking tags that measured real people across device,
browsers and apps. The data included people’s media consumption as well as their
behavioural data on the brand website including any conversions.
The collected data was sent to the and processed via the Datalicious OptimaHub customer
journey and media analytics platform for processing and marketing attribution analysis. The
Datalicious approach to marketing attribution is built using a Hidden Markov Model (HMM).
The HMM allows us to better tailor attribution impact to actual marketing strategies of targeting
customers at each customer journey stage differently. The model uses a nonlinear Bayesian
distribution and avoids the problems of assumption that are inherent to traditional linear
regression techniques.
The figures presented in this study are based on hundreds of millions of observed data points
and thus deliver a much more accurate and scientific measure of true attribution compared to
econometric modeling based approaches.
The study results are an average across multiple clients and may vary for your own company
and vertical so we highly recommend you implement your own people-based measurement
and marketing attribution solution for the best possible results for your business.
Facebook materials are property of Facebook and/or its licensors and may not be
reproduced, in whole or in part, without Facebook’s express prior written permission.
insights@datalicious.com datalicious.com/optimahub
18. For more information or to discuss how the Datalicious OptimaHub people-based measurement can
improve your cross-device marketing effectiveness, please visit datalicious.com/optimahub
or contact us on insights@datalicious.com.
People-Based Attribution