2 23702 getting-it_right_the_road_to_genuine_marketing_attribution

a WHITEPAPER from MEDIAPLEX

GETTING IT RIGHT:

THE ROAD TO GENUINE MARKETING ATTRIBUTIONBy Matt AnthonySenior Director – Analytics, MediaplexValueClick, Inc. (NASDAQ: VCLK)

GETTING IT RIGHT: THE ROAD TO GENUINE MARKETING ATTRIBUTION © COPYRIGHT 2012 MEDIAPLEX. ALL RIGHTS RESERVED. MEDIAPLEX.COM

PG.01 INTRODUCTION

PG.01 ANALYTICS SPENDING IS ON THE RISE

PG.02 ATTRIBUTION TAKES MANY FORMS. BUT ARE THESE FORMS ACCURATE?

PG.03 THE DANGER OF SIMPLISTIC MODELS

PG.04 AGENCIES AND ATTRIBUTION

PG.05 ATTRIBUTION IS JOB ONE

PG.05 DOING ATTRIBUTION RIGHT:

6 CHARACTERISTICS OF EFFECTIVE ATTRIBUTION APPROACHES

PG.06 THE CHARACTERISTICS IN DEPTH

PG.06 1. UNITED, CLEAN, HIGH QUALITY DATA

PG.06 2. THE RIGHT TEAM

PG.07 3. CUSTOM, ALGORITHMIC APPROACH TO ATTRIBUTION

PG.08 4. FOCUS ON CAUSALITY

PG.09 5. THREE LEVELS OF ANALYSIS AND RECOMMENDATIONS

PG.10 6. THE WILL TO ACT

PG.11 CONCLUSIONS

PG.12 ABOUT THE AUTHOR

PG.12 ABOUT MEDIAPLEX

01GETTING IT RIGHT: THE ROAD TO GENUINE MARKETING ATTRIBUTION © COPYRIGHT 2012 MEDIAPLEX. ALL RIGHTS RESERVED. MEDIAPLEX.COM

You’re proably familiar with the above quote from department store pioneer John Wanamaker. It’s entertaining of course, but it’s also proof that the marketing attribution challenge is as old as marketing itself. Every year, hundreds of billions of dollars are spent on marketing tactics – often with patchy (or even no) evidence that all that money is delivering genuine value to the business.

For decades, brands have struggled to identify accurate ways of assessing marketing impact. The advent of digital has inspired tremendous marketer interest in analytics and accountability.

One of the key drivers of this interest is the recognition that brands need to be more effective at understanding “what works.” A 2012 survey of marketers conducted by IBM showed that just 34% of marketing organizations have a “sophisticated approach to ‘investing’ marketing resources and ‘engaging’ customers across multiple channels.”ii

ANALYTICS SPENDING IS ON THE RISE

According to a recent survey of more than 500 leading marketers conducted by Christine Moorman for TheCMOSurvey.org, the percentage of total marketing budgets devoted to analytics is expected to grow by almost 80% in the next three years.

A 2012 study conducted by Google and eConsultancy showed that marketers have a variety of goals for their marketing attribution efforts, chief among them justifying budget and improving the marketing mix for better business results. Given this, accuracy

INTRODUCTION

i Wikipedia: http://en.wikipedia.org/wiki/John_Wanamaker

ii IBM The State of Marketing 2012: IBM’s Global Survey of Marketers http://www.slideshare.net/165yohodr/the-state-of-marketing-2012-ibms-global-survey-of-marketers-final

iii Christine Moorman and TheCMOSurvey.org 2012 http://cmosurvey.org/results/

“ Every year, hundreds of billions of dollars are spent on marketing tactics – often with patchy evidence that all that money is delivering genuine value.”

“HALF OF MY MARKETING IS WORKING, I JUST DON’T KNOW WHICH HALF.”

iii

i

8.0%

13.5%

0%

5%

10%

15%

Current Levels In The Next Three Years

Percent of Marke-ng Budget Spent on Analy-cs Percent of Marketing Budget Spent on Analytics


is paramount when estimating the revenue tied to particular channels or tactics, since inaccurate attribution would end up negatively affecting results.

Since such issues affect the daily lives of most marketers, it’s quite natural that interest in attribution has become rather keen.

