www.adrol l .comA

The Performance Marketer’s Guide to Retargeting: Part II

MARKETERS AND ATTRIBUTIONA DISCUSSION WITH AUSTRALIAN MARKETERS ABOUT THE CURRENT STATE OF ATTRIBUTION MODELING

Survey conducted by

and

www.adrol l .com1

Marketers and Attribution

Many whitepapers have been published over the past five years on the topic of marketing attribution modelling, and all have sought to provide a précis of what it actually means.

We wanted to examine attribution as well, but given the myriad challenges and vastly different levels of knowledge surrounding it, we decided against publishing yet another whitepaper, opting instead to publish a ‘discussion paper.’ We believe this approach will provide marketers with a more useful tool, as well as become a conversation starter and springboard to more open debate and (hopefully) action across the industry.

Besides, producing an authoritative whitepaper has become almost impossible as marketers increasingly adopt custom attribution models. We’ve joked internally that the disjointed nature of this topic could only ever give rise to a murky, grey paper.

Therefore, after two rounds of consultation with leading marketers in ANZ and a survey of 165 highly targeted people across the industry in February 2016, we were able to arrive at a series of solid conclusions which will hopefully take attribution conversations to the next level.

This discussion paper explores all of the attribution models available to a marketer and identifies those that are the most frequently used. We also take a brief look at who is clicking and why it’s important, and take a deep dive into who currently handles attribution on a day-to-day basis, as well as whether that model is effective. The paper also looks at how many marketers have moved away from the age-old first- and last-click model and what hurdles face those who haven’t. Finally, we ponder the future of attribution, looking at the forces that might change attribution models, and we conclude with a brief look at the current and possible future roles of media agencies in the attribution puzzle.

Our aim is that this paper helps us all move closer to a place where, as one senior financial institution marketer said, “We’re rewarding the right channels for the right input, capturing the right person at the right time, in the right place.” Effective, accurate attribution modeling has forever been the marketer’s dream and we hope this paper will help put us on the path to making it a reality.

INTRODUCTION

www.adrol l .com2

Attribution modelling is the process of identifying a set of user actions that contribute to a desired outcome and then assigning value to each of those actions.

Put more simply: attribution modelling enables marketers to track a customer’s online path to purchase and place a value on every touchpoint that person encounters.

Despite being regarded as an essential marketing discipline, due to its complexity the mere mention of ‘attribution modelling’ has the capacity to strike fear into the hearts of even the most hardened, multi-channel marketers.

Nevertheless, 92% of ANZ marketers acknowledge the importance of attribution modelling.1

WHAT IS ATTRIBUTION AND WHICH MODELS CAN A MARKETER ADOPT?

THE MOST COMMON ATTRIBUTION MODELS

Last ClickThe last click in the path to conversion gets the credit for the sale. If there are no other clicks in the path, the last click ‘wins’.

Equal Multi-TouchEqual weight is given to every interaction in the path to conversion.

Time DecayThe closer to the point of conversion the interaction is, the more weight it is given.

Position-BasedApplies custom weighting to different points in the conversion path, which typically gives more weight to the first and last clicks.

Pure CustomApplies custom weighting to all positions based on a variety of interactions. Potentially combines all of the above.

25%

CLICK

30%

CLICK

IMPRESSION

35%

CLICK

0%

IMPRESSION

25%

IMPRESSION

20%

IMPRESSION

IMPRESSION

15%

IMPRESSION

0%

IMPRESSION

25%

CLICK

10%

CLICK

IMPRESSION

CLICK

0%

IMPRESSION

25%

IMPRESSION

40%

40%

IMPRESSION

IMPRESSION

IMPRESSION

100%

IMPRESSION

20% 20%

20%

40%

30%

“There are few things more complicated in analytics (all analytics, big data and huge data!) than multi-channel attribution modelling.”

—Avinash Kaushik, Co-Founder and CEO, Market Motive2

Attribution Modelling: A Discussion Paper

www.adrol l .com3

Marketers and Attribution

Who Is Clicking?

Back in 2009, comScore released a much-cited paper entitled ‘Natural Born Clickers’.3 It suggested that only 16% of internet users were clicking on display ads and that an even smaller number (8%) accounted for 85% of all clicks. The ANZ marketers we spoke with when creating this discussion paper believe that comScore’s numbers have most likely fallen since then.

According to a detailed report from SmartInsights4 from April 2015, stats for click-through-rates (CTRs) for display advertising look like this:

↗ Global average for display ad is 0.06%

↗ Australia average for display ad is 0.07%

↗ Global rich media ad CTR is 0.27%

Who Is Measuring Attribution In ANZ?

