Log In


How to Troubleshoot Multi-Touch Attribution

Posted August 21, 2023

When organizations implement multi-touch attribution, there are often questions to answer and issues to address. You might turn it on and then discover inconsistencies and all of a sudden your data is now seen as unreliable and the confidence you have built among other departments and stakeholders is now out the window. Don’t panic. Typically when you hear someone say the data is broken, it means they don’t trust the data. Still, most of the time, you are either missing something in your model or you’re over-counting something like email sends. It’s so easy to want to put all the touchpoints in there, but for something to actually get revenue association, there has to be some sort of meaningfulness to the touchpoint, some sort of inbound response.

It’s important to always bear in mind that attribution models are still that – models. A quote that is often bandied about is that all models are wrong, but some are useful. The model is that you’re applying the same algorithms to all your touchpoints and in doing so, you’re comparing and contrasting your touchpoints and figuring out which channels or tactics are performing well and which aren’t. 

Adjusting is really the power behind the attribution model, less so the literal numbers you’re seeing in the model. The big mistake folks make is not understanding that attribution is meant to be directional, i.e. everybody’s expectations are set in that clicking on this email isn’t exactly what made this $100 million opportunity. There were salespeople involved. There were other things happening behind the scenes. However, there was in fact the person that had that opportunity, did indeed click on that email. Therefore, it is still useful, but using many models in tandem as well allows each model to tell a different story. So the model(s) you choose are very important.

Mostly, it’s coming back to the educational piece and ensuring that teams understand how their attribution model is set up and starting with asking why they think this is wrong. Attribution is a really helpful in a wide lens view, but a lot of times, you’ve to look at the details to understand and provide that context of why you’re getting the numbers that you’re getting. Once you’re able to do that, the light bulb normally clicks on for many teams.

What to look for when troubleshooting attribution issues

There’s another facet of today’s topic, which is adopting attribution. There can be one or two champions leading the charge on an attribution project. Then at some point, they have to take this model and socialize it around the others in the organization. There could be stakeholders from marketing, sales, and finance along with others. 

In such circumstances, what could be the pitfalls that these champions encounter as they’re trying to gain cross-functional buy-in? A lot of times, you have your marketing ops team building it and they have a different viewpoint than your demand generation team, your sales team, or even your RevOps team. Many times, one can be built in a silo and you can try hard to have all strategies in place. But with an attribution model, you need to ensure that there are stakeholders from other teams, especially Finance and sales, providing input throughout the process. 

This way, when the team has completed it, it doesn’t feel like you’re starting from scratch by presenting and answering their questions and keeping a list of things that you didn’t think about. You tend to see people complete it and then they don’t use it because the other teams didn’t buy in. It might have been sold as though it’s going to be solely a marketing solution, still, you need to have everyone in the room and ensure everyone has a seat at the table.

Another big piece is when you’re getting down to the granularity of setting it up where it can make sense, but explaining it can prove to be very difficult and so you do need to have your data and documentation in place. To be able to explain, you have to have the teams that are explaining it be able to talk about it to all kinds of stakeholders at different levels so that it is successful.

Marketing practitioners vs. executives

A lot of dashboards and reports that you end up building using the tool are really meant for what we call exploratory analysis and they don’t immediately tell you what are the insights that you should take from it. Therefore, there is a step further that people often skip in taking exploratory analysis and turning it into explanatory assists and then insights. They don’t have that burden of cognitive load which you just show them the dashboard and they instantly start pointing to things they feel are wrong. That is part of it.

Another controversial point is that the original intention of attribution wasn’t necessarily to illustrate marketing’s impact to executive audiences. The intention was aimed at helping marketing professionals understand what is and is not driving pipeline and bookings, allowing them to optimize ROI and put budget towards the most impactful marketing activities. We often see marketers going to executives with four, five, or six different models trying to walk them through what each model is depicting and ultimately, executives don’t care. All they want to know is how much better you did than last month, last quarter, or what is your total year.

How to look at key deals

Before we start examining the models, let’s look at marketing influence vs. sourced marketing influence. When you get a lot of these NTA platforms that aren’t only tracking the engagement curve, but any subsequent lifecycle stages that may be impacted by that engagement, you get great insights into what marketing influence looks like over time. And when you start showing the number, it is more influential vs. just showing source, since you can have cool things like the percentage of the total pipeline and revenue that was influenced month over month, quarter after quarter, etc. 

It is important to look at key deals that were either won or lost and back in and ask “who were all the people involved?”, “what kind of things did they engage with?”. This allows us to see if there are patterns, especially among those significant deals, that we can either repeat or improve upon. That is often enlightening when you look at an important deal and you see they attended this small event that we put on and they downloaded these pieces of content vs. this other opportunity that didn’t have as much engagement. And even just looking at won vs. lost deals and the number of touches and seeing that pattern of the more often we touch folks and hopefully it shows in your data, the more the likelihood they’re going to convert.

A mistake that is made often is not understanding that attribution is a backward-looking metric and it’s not designed to tell you where to allocate your budget. Attribution vendors are striving towards looking forward  on roadmaps now with the advent of generative AI and machine learning.

