6. Continued growth in digital marketing spend
is spurring interest in ROI measurement
U.S Online Ad Spend (in billions)*
$62
$58
$60
$53
$47
$50
$40
$40
$32
$30
$20
$10
$2011
2012
2013
2014
2015
2016
----------------------------Estimated ----------------------------
6
*Source: eMarketer, Jan 2012
7. Diversification of channels is a major
trend in digital marketing
U.S Online Ad Spend by Channel (in billions)
Search
Banner Ads
Classifieds/Directories
Video
Rich Media
Other
$50
$40
$30
$20
$10
$2010
2011
2012
2013
2014
2015
----------------Estimated ----------------
7
9. Attribution assigns credit for
conversions
Brand
Keyword
Search
Email Ad
NonBrand
Keyword
Search
Affiliate
Network
Banner
Social
Media Ad
Conversion
9
Lays foundation
for digital media
ROI measurement
Optimizes
Budget
Increases
ROI
10. Current approaches to attribution rely
on the marketer to assign credit
First - Click Attribution
100% credit
100% credit
Last - Click Attribution
Shared Credit Attribution
25% credit
25% credit
25% credit
Recency - Based Attribution
10
25% credit
10% credit
20% credit
30% credit
40% credit
11. Poll: Which attribution method do
you use?
First/Last Click
Shared Credit
Recency-Based
Algorithmic/Scientific
Other
None
11
12. Most digital marketers are using “traditional”
attribution – or no attribution at all!
IAB Poll*: which attribution method do you use?
None
44%
First or Last
Click
30%
Shared Credit
17%
RecencyBased
17%
Algorithmic /
Scientific
Other
12
11%
1%
*Source: Interactive Advertising Bureau – Digital Attribution Comes of Age
14. Current methods assign credit in a
biased and arbitrary way
First - Click Attribution
100% credit
Ignores Everything After
First Click
100% credit
Last - Click Attribution
Everything Is Ignored
Except Last Click
Shared Credit Attribution
25% credit
25% credit
25% credit
25% credit
10% credit
20% credit
30% credit
40% credit
Constant Weights Used
Throughout
Recency - Based Attribution
Subjective, for those with “Last
Click Separation Anxiety”
14
15. Our understanding of channel-level
performance is distorted by current attribution
Display Media
Display
Display
• Primarily an assist media
• Generates very few clicks
Search
Paid Search
Conversion
• Primarily a converter
• Generates many clicks
15
16. Advanced attribution approaches need to
address limitations of traditional methods
Traditional
Solution
Uses
incomplete view
& click data
Utilize all
available data
Ad Server data
logs
Requires
marketer
bias
Rely on data
to drive
results
Statistical
Analysis
Easy to
implement
16
Advanced
Maintain
implementation
ease
Integrate
findings with
media buying
17. Recent advances are enabling data-driven
attribution for advertisers and agencies
Demand for more precise
online ad spend ROI
Availability of
granular data
Advanced modeling
techniques
17
DataDriven
Attribution
Vendors
18. Independent studies confirm the rising
tide of attribution modeling
Forrester Wave: Interactive
Attribution Vendors
“Last click still survives but we
see accelerating momentum to
cast it aside and implement
interactive attribution”
“Since 2009…the market has
gained momentum”
“Benefits for digital marketers
include […] improved
multichannel
measurement, data-driven
spending and strategy
decisions, and media buying
leverage”
“Algorithmic attribution [is]
helping value marketing
efforts, identify and eliminate
waste, and optimize media
plans…to boost ROI”
18
IAB: Digital Attribution Comes
Of Age
“44% of interactive marketers
don’t have processes in place
to assign credit to their
efforts, indicating that adoption
of attribution has room to grow”
19. Scientific Attribution delivers on the
promise of full funnel measurement
Email
Display
Display
Search
Social
Rich
Media
Search
Display
Search
Purchase
Purchase
Purchase
Account for all touch points when assigning purchase credit
19
20. A scientific approach to attribution:
important factors
Advertising
expertise
• Digital agencies manage millions in online ad
spend each year
• Understand the challenges of online media
optimization
Customized
statistical
approach
• Build custom models to understand
true value of media channels
Unbiased
results
20
• No prior assumptions are made
• Results reflect the impact of each
channel on purchases
21. A statistical model unearths relationships
between a set of factors and a response
Moderate
Predictor
Strong
Predictor
Purchase?
