The 6 Requirements of Holistic Attribution & Actionable Insights

July 16, 2020

In this session we'll dive into the 6 requirements for generating actionable insights that drive effective consumer experiences, optimally allocating touchpoints across the consumer journey, and increasing overall ROI.

  1. Data Breadth - Understand the Entire Consumer Journey.
  2. Data Depth - Deliver the Right Content.
  3. Combine Channels and Messaging Analysis.
  4. Create Data Standards and Governance.
  5. Improve Customer Satisfaction, Value, and Costs.
  6. Analyze and Predict with Accuracy.

Mikel Chertudi (00:09): 
We'd like to welcome you to this next session of our Marketing Attribution Symposium. My name is Mikel Chertudi, and we are going to be discussing several important topics regarding performance measurement, actionable insights, and attribution. This is a topic that has been something that has been a challenge over the course of my two decade long career in marketing and digital and traditional media. And we are going to discuss, how can we better solve this problem? This challenge of measurement. My name is Mikel Chertudi. I am the General Manager of the Strala business and Chief Operating Officer here at ObservePoint. Just a little bit about my background, I have helped to lead and oversee large marketing organizations at large Fortune 500 organizations, and measurement—knowing where to invest a team's time, resources, dollars—has been a challenge over the last, you know, 20, 30 years. 

Mikel Chertudi (01:28): 
And as we attempt to improve customer experiences, our brand affinity, and ensure we're investing in the right areas to acquire customers, to engage existing customers, and drive loyalty and drive business and return on our investments. This is something that in this presentation, we hope to share six different elements with you that will help get this right. So there are going to be two components to our presentation today. The first is a brief segment on some research where we have asked leading CMOs VPs of marketing, heads of digital, heads of media, heads of marketing analytics, customer analytics, and marketing operations professionals—what are the challenges associated with performance measurement and how can we better improve this? Also, over the course of my career and working with many colleagues in the industry, we've created six core requirements that enable us to solve this problem. And here at ObservePoint recently we've acquired a technology called Strala and that platform enables us to do so. But what we've done in this session is we've abstracted out the requirements used in the platform from the technology itself. So we are going to be sharing with you the six requirements that have been built into the Strala platform, but that all of you can use on your own to improve this very critical area of measurement and insights and attribution. So we'll jump right in. 

Mikel Chertudi (03:20): 
So, this age old question that marketing has, where do we invest our time, our resources, our team's energy and efforts to improve growth, to improve our return on investment. And this quote here from Peter Drucker is still very applicable today. If you can't measure it, you can't improve it. And what we find in the industry is that this was a, some research that was done several years ago is that less than 11% of brands feel confident in the accuracy of their performance measurement and attribution. And what we found is, we've done some updated research here and we see that this is ticking up a little bit. And as we look at this research, now this was done by an independent research firm, just this last month. And we asked leading brands enterprises over $1 billion in sales, how they are feeling about their ability to accomplish attribution and to conduct performance measurement. 

Mikel Chertudi (04:29): 
And it's this question of why is getting measurement right so challenging? And one of the things that we've found is that still over half of us in the industry are left to our own devices and trying to do it ourselves by consolidating data. But also if we look over here, this 18% that we add to this 52%, what it shows is that almost 70% of us are still trying to do this ourselves. And what we are seeing is that we're still trying to measure the performance of our marketing in individual siloed channels. And we're relying on some of the siloed tools. So if you look at this 52% where we're trying to consolidate the data and the 18% where we're not even trying to consolidate the data. 70% of us are still trying to do this ourselves. What we see is that less than a quarter, 24%, are using a third party solution to try and accomplish this. 

Mikel Chertudi (05:33): 
And then what we've seen is that there are less and less of us who are relying on web analytics to try and do performance measurement as well as attribution. And what we've seen is that over the last 20 years, organizations try to do this in web analytics, but quite unsuccessfully. And there's one big reason for this. And that is that we have found that our experiences with our consumers are not just digital only, they're digital plus physical. They are digital plus traditional and physical events. And so trying to shoehorn physical measurement into web analytics is something that just does not work well. And so, as a result of that, we have to think more broadly about the tool sets that we're using. So we see that the large enterprises have found out that web analytics is not the solution for performance measurement because of the digital plus the physical. 

