Blog Post:Given the massive amounts being spent in the digital space and the expectations that marketers deliver bigger, better results than they did the day before, understanding what, specifically, moves the needle is key. And it’s those attribution models that help us paint the picture: where meaningful touches are taking place, how users are being exposed to your message across all channels and, overall, the optimal budget allocations that should take place within those channels. That’s why, now more than ever before, having a successful attribution model in place is key. In the past, we may have evaluated channel by channel with individual attribution models and experiences that were siloed from one another. However, consumers now interact with our brands across every possible platform; and to be effective, efficient and deliver relevant journeys, we need a holistic view that looks at each touchpoint through a cross-channel lens. In other words, if this consumer were influenced by multiple sources, shouldn’t every touchpoint get credit for the win? Think of it as the assist and the goal. Sure, one player scored, but without the perfect pass at the perfect time, would that ball have found its way into the net? Maybe — but maybe not. Understanding the Models: Rules-Based vs. Algorithmic When it comes to defining your brand’s attribution model, though, there are more considerations than just whether to have or to have not. Generally speaking, there are two categories of models: rules-based and algorithmic. A rules-based model is humanly defined, and as a result, fairly subjective. These models are based on easily understood assumptions we have such as first and last touch, equal touch and other clear-cut scenarios. For example, it was the last exposure that drove the sale or the first and last equally. Though easy to understand and interpret, the risk here is that perception is reality in rules-based modeling. Again, this is a highly subjective process, and if your perceptions are off, your entire attribution model could go out the window. At the same time, however, if your organization is just getting started with attribution modeling or is lower on the analytics-maturity scale, this more descriptive, lower consideration process could get the job done, at least for now. As your foundation grows, it’s easy to move from single-touchpoint attribution — last interaction and first interaction — to multi-touchpoint models that assign credit to various touchpoints based on existing rules. Adobe Analytics, for example, has seven standardized multi-touchpoint attribution models available. Moving further up the maturity scale, there’s the data-driven algorithmic approach. Here, attribution outputs are predicated based on data and the modeling of that data. Fractional attribution is assigned depending on one touch’s value relative to the others. To be successful, though, algorithmic attribution relies heavily on the richness of the incoming data. If the data is solid, you’re getting a comprehensive snapshot, but if it’s not, your results could be severely flawed. Consequently, this model requires a high degree of human interaction, despite its being rooted in machine learning. No matter how powerful your model, you’ll need context from a human analyst. This will also help curb the potential for flawed inputs, as it’s something that human marketers can likely spot quickly before they taint the outputs. The upside to algorithmic-attribution models? While there’s a bit more resource allocation and some heavy lifting, at least in the beginning, there’s no guess and check anymore. The attributions assigned give you a comprehensive look at the impact every interaction has on the end goal, helping you better understand what’s working, what’s not, and what influences what in this big, cross-channel puzzle. And that understanding has limitless potential for marketers and brands. Determining the Right Model for Your Organization So which to integrate? There’s no universally right answer, but instead, a series of benefits and considerations to weigh. The simple solution, of course, is rules-based attribution — and most common within that modeling is last click or last touch. It’s straightforward and completely clear cut: the last touchpoint gets the attribution. It was the final tipping point and the moment in time that made your consumer convert. Makes sense. Then again, there’s everything that happened upstream to consider: that initial email that got him thinking, the retargeting that reminded him repeatedly, and the exposures that happened along the way. Sure, he clicked on the display ad, but it was only after kicking around the idea for a few minutes, hours, days, or weeks — whether he realized it or not. Without those touches, would he have clicked through to buy the sneakers, book the plane ticket, or watch the video? Like the soccer goal, this one’s a maybe/maybe not. Algorithmic-attribution modeling digs into those experiences from start to finish, helping paint a better, more accurate picture while giving you a much richer view of media waste — a win/win even amidst the heavier lifting and greater resource allocation. So Which to Choose? I usually recommend that marketers ask themselves three simple questions to best steer the conversation:
  1. How much data do I have, and equally important, how many different marketing channels does that data extend across?
  2. Am I ingesting all of the data into the same platform?
  3. Do you have a persistent customer identifier?
If the answer to question number two is no — if you aren’t ingesting all of the data into the same platform — then it’s much easier to use a rules-based model. Having all the data sources on the same platform is a prerequisite for benefiting from a cross-channel algorithmic-attribution model.  If the answer to question three is yes — they can use it as the hub of the spokes between all marketing channel data. Imagine disparate data sources from all marketing channels, coming together as spokes on a wheel, finally connecting at the hub to be stitched together with the use of a persistent customer identifier.  Brands can more easily take action on stitching their different marketing channel data sources together to begin looking at paths to success events with consideration to user exposure across multiple channels.   However, on the route you opt for, there’s one more critical piece of the conversation: ensuring a single source of truth (SSOT). If some people within your organization are using rules-based attribution models and assigning credit to the last touch every time while others are using an algorithmic approach, you’ll wind up with multiple sources of truth, and with them, total analysis paralysis. Brands need to be aware that the attribution model they choose and communicate to their marketing channels will significantly influence the behavior and the tactics employed by those managing the delivery of their marketing activities. If you incentivize someone to achieve a flawed goal, you are essentially incentivizing them to behave poorly to win favorability against a poorly chosen attribution model. If brands identify a flawed model, the behavior in terms of how your marketing is managed will be flawed as a result. Think of your attribution model as the blueprint for incentivizing the behavior you want. So choosing a well-designed blueprint is paramount. While it seems daunting to stand up a rich attribution model, it’s worth the investment. It could be the difference in achieving sophisticated and data-driven marketing spend. Whose Goal is it, Really? Today’s consumer is far from one dimensional. She’s moving through countless experiences, platforms, extensions and influential touchpoints every day, all of which inform her journey toward your brand. Understanding attribution is a critical piece of optimizing your marketing efforts and best allocating your resources. However, given the cross-channel nature of today’s consumer, assigning attribution to the last destination will likely steer your future efforts in the wrong direction. It’s the soccer analogy: sure, sometimes you take the ball downfield and score with one swift kick. But more often than not, it’s a team effort that entails smart, strategic passing and crossing, and finally, a hard-earned goal. In that case, whose win is it, really? Being able to answer that question will catapult your marketing efforts ahead in a big way, ensuring you can best allocate resources, eliminate media waste and better architect meaningful journeys for your audience — now and in the future.
Author: Date Created:January 25, 2016 Date Published: Headline:Which Type of Attribution Model is Right for My Brand: Rules-Based or Algorithmic? Social Counts: Keywords: Publisher:Adobe Image:https://blogs.adobe.com/digitalmarketing/wp-content/uploads/2016/01/AdobeStock_66115658-e1453516319357.jpeg