User
Search
A
Yes
Yes
Yes
B
Yes
No
No
C
No
Yes
Yes
D
No
No
No
E
21
Display
Yes
Yes
Yes
22. Scientific Attribution also incorporates non-media
factors that play a role in purchase behavior
Landing
page
Seasonality
Industry
changes
Media
channels
Geography
Competition
Time of
day/
Day of
week
22
23. Understanding (and potentially controlling)
non-media factors helps boost optimization
Landing
Page
• Determine which landing pages drive
conversions
• Optimize and test based on results
Seasonal
Effects
Geography
23
• Key understanding of advertiser’s business
• Develop a seasonally-adjusted media
strategy
• Discover which states respond to different
channels/messages
• Increase geo-targeting efficiency
24. Scientific Attribution should be continuously
improved through collaboration
Data
collection
Everyone!
Advertiser/
Media
Agency
feedback
Media
24
Performance
reporting
Analytics + Database
Attribution
Modeling
Channel
optimization
Analytics
Analytics + Media
25. Create actionable results via “Full Click”
metrics
Media Channel
Last Click
CPS
“Full Click”
CPS
Difference
Google Brand
$24
$29
-21%
- budget
Google Non-Brand
$58
$36
+38%
+ budget
Bing Brand
$22
$23
-5%
Bing Non-Brand
$48
$29
+40%
Ad.com
$87
$94
-8%
$105
$76
+28%
Google Display
$65
$43
+34%
Rich Media
$41
$25
+39%
ValueClick
25
+ budget
28. BKV’s integrated approach to attribution
proved useful for a major Telecom client
Goal: increase media efficiency and
identify valuable “assist” keywords
Solution: Scientific Attribution
model, focusing on ease of
implementation
Result: more efficient media buys
28
29. BKV’s integrated approach to attribution
proved useful for a major Telecom client
1
• Discuss objectives
and goals with
client
2
• Build custom
model for client
• Collect/process
appropriate data
• Determine most
appropriate
modeling structure
& approach
29
• Extensive testing
process to verify
model findings
• Create actionable
recommendations
based on modeling
results
3
• Implement
Scientific
Attribution
• Report on
findings over
several weeks to
gauge impact
• Initiate
refinements for
future iterations
30. Scientific Attribution delivered tangible
results upon implementation
Scientific Attribution
implementation results
2% daily
spend
increase
30
22% daily
orders
increase
16% cost per
sale
decrease
31. Scientific Attribution will continue to
grow and evolve
Understanding mobile
marketing & cross-device
ROI
Determining the value of
social media presence
31
33. Scientific Attribution success relies
advance planning
1
2
Identify a partner for statistical modeling – with
extensive media optimization and analytics experience
3
33
Verify that all your online marketing is being tracked
and that cookie-level data is available
Create a detailed plan for implementation, including a
process for continual improvement
34. Scientific Attribution is a crucial
component of an online media strategy
Traditional methods
fall short
Industry-wide focus
to improve
digital attribution
34
Scientific Attribution
not biased
Case studies show
strong results
35. Thank you!
Visit bkv.com for more!
Dave Mundo
VP, Analytics Director
dave.mundo@bkv.com
@davemundo
35
Notes de l'éditeur
Introduce.
Aside from being a full-service media agency, BKV’s Analytics team offers a full suite of marketing analytics solutions focused on both online and offline media. We work with clients to determine how their online/offline media drives their sales/conversions.
First we will discuss the concept of digital media attribution and why it is important for advertisers and marketers to understand it and get it right.Then we will explain a data-driven, scientific approach that ultimately results in efficient investment allocation across all digital media touchpoints.We will follow that with a recent case study for a telecom advertiser to show how we implemented Scientific Attribution and how it benefitted our client.Finally, we will discuss the best practices for getting started with advanced, scientific attribution. This is meant for advertisers who want to use this powerful approach to execute digital media as efficiently as possible.
With that, let’s get started…(build)our goal in this first section is to define attribution and review commonly-used traditional attribution approaches.
Why do advertisers need to understand digital media? One reason is its steady historical and projected growth.As more marketing dollars shift to digital, there is a growing need amongst marketers to obtain accurate ROI for their digital investments, especially when it comes to budget optimizations.For years marketers have analyzed data and conducted research to measure the impact of offline media. Digital media needs that same attention.
Digital marketers are increasingly spreading their budgets out across different online media channels in an effort to reach their audience.As a consequence, marketers now need to be able to compare effectiveness between different channels, even though they may work differently and have different objectives.Next, let’s explore how these different online media types fit into the sales funnel and what that means for attribution.
To measure ROI in digital marketing, we need to focus on the process of attribution. In other words, how do we attribute or assign credit to an individual digital media channel for a conversion or a purchase? A conversion or purchase could mean an online purchase, a lead, a lead or donation for a non-profit, etc.(build)Intuitively, we all understand that attribution has a huge impact on marketing decisions. The media tactic that receives the credit for the most sales is likely going to have a high ROI, which means that we’re likely to continue allocating marketing dollars towards that tactic in the future. As we optimize digital media budgets, we will benefit from an improved digital ROI.Next, we’ll look at how most digital marketers are currently handling attribution.