Mikel Chertudi (06:36): 
And so if we go to the next area of research, what we find is that that 11% of not feeling high confidence has improved slightly. And we see that only 17% have high confidence in the level of completeness of our accuracy for attribution and performance measurement. And we still see that the large, vast majority of us, 65% have partial confidence in the completeness and the accuracy of our performance measurement and that, you know, still 18% of us have very low confidence. And if we correlate this data to the previous slide, what we see is that there's a large contingent of us who are trying to consolidate the data, or who are trying to rely on those individual channels. It's almost perfectly correlated that 70% of us who are doing this ourselves and only having partial confidence in our accuracy and in our approach. 

Mikel Chertudi (07:44): 
So there are better ways, and that's what we want to dive into and share with you today. Now one last piece of research here, what marketers need versus what we have. So what we've seen is that 82% of leading marketing organization believe it is critical to merge customer data across devices. That's including mobile tablet, computer IOT networks, one half of enterprises state that the tools leave data gaps and don't provide meaningful insights for those of us who are still using enterprise solutions. And we're turning to a solution. We're still finding that these tools are leaving gaps. And there's some reasons for that. What we see is that still with the tool sets that are out there for the quarter of us that are using these tools, we still are not able to merge data across these experiences, whether it's mobile or tablet or screens, for desktops, for example, and we're not even involving our IOT that are starting to become more prevalent. 

Mikel Chertudi (08:51): 
And we're not looking at that offline world with phones, in our stores. Print is still being left out, as well as the shopping experience in retail environments, whether that's at a bank, at a store. And across all of these networks, we are not doing this well. And as a result of that, we know it's critical to merge these multiple devices. We know it's critical to merge these multiple experiences across digital and physical, but the tool sets just aren't enabling us to do that well. And the last thing, and we're going to be going into a deeper dive on this, 94% of leading marketing organizations state that's setting up campaign IDs, the touchpoint metadata, that is just the crucial means to an end of getting good measurement, is a time consuming, soul crushing task. And so as a result of that, we have to do better. We need better ways of doing this. Many of us are still relegated to using, in the digital side at least, UTM parameters, spreadsheets, homegrown systems, to set up that touchpoint management and those campaign IDs. And as a result of that, it's just not working, especially when we try and apply that to the physical world. So let's jump into these requirements and what we can do to improve. 

Mikel Chertudi (10:14): 
Okay. First, this is where we are trying to solve an area where we are falling down in a big way. Number one, data breadth. We need to ensure that we are capturing every customer touchpoint across the customer journey. So there are three critical areas to getting measurement, right? Number one, understanding whether it's an anonymous or a known customer, and we need to pin that known customer identity or that anonymous customer identity and have the ability to merge those two, if they are not the same, across devices. And we need to be able to associate multiple touchpoints to that customer. So the first, in this example, might be paid search. They're looking for something, and then they may sign up and we send them an email. That's our second touchpoint Third, they may engage with us with a social post. Fourth, they may come back to our website and engage for a period of time. 

Mikel Chertudi (11:22): 
There, they may download our mobile app and engage in an experience there. Then they may attend a physical event. That's our sixth touchpoint, and then they may receive a call or actually call into our call center. That's a seventh touchpoint. What we find is that many of these touchpoints in the customer journey, for example, are not even being considered or captured. And as a result, our attribution is lacking. And so when we try and pin missing touchpoints, or just a partial set of the customer journey to these key outcomes and these conversion events, which is the third ingredient to successful measurement, what we find is that our attribution, our performance measurement, our ability to run simulation analysis and predictive capabilities just are not right. And a lot of companies ask us, they say, well, we've got artificial intelligence and that will cover the gaps. 

Mikel Chertudi (12:22): 
The problem is that artificial intelligence machine learning algorithms, all they are doing is looking at the patterns across these journeys. And if there are points of data that are missing, no missing data and no algorithm can make up for the missing datasets to help predict and enable us to know where we should be investing. So this is a big challenge. We have to ensure that every customer touchpoint, both digital and physical is being captured. And there are ways to do this. And we'll talk about that. Next, data depth. We all hear about channel attribution, but what about content attribution? We're always trying to decide, should we invest more in paid search, more in TV, more in social, more into our call centers or apps, chat events. And that is a struggle if we're not able to get this right, but what about all of our promotions? 