Given the massive amounts being spent in the digital space and the expectations that marketers deliver bigger, better results than they did the day before, understanding what, specifically, moves the needle is key. And it’s those attribution models that help us paint the picture: where meaningful touches are taking place, how users are being exposed to your message across all channels and, overall, the optimal budget allocations that should take place within those channels.

That’s why, now more than ever before, having a successful attribution model in place is key. In the past, we may have evaluated channel by channel with individual attribution models and experiences that were siloed from one another. However, consumers now interact with our brands across every possible platform; and to be effective, efficient and deliver relevant journeys, we need a holistic view that looks at each touchpoint through a cross-channel lens. In other words, if this consumer were influenced by multiple sources, shouldn’t every touchpoint get credit for the win? Think of it as the assist and the goal. Sure, one player scored, but without the perfect pass at the perfect time, would that ball have found its way into the net? Maybe — but maybe not.

Understanding the Models: Rules-Based vs. Algorithmic

When it comes to defining your brand’s attribution model, though, there are more considerations than just whether to have or to have not. Generally speaking, there are two categories of models: rules-based and algorithmic. A rules-based model is humanly defined, and as a result, fairly subjective. These models are based on easily understood assumptions we have such as first and last touch, equal touch and other clear-cut scenarios. For example, it was the last exposure that drove the sale or the first and last equally.

Though easy to understand and interpret, the risk here is that perception is reality in rules-based modeling. Again, this is a highly subjective process, and if your perceptions are off, your entire attribution model could go out the window. At the same time, however, if your organization is just getting started with attribution modeling or is lower on the analytics-maturity scale, this more descriptive, lower consideration process could get the job done, at least for now. As your foundation grows, it’s easy to move from single-touchpoint attribution — last interaction and first interaction — to multi-touchpoint models that assign credit to various touchpoints based on existing rules. Adobe Analytics, for example, has seven standardized multi-touchpoint attribution models available.