The low click-through rates and the changing nature of how consumers engage with advertisements and content online underpins the growing need for a marketer to truly understand the path to purchase and the relevance of an attribution model, but what percentage of organizations are measuring attribution and who is in charge of it within these companies?

The answer was revealing, showing that two-thirds of ANZ marketers have an attribution model in place: 66% of ANZ marketers (from both the brand and agency sides) are measuring attribution, while 34% declared that they have yet to implement a model, according to the AdRoll/AdNews survey.

For those companies with an attribution model, 64% of respondents say that attribution is handled by someone in the marketing department of their company, and interestingly 12% said that it is overseen by the CEO, suggesting that those higher up in the corporate ladder are realizing the importance of attribution modelling. This notion is clearly highlighted by the fact that 72% of those who handle attribution described themselves as managers, senior managers, directors and specialists. However, only 6.6% of those working in attribution would call themselves an ‘analyst’. This certainly chimes with input we received from marketers who felt there is an increasing need for analysts, particularly data analysts, to come to the attribution table.

THE LINE FROM CUSTOMER TO MARKETER—WHO IS CLICKING AND WHO IS MEASURING ATTRIBUTION?

DOES YOUR COMPANY CURRENTLY MEASURE ATTRIBUTION?

WHO HANDLES ATTRIBUTION AT YOUR COMPANY?

66% Yes

34% No

Marketing/advertising64%

12%

11%

4%

9%

CEO

Sales

Engineering

Other

www.adrol l .com4

Marketers and Attribution

From our discussion:

“There are definitely more mini touchpoints than ever before and it’s not the last click that does it. That click maybe made it easy to buy but that’s not the magic solution to all of this.”—Mariana Thomas, Head of Marketing, Tarkett

“Fundamentally the reason why last-click is still used is that the people running the e-commerce sites are the web marketers and they’re traditionally concerned about online conversion and last-click to them is really important. So what they’re doing is they’re optimizing everything to last-click because that’s what’s bringing in the traffic, but they’re not necessarily filling the funnel with more prospecting.” —Jodie Sangster, CEO, ADMA

Marketers have no doubt encountered numerous articles, whitepapers and so on about the limitations of first- and last-click attribution modelling and why it fails to adequately reflect the online consumer journey.

However, the AdRoll/AdNews survey revealed some interesting statistics about the current state of play, and how advanced the ANZ market is becoming with 66% steering away from the first- and last-click model.

34% of ANZ marketers continue to use first- and last-click attribution, but interestingly 66% are using alternative models including Linear (10%), Time Decay (12%) and Custom (22%) —the remaining 22% use undisclosed methods.

From our research, we see that custom methods are employed by the second-largest segment of those who engaged in the survey, indicating that a large portion of the market acknowledges that a Custom Model is the best solution for their brand.

“Different consumer behaviour means vastly different conversion paths and values of the interactions along the way for each path, making a single model difficult to employ. For example, people searching for phone and internet service may have a

longer conversion cycle and click on more ads during the process than those looking for a better rate on home electricity. Each path to conversion is distinctive and original, built from various touchpoints throughout the funnel and, as such, a proper attribution model must be adaptive and account for these activities and assign value accordingly ... Weep a little for that inconsiderate ‘friend’, last-click attribution.”

— Avinash Kaushik, Co-Founder and CEO, Market Motive5

ARE ANZ MARKETERS MOVING AWAY FROM THE LAST-CLICK ATTRIBUTION MODEL?

WHAT ATTRIBUTION METHOD DO YOU CURRENTLY USE?

34% First- and Last-Click

10% Linear

12% Time decay

22% Other

22% Custom

www.adrol l .com5

“From a practical perspective, the last-touch method is easy to understand and execute. In fact, it used to be the only attribution standard implemented in Google Analytics, in the beginning. While other methods exist and have been applied more recently (e.g., algorithmic multi-touchpoint models), not every CMO may have the skills in their team to run statistical models or the budget to buy a third-party solution.

And it is easy to see why some companies are hesitant to invest too much yet. Since the whole area revolving around attribution is fairly new, there is little evidence revealing the monetary benefits of using more sophisticated methods versus free tools (also, these days site-analytics providers offer more flexible allocation rules that consider all touchpoints, such as equal weighting of all channels). No surprise that clients ask themselves: are the efficiency gains worth the extra efforts and investment in advanced analytics?” —Nico Neumann, Senior Research Analyst, University of South Australia6

THE FIVE HURDLES:The AdRoll and AdNews led survey revealed that there are five common hurdles facing marketers that are seeking to move away from the first- and last-click attribution model.