Mikel Chertudi (13:24): 
What about our videos? What about our content and our learning, our eBooks? What about, in this example here in the yellow, we're showing multiple dimensions. So the top dimension is our channel, but what about publishers? Most of us are pretty good at doing medium and source, but we're not so good at adding other critical areas of measurement to the touchpoint and content attributes. So in these yellow areas, we're looking at, for example, in a retail example, a Spring sale—a buy one, get one free, 20% off, home set up, free shipping. So if we're not able to consistently measure and pin and associate content attributes to our touchpoints, we're only getting half of the measurement and only half of the ability to optimize our outcomes. So this is critical as well, which then goes into the next area. And what we're highlighting here is what's missing. 

Mikel Chertudi (14:27): 
So let's assume that we've got multiple touchpoints across our consumer journey. So this might be one touchpoint and associated to that again, paid search is the channel, Google is the source or the publisher. We may have what type of promotion this is for content type. And then we've got all of our subsequent touchpoints. On average, we see that there are roughly 13 different touchpoints prior to a consumer transaction. If we extrapolate that to B2B sequences and scenarios, multiply that by the number of people that you typically see on your account, it could be 7 or 10. That means that we're seeing over a hundred touchpoints on an account and a B2B scenario. Think of all of this missing data that we're not even capturing, such as message theme, asset types, what products we're promoting, calls to action. So this becomes a missing part of the equation, and it's really limiting our ability from an attribution and measurement perspective. 

Mikel Chertudi (15:26): 
So when we are able to fill out all of these content attributes, our analysis becomes very powerful. This is the third area and requirement, combining channel and content analysis to optimize our consumer experiences, to improve growth, acquisition, ROI engagement. And when we have all of these, now we're able to do analysis. These touchpoints are the foundation and the building blocks, and the metadata, the classifications, the attribute values associated with them are the ways in which we slice and dice our reports. So if we've got channel here, we're able to do channel analysis. If we've got source or publishers and the touchpoints, we're now able to do publisher analysis and optimize across different publishers, including our own brand. If we've got content types, this is a B2B example, promotions, eBooks, video scripts, demos. Now we're able to do content analysis. And what about all of those missing content attributes that we're not even associating to our touchpoints? That's all of this missing analysis over here. And what we see is that our most sophisticated customers are using up to 30 plus attributes, not just medium, not just source, and it's enabling them to do much more in depth analysis and optimization across that experience. Okay. 

Mikel Chertudi (17:00): 
Let's talk about the fourth area. How do we get this right? Create data standards and governance. Now we know that that touchpoint system of record is critical. Again, some of us are using spreadsheets. We are using homegrown systems to capture UTM parameters, campaign IDs, and we need to do this for all of our systems, whether this is for display voice of customer analytics, offline paid search, paid social, our website, email, our CRM systems. We need one central repository. And not only should we just be using the somewhat over-simplistic standard set by UTM channel source, medium publisher, we need to be looking at those content attributes and we should be embedding those content attributes in that system of record. And so, as we think about how can we do that? One of the things that we've set up as part of our Strala platform is that we have this very simple enterprise approach to creating all of these systems where we can add additional attributes. 

Mikel Chertudi (18:12): 
So here is the all famous content or channel or medium. So here apps, direct mail, display, email events, we've got brief descriptions, campaign themes. These are things that we are not tracking well today. What about content types should this enable us to know, should we be investing more in videos? Should we be investing more in our different static assets, our eBooks, other elements here? So this becomes really critical for us to do this. Now, what about different stages of the customer journey like awareness, active use, retention? So this is really critical when we start adding all of these attributes. This is something that we advocate. Start adding these elements to your own systems of record, start using and adding additional attributes and values to your spreadsheets that you're using for tracking. This is critical. This will allow us to do this across all of our systems, across all of those customer touchpoints. 

Mikel Chertudi (19:18): 
We should be adding these capabilities, these IDs, to our call center scripts. We should be adding these to vanity URLs for print or TV. So this is really important. Okay, let's talk about the fifth requirement. 72% of the market is still stuck on last-touch attribution models. Now, if we take the prior stat that roughly 13 touchpoints occur prior to a conversion event or an outcome, we are indexing very limitedly in our ability to do good measurement. And if we are only focused on that last touchpoint and giving credit to that last touchpoint across all of our journeys, there is a whole world of touchpoints that we're not giving credit to. In one of my previous roles, I was responsible for a large marketing organization's media and digital marketing. And what we found is that we were using a last-touch model. 