ATTRIBUTION TAKES MANY FORMS. BUT ARE THESE FORMS ACCURATE?

According to the IBM survey, 73% of marketing organizations attribute marketing effectiveness to specific tactics in some way. But delivering accurate sales impact attribution across media touchpoints is rather complicated.

The vast majority of marketing organizations use one (or more) of five attribution models:

• Last-Click (or last View): Gives full credit to the final marketing event immediately preceding a conversion.

• First-Click (or First View): Gives full credit to the first marketing event to which a user is exposed during a campaign.

• Even Weighting: Assigns an equal fraction of the credit to every marketing event the customer is exposed to prior to converting.

• Business Rules Weighting: Applies a set percentage of the credit to each marketing event preceding a conversion based upon a predetermined weighting scale.

• Exponential Weighting: Applies a progressively larger percentage of the credit depending on how close marketing events were to the time of conversion.

Each of these models is called “rules-based,” meaning that the model is assumed/assigned by the brand or agency, rather than being derived based upon a comprehensive analysis of a brand’s data.

“ The vast majority of marketing organizations use one (or more) of five rules-based attribution models.”

iv eConsultancy and Google Analytics: Marketing Attribution: Valuing the Customer Journey http://services.google.com/fh/files/misc/marketing_attribution_whitepaper.pdf

62% 57% 47%

36%

28% 36% 42%

39%

10% 7% 11% 25%

0%

25%

50%

75%

100%

Jus0fying Digital Spending

Create Most Effec0ve Media Mix

Understanding Funnel and Sales Cycle Length

Correctly Determining Affiliate

Payments

High Priority Medium Priority Low Priority

Marketer Goals for Attribution


The Google and eConsultancy study revealed that the majority brands and agencies use the simpler methodologies to address the attribution challenge. Specifically:

•54% of agencies and brands focus on last-click attribution (e.g., crediting an entire sale to the last touchpoint before purchase).

•41%ofagenciesand25%ofbrandsmanuallyassignweightingstodifferentchannelsandapplythosetotheirimpactanalyses.

•41%ofagenciesand24%ofbrandsassignfullcredittothefirst-clickaconsumermakesonamarketingexperience.v (Note, figures add up to greater than 100 because companies may use more than one method across businesses, campaigns, or time periods.)

In practice brands and agencies tend to choose the simpler methods for attributing sales credit, even though their leaders recognize that these approaches likely distort the true value of tactics. The simple methods distort because they are rules- rather than evidence-based. Attribution “credit” is assigned rather than demonstrated through statistical analysis. For example, while last-click is the most common attribution methodology, it is also the least trusted, according to that same survey.

THE DANGER OF SIMPLISTIC MODELS

The problem with these models is that they are simplistic, meaning that they make broad assumptions about the data to simplify the math of attribution. These attribution models are typically selected because they are easy to implement, not because they are accurate. It is natural for business people to want a simple solution, but simple and simplistic solutions are two very different things.

v IBID http://services.google.com/fh/files/misc/marketing_attribution_whitepaper.pdf

vi IBID http://services.google.com/fh/files/misc/marketing_attribution_whitepaper.pdf

vii IBID http://services.google.com/fh/files/misc/marketing_attribution_whitepaper.pdf

vi

54%

41%

41%

35%

13%

29%

19%

5%

54%

25%

24%

20%

16%

10%

9%

7%

0% 10% 20% 30% 40% 50% 60%

Last Click

Customized by Channel

First Click

Unique Methodology

Not Sure

Linear

ExponenJal

Other

Agency Brand

Most Common Methods of Attribution

vii

5% 13% 17%

9% 8% 25%

18%

49% 64% 69%

61% 55%

29% 27% 23% 14% 14%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

100%

Customized by Channel

Unique Methodology

Linear First Click Last Click

Very EffecIve Somewhat EffecIve Somewhat IneffecIve Very IneffecIve

Marketer Assessment of the Effectiveness of Attribution Models


I say these models are simplistic because for them to truly be accurate, then virtually every one of your customers would have to share a common “path to conversion”. That is absurd. And besides, if you knew the correct attribution weightings, why would you go to the trouble and expense of buying a rules-based attribution tool? It would add little value.

