Play Ball: Why Data-Driven Attribution Is a Home Run


Share on LinkedIn

One of the best kept secrets in online marketing is that most campaign attribution data is completely wrong.” – Eric Peterson, CEO Web Analytics Demystified

There’s an old joke about marketing. A CEO looks across the boardroom table, frowns, and says, “I know that half the money I spend on marketing is wasted. The trouble is I don’t know which half.” This has been the unfortunate reality in marketing for decades—and still is the case in many companies. We know that marketing works. But oftentimes not how or why. And, almost certainly not which marketing investments are wasted or less effective. Before the digital and social era, it was almost impossible to understand with any certainty which marketing tactics were really moving the needle. As Yogi Berra once said, “You’ve got to be very careful if you don’t know where you are going, because you might not get there.”

Fortunately, times have changed, and CMOs now have the opportunity to more accurately measure, improve and predict the results from their marketing investments. Customers are rapidly adopting digital devices such as smartphones, tablets and wearables. Media is increasingly digitized. And, social media usage is growing across almost every consumer segment. The net result? We now have an unprecedented amount of data with which we can understand how marketing influences customer behavior across the full lifecycle of interactions with a company.

Why Data-Driven Attribution?

One of the hottest topics in marketing right now is data-driven attribution, the process of assigning credit to marketing events across multiple interaction channels. Most companies are in the early stages of accurately attributing marketing results to their marketing investments, and the practice of simply assigning all the credit to the last marketing touch is still common. However, these companies are missing out on an incredible opportunity. By some estimates, top performing marketing organizations are five times more likely to use advanced attribution models.

Simply put, data-driven attribution matters because it helps us develop marketing programs that cost less and deliver more. There are two primary reasons for this. The first reason is that consumer behavior has shifted dramatically in the past few years. McKinsey estimates that 56 percent of consumer journeys now include multi-channel, multi-device interactions. The traditional concept of the marketing funnel is now ancient history. It has been replaced by decision journey—the non-linear, multi-channel and consumer-driven path to purchase.

And, the second reason is quite simply that we now have the data and the capabilities to better understand how marketing drives purchase behavior. One of the wonderful things about digital is that it is an inherently measurable medium. As more advertising and marketing shifts into the digital world, we have more data that we can use to evaluate effectiveness. Marketers can now measure how campaigns perform for different customer segments and optimize their media buys and creative elements over time to improve effectiveness. So, we have more and more data, but what about capabilities? There’s good news on that front as well. Cloud computing, big data technologies, and growing ranks of data scientists are combining to deliver the data, visualizations, and algorithms required for data-driven attribution. These forces are also driving down the cost of these capabilities, and enabling even mid-sized and small companies to benefit.

How to Make It Happen?

Companies that succeed with data-driven attribution—and data-driven marketing in general—are those that take a systematic, disciplined approach to identifying all the touchpoints they have with prospects and customers and creating an integrated view. These companies tend to treat customer data and insight as a core competency and source of competitive advantage. They invest in the hard work to identify marketing channel data, integrate it, and marry it to their customer profiles.

Step 1 – Inventory sources of marketing and customer data

Once marketing organizations take a closer look at relevant data sources, many discover that, while they have volumes of data internally, much of the data they need has yet to come through the front door. It remains locked away with advertising agencies, web analytics vendors, or other third parties. The first step in building data-driven attribution capabilities is to inventory the data sources that are needed—including those that are already available and those that must be pulled into the company.

There are many potential data sources to consider. To help jump start your thinking, here’s a checklist of some of the most common examples:

  • CRM
  • email campaign
  • direct mail campaign
  • web and mobile clickstream
  • call center
  • paid search
  • organic search
  • affiliate marketing
  • social media
  • Internet video
  • Internet radio
  • display ads
  • mass media
  • stores / branches
  • marketing automation
  • marketing cost
  • customer/prospect demographics
  • customer/prospect behavior

Step 2 – Integrate data and build an event stream

Once you have a good handle on all of the potential data sources that are relevant to your data-driven attribution efforts, the next step is to begin to integrate some of that data. Start with data from your most important marketing channels and go from there. These sources can be consolidated using traditional data warehousing techniques; however, the large volume and unstructured nature of some of the data sources will benefit from big data technologies.

The data foundation for attribution is the marketing event stream. Essentially, this is a data structure that ties together all marketing interactions that occur with each individual. It requires detailed, atomic-level data from each marketing channel and then matching that data up with individual prospects and customers. In some cases, information about the customer or prospect is known and in some cases the prospect is completely anonymous. One of the core challenges is identity resolution—tying online and offline interactions to individuals who could be engaging across multiple channels and devices.

A marketing event stream may look something like the example below, where multiple interactions over multiple days occur prior to conversion.

Step 3 – Develop rules-based attribution models

The next step in the process is developing your first attribution models. The best starting point is simple, rules-based attribution. There are several examples of this illustrated in the diagram below. Last touch attribution means that we assign all of the credit for a conversion event to the last marketing touch that occurred. First touch is the opposite of that, all the credit goes to the first touch. And, even-weighting is assigning equal credit to all the marketing touches in the event stream. Each of these rules-based models is imperfect and incomplete; however, they are good places to start and develop a better understanding of conversion paths.

Step 4 – Test & learn; Develop algorithm-based attribution models

Evolving to a more sophisticated and more accurate approach to attribution often means developing an algorithm-based model. Many companies find that a test and learn approach helps refine a model that does a much better job of assigning credit. Test your hypothesis using a small percentage of your marketing budget and analyze the results. The cost and effort associated with this approach are challenging; however, the payoff can be huge and a source of competitive advantage.

 Step 5 – Ensure that insights lead to action 

The key to success with data-driven attribution is ensuring that insights lead to better decisions and actions. The process of integrating data and building attribution models is difficult, time-consuming and costly. And, all of that investment is wasted if insights never see the light of day. To avoid this situation, focus on people and process aspects of data-driven attribution, not only data and technology. Proactively think through the business processes that will need to change: everything from marketing strategy, to budget planning, to campaign optimization. How will data visualizations be distributed? Do you have the right people available who can review the data and craft a compelling business story that influences decision-making? What existing KPIs and measures will be disrupted? Who is likely to support this change? Who is likely to resist it?

Next Steps

Developing a fully optimized data-driven attribution capability can be difficult and costly, but getting started doesn’t have to be. Companies can often begin with data from their most important marketing channels and use simple conversion path analysis and rules-based attribution techniques. This is a great way to prove the value of data-driven attribution and assess the people, process and technology changes that are needed.

Another key consideration is whether to outsource this capability or to keep it in house. Many companies are recognizing that customer and marketing data is a source of competitive advantage. And, these companies are investing more to build out related capabilities as an internal core competency. However, other companies find that they can outsource to an attribution vendor who brings their marketing channel data together and produces insights with a more affordable approach. This is an important decision, and senior leaders should ensure that it aligns with their overall business strategy and competitive capabilities.

Data-driven attribution is an important part of the future of marketing. It can help make marketing more efficient, effective and customer-focused—it will allow you to guess less and know more. And who wouldn’t like that?

Republished with author's permission from original post.

Dave Birckhead
Dave is the Global Head of Marketing Technology at Spotify. He has worked with numerous Fortune 500 companies to bring about marketing technology solutions that optimize business performance, accelerate innovation and enhance marketing. You can find Dave on Twitter, LinkedIn and Google Plus.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here