1. If you can’t have the Holy Grail, then it’s easier to stick with the old formula.

The general consensus among marketers is that until we can link online and offline attribution—the Holy Grail—then first- and last-

click model is better than nothing at all. Of course, Facebook7, Google8 and every other data firm and retailer in between are currently in the process of closing that loop.

One marketer from a major Australian financial institution told us, “Everyone is talking about path to purchase, but really it’s just behaviours and touchpoints. Attribution is just choosing a metric, whether that’s dwell time or something else. We pick a metric because a click is so much more than an impression, but how do you trace that? And let’s face it, the Holy Grail is linking online-offline, so that we know that the person sitting there one night watching your TV ad is searching for your product the next day.”

2. Fear factor: What happens if we spend the money, switch models and gain nothing from it?

One leading ANZ marketer summed this hurdle up neatly: “I think there’s a little bit of a fear factor because if we change things, it has

ramifications. If I change something further on along down the line and I actually take myself away from last-click, there are implications for that, so I’m going to have to be really sure about it.”

3. Lack of Resources

Without a shadow of a doubt, every marketer we spoke with intonated that a switch from one model to another would require resources that simply aren’t available to them.

One marketer from a leading Australian travel brand said: “A lack of resources is the problem; it is difficult to build a business case to justify more resources without

WHAT ARE THE CHALLENGES PREVENTING ANZ MARKETERS FROM MOVING AWAY FROM A FIRST- AND LAST-CLICK MODEL?

Attribution Modelling: A Discussion Paper

www.adrol l .com6

Marketers and Attribution

providing data-driven insights that show that a change in our attribution model will result in spending our marketing budgets more efficiently.”

4. Drowning in Data

Issues involving data were cited as a common problem facing those wishing to make a switch: lack of data, too much data, the need for a data partner and not enough budget to hire a data partner were all

listed as barriers to change.

It was agreed that capturing the right data is the only way to derive better results from attribution modelling, and this poses a big challenge. There was an implicit understanding that agency data, media data and sales data need to be merged into a comprehensive and understandable format, which respondents felt was far from easy to achieve.

One marketer from a computer hardware brand told us: “It can be quite expensive just to set up: the tracking itself, the extraction and then the manipulation and the analysis. I’m not sure if there are any providers out there who can do all of these things in an expert way.”

5. C-Suite Says No

The majority of the 34% stuck using the first- and last-click model said that they were unable to move forward because of a lack of understanding from C-suite. The marketing team, a digital marketer

or analyst is more often than not responsible for attribution modelling (56% according to our survey), so the message rarely travels up the chain to C-suite about the underperforming nature of first and last click.

One marketer said: “As a marketer, you will instinctively have an understanding of how attribution affects your business, but the people above you will want proof. The business requires proof at all levels and therefore it becomes really hard to sell because you also don’t always have all the answers. But you still have to try. Raising investment even for a trial isn’t easy.”

Another ANZ marketer told us: “C-suite thinks it is simple and that a switch can just be flicked. They just ‘want it done’ but some people don’t understand why doing it the right way is important rather than just putting a solution in place. Many ad-tech vendors do simple sales pitches that don’t adequately address the fundamental data management needs of organisations. There is no cookie-cutter approach.”

According to Deloitte: “Attribution is about more than accountability and effective allocation of marketing resources. In order to realize this value, the digital attribution endeavour requires alignment and integration across a number of functions to be successful: the CMO and marketing team, the CIO and IT team, the CFO and finance team, the CEO and third-party vendors.” 9

www.adrol l .com7

Marketers and Attribution

The good news is that 66% of ANZ marketers have moved away from first- and last-click attribution, with the majority employing a custom solution. But even though the market is headed in the right direction there are still factors which inhibit the sector from reaching optimum performance levels. Our research indicates that there are a number of factors to consider when adopting an attribution model and that this will continue to change as the market evolves over time.

Tweaking as your campaigns change and grow

67.7% of those who took part in the AdRoll/AdNews survey stated that they change their attribution model more frequently than every 12 weeks. However, as buying habits change online, it’s recognised that an attribution model should be tweaked as campaigns progress instead of at a predetermined interval. Obviously there are budget limitations that will prevent or delay making changes to the current attribution model, and this was certainly recognised by survey respondents.

One of the marketers we surveyed stated, “Multi-channel attribution modelling and analysis is not a one-time effort, it is something you’ll do all the time. Not every day, but at least do an operational review every two weeks and a strategic review (with recommendation for changes) every month.”

The evolving nature of consumer behaviour online

As this goes to press, the latest campaign to catch the lustful gaze of marketers is the collaboration between Burberry and Snapchat.

Burberry created a Snapchat ‘sponsored’ Discover channel to promote its latest fragrance, Mr. Burberry. It has received high praise from many quarters.10

Similarly, Subway has been causing a storm in Australia with its video content ads starring major influencers.