Another issue with rules-based models is that they minimize or ignore the importance of demographic, psychographic, and behavioral user characteristics and their impacts on conversion. Rules-based models are focused primarily on an imagined sequence of marketing events, even though extensive research has demonstrated that set paths don’t actually exist.

Using a simplistic attribution approach does make the attribution math easier, but using such a model without first demonstrating its validity is inherently dangerous. We might presume, for example, that spreading credit evenly across all marketing events leading up to a conversion would be more accurate than giving all credit to the final click.

Sounds sensible enough, right? It seems logical that preceding events had to have had at least some role in causing a conversion.

But did they? Without actually analyzing a brand’s data, that assumption is just as unproven as last-click or any other attribution model. It can be just as inaccurate – perhaps even more inaccurate than last-click.

The whole point of attribution is to know rather than guess. To attach revenue estimates to various ad channels and tactics with the ultimate goal of changing overall strategy and marketing mix. Getting it wrong means making changes and choices that could be bad for your business!

You may also be familiar with A/B testing methodologies that demonstrate the incremental value of marketing within a particular channel. They are common at the individual media vendor level, but it’s impractical to use them at the media plan level or in a cross channel analysis at this time.

AGENCIES AND ATTRIBUTION

Brands that rely on their agencies to measure and properly attribute credit to marketing investments need to be cognizant that most agencies use one or more of the rules-based methodologies. That’s even though most agency and brand people believe they aren’t very accurate.

That’s not a criticism of agencies. Unless you are paying a separate line item for analytics-based attribution in your agency contract, your agency’s margins likely force them to use simplistic

“ The whole point of an attribution is to know rather than guess.”


models. Expecting agencies to accurately deliver attribution without paying for it is absurd. Agencies are businesses after all.

Historically, many clients have been unwilling to pay additional fees to their agencies for analytics-based attribution. As a result, their agencies have been forced to choose unproven rules-based approaches. They do their best with the resources you give them.

If you expect your agency to deliver a genuine attribution solution, you’re going to need to pay separately for that service. Today’s lean commissions and fees simply don’t allow room for it.

ATTRIBUTION IS JOB ONE

If maximizing marketing ROI is your leading objective, then finding out precisely how to attribute credit for brand sales should be your most important challenge for the year. While simplistic solutions have an appeal, they ultimately put your brand’s success at serious risk. And here’s the rub: basing your action plan on unverified or simplistic models is a completely unnecessary risk. It’s actually possible to scientifically estimate the business impact of all marketing touchpoints, and to use this information to optimize your marketing allocations.

Doing attribution right isn’t a problem waiting for someone to devise a solution. The solution is here today. You can do it right now, and enjoy its benefits right now. Let me take that statement one step further. You NEED to do it right now. Because as business goals get tougher, achieving them will take greater marketing effectiveness, made possible by genuine insight.

DOING ATTRIBUTION RIGHT: 6 CHARACTERISTICS OF EFFECTIVE ATTRIBUTION APPROACHES

We’ve talked a lot about how not to go about attributing marketing impact. Let’s spend the balance of this paper defining how to do attribution right. On a fundamental level, a bona fide

“ As business goals get tougher, achieving them will take greater marketing effectiveness, made possible by genuine insight.”

“ Expecting agencies to accurately deliver attribution without paying for it is absurd. Agencies are businesses after all.”


attribution solution must have the following six characteristics:

1.UNITED,CLEAN,HIGHQUALITYDATA

2.THERIGHTTEAM

3.CUSTOM,ALGORITHMICAPPROACHTOATTRIBUTION

4.FOCUSONCAUSALITY

5.THREELEVELSOFANALYSISANDRECOMMENDATIONS

6.THEWILLTOACT

Let’s examine each of these points individually.

THE CHARACTERISTICS IN DEPTH1. UNITED, CLEAN, HIGH QUALITY DATA

In order for a marketing attribution solution to live up to its full potential, it must be able to bring together all of your data – cross channel marketing data, audience data, site data, and of course conversion data (online and offline). All of this information must be combined to create a unified cross channel dataset. The goal here is to create a “user-eye view” in the data. To achieve this, a unified User ID system (or an accurate way to associate multiple system IDs to each other) is essential.