Let’s look at current commonly-used attribution models.(Start animation and explain each model.)Looking at the current approaches to attribution, we’ll notice right away that they all have one thing in common: they rely on marketers to decidehow credit for sales is distributed. In the agency world, the last-click attribution method is probably the most popular, and has been the cornerstone of ROI measurement since the inception of digital marketing.
Now we’d like to poll all of you to find out which attribution methods are most common among our audience.
Attribution methods are beginning to gain traction, with a small % of respondents opting for advanced, scientific approaches.
Now let’s explore Scientific Attribution, and understand(build)why marketers need it, how it works, and how to execute it.
Let’s look at current commonly-used attribution models.(Start animation and explain each model.)Looking at the current approaches to attribution, we’ll notice right away that they all have one thing in common: they rely on marketers to decidehow credit for sales is distributed. In the agency world, the last-click attribution method is probably the most popular, and has been the cornerstone of ROI measurement since the inception of digital marketing.
Current attribution models can easily bias results towards “bottom of the funnel” digital media channels. For example, Display generates few clicks and tends to get shortchanged in many attribution approaches. Paid Search acts as a converter and in some cases may benefit from the a priori biases inherent in traditional attribution techniques.The same can be said for other digital media touchpoints (SEO, affiliates, etc.). Quite simply, traditional approaches can lead to subpar optimization and digital media ROI.
So clearly there are limitations with traditional approaches. Any advanced, scientific approach to attribution needs to directly address these shortcomings.(build)Any advanced solution that can reliably deliver on these three points will result in an improved data-driven measure of digital media effectiveness.
Besides media agencies getting serious about scientific media attribution, many vendors dedicated to advanced attribution have surfaced. Therefore, there are numerous options and approaches for advertisers searching to “solve” digital attribution.Demand for a more precise ROI read, availability of granular ad server data, and advanced modeling methods are driving this.
Advanced scientific attribution (aka algorithmic attribution) is gaining increased traction in the industry, as we’ve seen from both the influx of scientific attribution practitioners and industry research reinforcing its importance.Forrester recently published a report on the state of scientific attribution as well as highlighting a sample of vendors in this space.The Interactive Ad Bureau similarly released a study underscoring the importance and likely growth of advanced attribution techniques.
Historically the impact of “top of the funnel” digital media interactions has been an unknown.(build)Scientific Attribution, however, delivers on the promise of measuring the impact of all media touch points and assigning credit appropriately.
We have identified factors that we feel are important when tackling attribution.Digital media expertise is crucial when attacking scientific attribution in order to ensure the results are actionable and any budget reallocations are realistic based on advertisers’ other initiatives.Work with a nimble vendor who can create custom analyses based on an advertiser’s business environment and unique complexities.
We realize talk of “statistics” often frightens people away and leaves them uttering the term “black box”. This thinking greatly underestimates marketers, as many are able to understand even complex data analyses and statistics.Scientific attribution modeling is actually not very difficult to understand.(explain slide)Now envision this for millions of user interactions and dozens of predictors. With a dataset so large, simply eyeballing relationships as we have done here is virtually impossible. Therefore, we use statistical techniques to help identify which predictors are the most important.
I mention “dozens of predictors”…this is because we are using digital media execution as a possible predictor or purchase, as well as non-media factors.
Scientific attribution is a continually evolving process that requires regular modeling refreshes that correspond to a business’ changing landscape and competitive environment.
RAFUsing traditional media management approach—if GOAL is a $40 CPA (last click only shows 2 winners, we would cut off the rest); we would be solely running within branded SEM which we know is a bottom of the funnel PULL TACTIC—we would be closing off all the media that may be bringing audiences (esp NEW audiences) into the funnel IE: non-brand SEM, display and other engagement and PUSH centric mediaOur scientific attribution uses a slew of variables that then clearly communicate an adjusted CPA. The importance here is that the findings are really easy to understand and test against. We are looking at the channel level here, but our models go down to the placement or keyword level. We now see 5 winners , some on the fence, and underperformers.(build)Once we have the data, the art of media management comes in—we need to ensure media is as stable as possible in the channels where we buy from – no or minimal increase in CPCs for SEARCH and for display that we can buy more inventory at same prices...network mix can change -- so need to have transparency and partnershipwith sites/networks you work withWe then construct a test to see if same spend levels between the test period and model period are increased and improved. So, does it WORK?
Findings only as good as the data we collectRight now there is a disconnect between desktop/laptop and mobile; however we are working with a couple of vendors that are connecting the dots—Drawbrige and TapadFB is making huge strides in attribution trackability, including third party tracking and pure play firms, however an easy step is to include view pixels on all custom tabs and to also employ third party redirects on all important clickable pieces of content, it is also important to potentially have longer windows in social…