Mikel Chertudi (20:28): 
And as soon as we went to a multi-touch model, whether it was a multi-touch linear, a time-decay model, even a machine-learning or AI-driven algorithmic model, what we saw was magical. We saw that by not over-indexing on that last touch and ensuring we were evaluating all the touchpoints, we saw over a 30% increase in our return on investment and our growth with the same investment on a year over-year investment in our marketing programs, our channels, our media. So just by getting the model right, and applying multiple models to answer different questions, we were able to ensure that we were looking at this much more holistically and much more accurately. This is critical. And there are many of us who are attempting to bring in PhDs and data scientists to do this. We should keep investing here. Now again, when we are applying models, what we are doing is taking five, ten, a thousand, ten thousand, a hundred thousand, millions of consumer journeys with all of these touchpoints. 

Mikel Chertudi (21:51): 
And we are trying to discover patterns for these touchpoints that lead to outcomes. Now, some of these don't lead to outcomes, and that's where some of our algorithmic models can pick this up and understand which of these are not leading to outcomes. So this is critical. Again, artificial intelligence cannot make up for missing touchpoints. And so this is very important. So as we think about applying multiple models, one of the things that's critical is that again, these models answer different questions here. I'm just highlighting something in our Strala platform capabilities by ObservePoint where it's important to look at multiple models. So we can look at an algorithmic model for content effectiveness by channel, we can apply Bayesian models, which is an algorithmic model, or a Markov chain for simulation. We can apply a last-touch and compare it to a time-decay model or a last-touch to a more sophisticated model. 

Mikel Chertudi (22:57): 
This helps us to answer different questions, different models have different strengths. So what we are showing here is, these are the models that we use within the ObservePoint Strala platform. Last-touch, first-touch, multi-touch, time-decay and again, these very critical machine learning and algorithmic models and their strengths. We shouldn't be using just an algorithmic model all the time. It doesn't answer all the questions. That's where we espouse using multiple models because they have different strengths and weaknesses associated to them. And so we'll leave this here as a leave behind, you can come back and look at some of the details within the different models and when to use, and when not to use them. 

Mikel Chertudi (23:46): 
Okay. Last but not least, the whole purpose of this is to optimize and take action based on the right insights. So what we're attempting to do is we see the level of insights tend to improve with the amount of data that we have. And what we're trying to do is use descriptive analysis, diagnostics, predictive, and prescriptive. And so what we've done is that we've created this catalog of insights and it's based on questions. And then what we do is we pair the insights to these different types of models and capabilities that enable us to answer the right question for the right element that we are trying to optimize for. So for example, we need to, in diagnostic questions, this is where for example, attribution comes in, what is my return on investment overall, and for specific communication channels and content such as print calls, email website, app, physical events. 

Mikel Chertudi (24:57): 
That is a question that can be answered by attribution models, whether it's a multi-touch time decay, multi-touch linear or last, to know what brought someone in last, or what brought them in first. Another question, which touchpoint combinations of channels and content have the greatest impact on driving more efficient consumer behavior outcomes, orders, sales, customer satisfaction. So again, this is an attribution question. It's looking back, and again, you can apply those attribution models. As we start getting more sophisticated into more of the predictive and prescriptive questions, we're going to use different types of analysis. For example, how do I increase ROI and overall outcomes, spanning channels and content. If I redeployed investments from less expensive to more expensive content formats, such as static website to instructional onboarding videos with the shift in consumer behavior, justify the expense and ROI, and then prescriptive questions. 

Mikel Chertudi (26:01): 
This is where we start to get into simulation or algorithmic models, which communication channels should I utilize to maximize consumer outcomes? What are the specific message themes I should be using with different consumer segments to drive specific outcomes such as sales or improved experiences? Where can I cut budget, but still generate the same results? And so these are very important to create this catalog of questions and align it to the previous set of elements and models. And so this is critical. So in summary, these six things should enable us to be much more robust in our measurement and ensure that we are getting performance measurement, attribution, and predictive capabilities right. To optimize our media and to generate much better consumer experiences for our customers. So thank you for attending. And if you have more questions, please let us know. We'd be happy to answer them and spend more time with you. Thank you.

Previous Video
Achieving Actionable Insights: Attribution Best Practices
Achieving Actionable Insights: Attribution Best Practices

The Search Discovery Panel discussed what they've learned about attribution best practices at the Marketing...

No More Videos