It’s critical that you identify a solution that can integrate everything effectively. Data must be standardized, cleaned, and related to other data in order to deliver maximum value.

What makes it such a difficult data management challenge is that different platforms utilize different data structures, formats and taxonomies. For example, working with Affiliate and Lead Gen data is rather different from working with Search or Display data. After being combined, data must be properly standardized, cleaned, and classified in order to deliver maximum value. Given all this, it’s critical that your attribution solution has the data management capabilities to integrate everything effectively.

For Mediaplex this is not a new challenge. We’ve been collecting and managing data across sites, Display, Search, Video, Mobile, Affiliate and Rich Media for years. But whether you hire us, build a team in-house or work with another vendor, ensure that these data management capabilities are covered off VERY thoroughly. They are absolutely critical to laying the right groundwork for accurate attribution.

“ It’s critical that you identify a solution that can integrate everything effectively.”


2. THE RIGHT TEAM

To perform statistically sound analysis on brand data, you need – guess what? – a statistician. Perhaps more than one. The mathematical challenges of achieving accurate marketing attribution are huge. The processes required to derive insights and answers from gigantic cross channel data sets are beyond the capabilities of most marketing generalists. It takes far more than being a whiz at Excel.

Similarly, it challenges credulity to think that a pretty “attribution dashboard” alone can deliver true analytics. The biggest issue with SaaS-based D-I-Y solutions is that almost by definition they must leverage surrogate measures, incomplete math, and simplistic models to make themselves easy to implement and use. The uniqueness of your brand’s business gets lost when you use cookie cutter platforms.

There are many great analysts in the world who aren’t PhDs in Statistics or Mathematics. But analysis is different from analytics. And revealing the true business impact of marketing tactics requires analytics, not reporting or analysis. I’m not being pedantic here. The words simply describe different things.

If you accept that you need statisticians to do the job, your next challenge is to get them on your team. You can hire them yourself or you can work with a solution provider that has an in-house team you can leverage.

In my view, outsourcing is preferable for most companies. Here’s why. In addition to the infrastructure costs necessary for warehousing, storing & processing the data, the cost of acquiring a first rate analytics/statistics team is very high. Demand for high quality statisticians well outstrips supply. According to a 2011 McKinsey Global Institute study, by 2018 the US “faces a shortage of 140,000 to 190,000 people with deep analytical skills as well as 1.5 million managers and analysts with the know-how to use the analysis of big data to make effective decisions.” viii

viii McKinsey Global Institute, 2011

ix IBID

190,000

-‐

50,000

100,000

150,000

200,000

250,000

300,000

350,000

400,000

450,000

500,000

2016 Deep Analy5cal Talent Supply

Low Es5mate -‐ 2016 Demand

High Es5mate -‐ 2016 Demand

The Deep Analy,cs Talent Gap

140,000

285,000

The Deep Analytics Talent Gap

ix

“ It takes far more than being a whiz at Excel.”


When you outsource to a vendor that already has a top tier team and a world-class data infrastructure, you avoid the capital investment, hiring and salary expenses as well as the time lag for finding and onboarding qualified individuals. You can be acting on insights and optimizing your plan far more quickly with an outsourced model.

3. CUSTOM ALGORITHMIC APPROACH TO ATTRIBUTION

There is no shortage of companies claiming to offer a SaaS platform that can solve all of your attribution problems. Unfortunately true attribution is both custom and people intensive. You cannot apply a finite set of rules to every company. Nor can you “guesstimate” the proper credit allocation with any degree of accuracy. Guesstimation is guess work after all.

As we observed earlier in this paper, there are many possible paths to conversion, not one. Further, consumer paths are constantly evolving and changing. Don’t rely on any attribution “solution” that depends upon on unverified “rules of thumb” to attribute impact to marketing vehicles. To get this right, statisticians must examine your actual brand data to identify the causative effects of various marketing event types.

An algorithmic solution is the choice of true statisticians. It leverages your actual in-market data to derive a unique, brand-specific model that can then accurately quantify the effects of your marketing tactics. It reflects your category, your specific customers, and the state of your brand. It can be optimized to the specific metrics you care most about.