These are both fascinating plays which move way beyond digital display advertisingand will no doubt have a huge effect on the placement and effectiveness of corresponding digital campaigns.

So the question is how will attribution fit into this brave new world? Indeed, marketers will have to learn how to build their attribution models—and tweak them—according to new marketing practices and new emerging channels, like Snapchat.

Many marketers involved in the roundtable discussions which led to the creation of this ‘discussion paper’ said that the oncoming arrival of Beacons serves to highlight one area which will call for a dramatic improvement in attribution modelling. The data collected from Beacons will provide a key step in building the offline-online Holy Grail of attribution.

Media Mix —the role of agencies

The number of replies to the survey were evenly split between those working client-side and those who work at media agencies and/or tech vendors.

Agencies clearly play an interesting role where attribution is concerned. First and foremost, they have a clear vision as to how all media channels come together and how spikes in one area (for example, a free-to-air TV campaign) can have a wider impact on where money is spent online.

Agencies also have the resources to collect a lot of data, which begs the question of whether a brand wants the data which drives its attribution decisions sitting with a third party. There is a lot of debate over the handling of this data, with many brands admitting they are unable to budget for the collection, housing and analysis of data, but that they also sit on a lot of proprietary data which could be overlaid with the data an agency collects. A lot of brands are calling for the more efficient swap of data between the two parties.

According to the Adroll/Adnews survey, 41.4% of brands are active in attribution modelling, 21.9% collect data from attribution modelling and pass it on to the brands, while 36.9% have no involvement whatsoever. Perhaps the brand has requested it stay that way, but some have called for more open dialogue between the two parties.

As the importance of attribution grows, agencies are in a good position to take advantage of it by developing expertise to provide a useful resource to brands.

WHERE DOES ATTRIBUTION MODELLING GET COMPLICATED, AND HOW WILL FUTURE TECHNOLOGIES IMPACT YOUR MODEL?

www.adrol l .com8

Marketers and Attribution

There is opportunity for agencies, tech vendors and brands to come together to openly talk about improvements that would help foster the evoltuion of attribution models; from the need for more education on attribution to more engagement from c-suite to more investment in the data that backs attribution modelling decisions.

The market demands further exploration—22% of those surveyed declined to comment on which attribution model they are using, while 22% are using a custom model the details of which have not been revealed.

Regardless of the fact that every brand is entitled to a unique attribution model, there are still industry-wide learnings that could be determined and applied for the greater good of the overall market.

How to create a seamless experience between online and offline also came up as a constant sticking point for most brands, as did how to successfully track the flow from one to the other. One marketer said that a brand should know its online presence as well as its offline, but there was a general feeling that technology in Australia for the time being is some way off helping to break down this barrier.

There was a unanimous feeling also that creatives should be supplied with better briefs to create campaigns which can be targeted more effectively based on careful analysis of data and closer attention to attribution. There was also a sense that the talent pool in Australia for these types of people is small and needs to be expanded.

Lastly, there was an acceptance that first and last click may be the simplest route and that change requires investment and understanding from those further up the chain. It was also noted too that a new breed is taking the helm of marketing in ANZ and this will only serve the attribution-modelling discipline better.

And all who were involved in the AdRoll/AdNews research were convinced of one point: marketing attribution demands serious, urgent attention.

CONCLUSION —WHERE DOES THE CONVERSATION NEED TO GO?

REFERENCES1. AdRoll ANZ State of the Industry 2016

2. http://www.kaushik.net/avinash/multi-channel-attribution-modeling-good-bad-ugly-models/

3. https://www.comscore.com/Insights/Press-Releases/2009/10/comScore-and-Starcom-USA-Release-Updated-Natural-Born-Clickers-Study-Showing-50-Percent-Drop-in-Number-of-U.S.-Internet-Users-Who-Click-on-Display-Ads

4. http://www.smartinsights.com/internet-advertising/internet-advertising-analytics/display-advertising-clickthrough-rates/

5. http://www.adnews.com.au/opinion/why-are-last-touch-models-still-in-place

6. https://www.facebook.com/business/help/298717656925097

7. https://support.google.com/adwords/answer/2998031?hl=en-AU

8. http://cmo.deloitte.com/wp-content/uploads/2015/07/Digital-Marketing-Attribution-Myth-vs.-Reality.pdf

9. http://www.marketingmagazine.co.uk/article/1390179/why-burberrys-sponsored-snapchat-discover-channel-marketing-brilliance

10. http://mumbrella.com.au/eat-well-go-subway-repositions-and-launches-australias-biggest-ever-influencer-campaign-313807