Moving further up the maturity scale, there’s the data-driven algorithmic approach. Here, attribution outputs are predicated based on data and the modeling of that data. Fractional attribution is assigned depending on one touch’s value relative to the others. To be successful, though, algorithmic attribution relies heavily on the richness of the incoming data. If the data is solid, you’re getting a comprehensive snapshot, but if it’s not, your results could be severely flawed. Consequently, this model requires a high degree of human interaction, despite its being rooted in machine learning. No matter how powerful your model, you’ll need context from a human analyst. This will also help curb the potential for flawed inputs, as it’s something that human marketers can likely spot quickly before they taint the outputs.

The upside to algorithmic-attribution models? While there’s a bit more resource allocation and some heavy lifting, at least in the beginning, there’s no guess and check anymore. The attributions assigned give you a comprehensive look at the impact every interaction has on the end goal, helping you better understand what’s working, what’s not, and what influences what in this big, cross-channel puzzle. And that understanding has limitless potential for marketers and brands.

Determining the Right Model for Your Organization

So which to integrate? There’s no universally right answer, but instead, a series of benefits and considerations to weigh. The simple solution, of course, is rules-based attribution — and most common within that modeling is last click or last touch. It’s straightforward and completely clear cut: the last touchpoint gets the attribution. It was the final tipping point and the moment in time that made your consumer convert. Makes sense.

Then again, there’s everything that happened upstream to consider: that initial email that got him thinking, the retargeting that reminded him repeatedly, and the exposures that happened along the way. Sure, he clicked on the display ad, but it was only after kicking around the idea for a few minutes, hours, days, or weeks — whether he realized it or not. Without those touches, would he have clicked through to buy the sneakers, book the plane ticket, or watch the video? Like the soccer goal, this one’s a maybe/maybe not. Algorithmic-attribution modeling digs into those experiences from start to finish, helping paint a better, more accurate picture while giving you a much richer view of media waste — a win/win even amidst the heavier lifting and greater resource allocation.

So Which to Choose?

I usually recommend that marketers ask themselves three simple questions to best steer the conversation:

  1. How much data do I have, and equally important, how many different marketing channels does that data extend across?
  2. Am I ingesting all of the data into the same platform?
  3. Do you have a persistent customer identifier?

If the answer to question number two is no — if you aren’t ingesting all of the data into the same platform — then it’s much easier to use a rules-based model. Having all the data sources on the same platform is a prerequisite for benefiting from a cross-channel algorithmic-attribution model.  If the answer to question three is yes — they can use it as the hub of the spokes between all marketing channel data. Imagine disparate data sources from all marketing channels, coming together as spokes on a wheel, finally connecting at the hub to be stitched together with the use of a persistent customer identifier.  Brands can more easily take action on stitching their different marketing channel data sources together to begin looking at paths to success events with consideration to user exposure across multiple channels.  

However, on the route you opt for, there’s one more critical piece of the conversation: ensuring a single source of truth (SSOT). If some people within your organization are using rules-based attribution models and assigning credit to the last touch every time while others are using an algorithmic approach, you’ll wind up with multiple sources of truth, and with them, total analysis paralysis.

Brands need to be aware that the attribution model they choose and communicate to their marketing channels will significantly influence the behavior and the tactics employed by those managing the delivery of their marketing activities. If you incentivize someone to achieve a flawed goal, you are essentially incentivizing them to behave poorly to win favorability against a poorly chosen attribution model. If brands identify a flawed model, the behavior in terms of how your marketing is managed will be flawed as a result. Think of your attribution model as the blueprint for incentivizing the behavior you want. So choosing a well-designed blueprint is paramount. While it seems daunting to stand up a rich attribution model, it’s worth the investment. It could be the difference in achieving sophisticated and data-driven marketing spend.

Whose Goal is it, Really?

Today’s consumer is far from one dimensional. She’s moving through countless experiences, platforms, extensions and influential touchpoints every day, all of which inform her journey toward your brand. Understanding attribution is a critical piece of optimizing your marketing efforts and best allocating your resources. However, given the cross-channel nature of today’s consumer, assigning attribution to the last destination will likely steer your future efforts in the wrong direction. It’s the soccer analogy: sure, sometimes you take the ball downfield and score with one swift kick. But more often than not, it’s a team effort that entails smart, strategic passing and crossing, and finally, a hard-earned goal. In that case, whose win is it, really? Being able to answer that question will catapult your marketing efforts ahead in a big way, ensuring you can best allocate resources, eliminate media waste and better architect meaningful journeys for your audience — now and in the future.