Further, the dynamic nature of the media environment is constantly changing the roles that different media play in a customer’s decision journey. If you develop an allocation based upon the data for a particular period of time, its validity will likely be very high for a period immediately following. But over time an allocation loses its validity as the digital environment and consumer behavior change. Consider, for example, how Social has upended the conversion process in so many categories. The easy availability of expert– and crowd-based recommendations has dramatically altered the buying landscape.

And Social isn’t the only vehicle that is dramatically altering the journey. As Search becomes more multiplatform and begins to include more visuals, its role is radically changing as well. Similarly, Rich Media is enhancing the role of Display in some categories – a banner with a video player has markedly different impacts than a static GIF format.

“ The dynamic nature of the media environment is constantly changing the roles that different media play in a customer’s decision journey”


The marketing effects of different tactics are constantly evolving, so marketing attribution is an ongoing commitment. Make sure that the attribution approach you choose can accommodate the dynamic nature of digital so that you can adjust to new market realities.

4. FOCUS ON CAUSALITY

The ultimate goal of any marketer is to discover what mix of marketing investments drives the most conversions. What makes marketing attribution so complex is that correlation looks a lot like causation. It is far easier to observe that converters seem to share a common trait than statistically proving that that trait is instrumental in causing the conversion. Just because exposure to a particular marketing event is common among conversions (correlation) doesn’t mean it actually influenced to those conversions (causality).

It might be that exposure to a particular tactic is simply common across the entire population. That investment may have in fact been wasteful. It’s the “post hoc ergo propter hoc” fallacy. “After this, therefore because of this.”

There are proven statistical methods that can ensure you don’t fall into this trap. In order to deliver real insight into “what works,” aka causality, you need to meet a certain burden of statistical proof. You need to understand the differences between converters and nonconverters to tease out events that correlate with conversion from those that actually cause it.

The quest for causality – a fundamental marketing attribution essential – requires the right inputs and the right analytics. Without them you may easily mistake correlation for causation and may actually allocate resources in a way that diminishes return.

5. THREE LEVELS OF ANALYSIS AND RECOMMENDATIONS

When most marketers think about the value of marketing attribution, they focus primarily on its ability to help right-size budget allocations across media. Their hope is that it can provide recommendations on increasing or decreasing the size of each channel budget to achieve maximum results. That’s the modern expression of John Wanamaker’s question.

This is a key part of what marketing attribution should do. But there are other parts as well. Attribution should provide answers to three questions, not just one:

“ Attribution should provide answers to three questions, not just one.”

“ What makes marketing attribution so complex is that correlation looks a lot like causation.”


1.HowshouldIallocatemyspendingacrosschannelstoachievemaximumROI?

2.HowshouldIallocateresourcesacrossvendorstoachievemaximumROI?

3.Whatistheprecisevalueofareal-timeimpressionforaparticularperson?

The way a statistician thinks about attribution is to use data analysis to estimate the effects of different marketing events while controlling for other factors. The reason is that if we can get good estimates of the revenue being generated by each event, we can create good estimates of the relative value of each marketing activity and how to reallocate our spend to improve results.

Channel level allocation helps marketers fine tune their investments across various marketing vehicles. This can be extremely valuable counsel as certain simplistic attribution models, like last-click attribution, favor so called “bottom of the funnel” vehicles.

Vendor level allocation helps us choose partners that are delivering genuine, incremental marketing value to a plan. It helps us understand if a step on a consumer journey is actually causal. For example: using a last click attribution model, Paid Search often gets all the credit for sales. Search plays a significant role in some cases. But a large percentage of consumer searches are what is called “navigational searches,” meaning that the search engine is simply used to make it easier to visit a website to buy something they were already going to buy. In those cases, Search is getting “credit” for sales it really didn’t drive.

The third, impression-level analytics and recommendations, relate specifically to the real-time bidding (RTB) media exchanges that have emerged as an important media buying channel. A robust marketing attribution solution should be linked to your demand side platform (DSP). That way, as an impression becomes available on the exchange your tool has a precise way of assessing its value and bidding accordingly.

6. THE WILL TO ACT

OK, this attribution requirement is a little different from the others because it relates to how you get maximum business impact from genuine marketing attribution.

Something happens on the way to an analytics-based brand strategy in many organizations; the focus drifts to verifying the relevance and validity of our current strategies, rather than identifying the best approaches for growth.

The challenge of analytics-driven marketing is that it requires commitment and alignment across


an organization. There can be no sacred cows, no pet tactics. Or at least if there are sacred cows, we need to recognize that that’s exactly what they are – that there are reasons other than ROI for why we do them.

For some people reading this, a fact-based approach to tactic selection will seem a breath of fresh air. So many marketers I have spoken with have said that their efforts to optimize marketing are hindered by fad-driven decision-making higher up. “We need a widget!” “We need an app!” “Why doesn’t our toilet paper brand have a Twitter account?”

As a statistician, I am a purist. If I had my druthers, everything would be quantifiable and measurable. Not everything is quantifiable yet, but an organization aligned to the value of analytics will get closer and closer to that goal over time. An analytics-based approach demands that we recognize that better information has given us new insight. Knowing and acting on that knowledge is a good thing. Knowing and not acting is irresponsible.

One final note: as we put marketing attribution insights into practice, we need to ensure we don’t throw out babies with the bathwater. A pure analytical approach might suggest that some tactics are ineffective and need to be reconsidered. But this in no way questions the worth of individuals who were charged with implementing those tactics. They might, for example, have been incredibly adept at getting extraordinary results from a tactic even if it wasn’t the best use of funds. For that, the individual deserves praise, not condemnation.

CONCLUSIONSMarketing attribution is absolutely critical to your brand’s future success. No brand has resources to waste pursuing marketing approaches that aren’t driving maximum ROI.

In fact, in my view measurement is the #1 issue holding back digital marketing from scaling to a $100B industry. Marketers can’t make wholesale shifts in channel allocation from traditional to digital without proof. And real proof only comes from algorithmic causal attribution.

The sobering reality is that most brands do not currently attribute marketing effects in a genuine, fact-based way. It’s also unfortunate that most of the “marketing attribution” tools and platforms out there

“ No brand has resources to waste pursuing marketing approaches that aren’t driving maximum ROI.”

“ The challenge of analytics-driven marketing is that it requires commitment and alignment across an organization.”


aren’t really capable of doing the job right. But all is not lost! Believe it or not, genuine marketing attribution is quite achievable. And in my view essential if marketing is to deliver on its full potential for brands.

We hope that this white paper has made it a little easier for you to vet potential partners and choose the best marketing attribution option for your business. I’d love to hear from you – about your thoughts and experience in the world of marketing attribution or in the process of choosing a partner. Please don’t hesitate to email me at [email protected].

ABOUT THE AUTHORMatt Anthony is the Senior Director, Analytics at Mediaplex, the technology and analytics arm of ValueClick, Inc. (NASDAQ: VCLK). Matt’s responsibilities include analytical technologies development for the main measurement and optimization tools that make up the backbone of the Mediaplex analytics framework, as well as overseeing the Analytics Services business, where Mediaplex offers full-service and custom analytical engagement support to its clients. Matt joined Mediaplex in 2011, coming over from The Sports Authority, Inc. where he served as the Director, Analytics for the eCommerce division. Prior to that, Matt was Director, Research and Development at Datalogix (f.k.a. NextAction), where he oversaw algorithm and technological research supporting their full product line. Matt has an MS, Statistics from Harvard University, and BS, magna cum laude in Mathematical Social Science from Dartmouth College.

ABOUT MEDIAPLEXMediaplex, a division of ValueClick, Inc. (NASDAQ: VCLK), is the technology-empowered analytics and optimization provider that scientifically identifies the best way for brands to invest marketing resources and then executes against those insights with a suite of advanced campaign management tools. Its unique software with service model delivers industry-leading technology coupled with the deep analytics expertise necessary to answer Marketing’s most mission-critical question: how to maximize ROI from marketing investments. With all the essential software and services needed to correctly collect, manage, interpret, and action data, Mediaplex’s end-to-end solution is the first to make genuine data-driven marketing a reality.

For more information visit Mediaplex.com or contact us at 1.877.402.PLEX (7539).