What touches are missing? Why are we skeptical of the numbers? Dissect your buyer journey. Get buy-in from other groups. Every model is different. Everybody’s got a different use case. It just depends on how the business is structured.

And do you even have the right infrastructure in place to capture those touchpoints? Are you tracking calls? Are you on Outreach or Salesforce? What is tracking email responses and then how do you get those touches into the model? There is an art and science involved and it is very much an organization-by-organization thing. Sometimes, there’s some skepticism from the other parts of the revenue organization. And some organizations are chomping at the bit and want to see which sales cadence is outperforming another sales cadence. Maybe there is a cadence going to stage two in the funnel and they want to dissect that. So there is some utility in thinking holistically about the touches. But perhaps starting small with marketing and it is an iterative process bringing some of those other organizations.

Understanding UTMs and its role in attribution

Implementing attribution provides an opportunity to train the different teams on what UTMs are and their importance. Ultimately, putting together attribution platforms actually shows discrepancies in the way that different teams are using UTMs. And all of us agree that bad data in is bad data out. UTMs are  web bases that are used to understand how someone is coming to your organization, how someone’s coming to your domain specifically. 

With attribution, you’re trying to fill in the gaps of both your online and offline engagement and UTMs are a huge part of that online piece of the picture.  With attribution, you want to understand what is driving the most engagement with your website. You want to understand where everybody’s coming from and be able to group that information together for easier reporting. With the UTM strategy, many MOPs folks in particular, who are the ones setting up the attribution platforms, don’t really understand how UTMs are used and their importance.

The biggest piece of that is data quality, being able to group those engagements together, and understanding the difference between a medium, a source, and a campaign and how those pieces fit together. For example, you could have a team that’s using Google Adwords. You could have someone with the value of CPC vs. PCP and understanding why those two values are different, what their meaning is, and having that organization across all of your teams so that when those values come in on your engagement, you’re able to understand where that is coming from.

And it is also easier for your analytics team to put together dashboards and visualizations on those values. Therefore understanding UTM is understanding best practices, and how it is used. UTM is really important for marketing ops teams and is probably one of the biggest things that I see which isn’t well understood when setting up attribution platforms.

A lot of times connecting top of the funnel to bottom of the funnel can be difficult because many digital marketers are focused on the top of the funnel. And with tracking UTMs and having that go all the way into Salesforce, tracking your campaign members and being able to connect that to opportunities and drawing insights, not just how many views your ad got, but how many warm fills you got. Seeing whether that actually drove the needle.

For example, this company had an asset on our website that was very popular every year and pretty much of the entire database was downloaded every year. It looked like the highest-performing campaign, but it was a volume. Yes, high performing in terms of actual influence, but not as high as something that had a little less volume, but more impactful. So understanding that nuance. Going back to the point about exploratory vs. explanatory analysis, curating the insights for that group and drawing out the patterns you’re seeing can help them connect the dots a bit more than just letting them go wild on a dashboard.

The two disciplines are much more converged in the business-to-consumer model. If you’re advertising or selling shoes, then you are going to be compensated if they convert. As a digital marketer, we know that in B2B, there is a much longer sales cycle. So those top-of-the-funnel folks are focused on getting the clicks, the form fills, or whatever metric they’re using. However, what attribution allows you to do is start looking all the way back to the top-of-the-funnel activities via these UTMs. 

But in B2B it’s no longer buying a pair of shoes but purchasing a six or seven-figure software that takes multiple buyers, clicking on many different ads and offers. Therefore, if you are in marketing ops and you just use Marketo all day and don’t talk to the digital team, start making friends with them to understand UTMs because it will make your attribution project more successful when you implement it.

Marketing operations and data warehouses

Whether you’re hunting for a job right now or not, everybody sees that the Director of Marketing Ops who also runs the website does analytics and also manages the CRM. Even though this profession has been around for some time now, it’s still relatively young and many organizations take advantage of it. The big thing is getting out there and learning something. If you don’t have the budget to hire an analyst, scientist, engineer, or an army of MOPs people, bolster your resume, go talk to chatGPT for thirty minutes more often.

The concept of multi-touch attribution is well-defined. The problem lies in tying it back to the business in a way that makes the most sense for what we are trying to do as a marketing organization. What are our goals? And that is where the complexity lies. You don’t need to be a 20-year veteran and a Sequel engineer or a data scientist to create Salesforce dashboards to look at marketing influence. Just the curiosity and drive to be a lifelong learner is going to be your biggest ally in MOPs. 

Marketing ops, in general, becomes the jack of all trades where we are looked upon to help make recommendations on diverse things including figuring out how platforms work together, and how to maintain data quality. Supporting a lot of conversations about the costs of a lead, its lifecycle, and how to enrich it, requires MOPs to have a breadth of knowledge across a wide range of topics.

Specifically, when it comes to data warehouses and BI tools, it’s important for teams to have an understanding of their functionality, at least, since a lot of the work that’s done is directly impacting how that data is populated in those tools and what it’s going to look like when people are manipulating it. Multi-touch attribution opens the door to so many things that people are excited to get into BI tools, data warehouses and figure out what they can do beyond just a CRM or marketing automation platform. Multi-touch attribution is just the beginning and we need more to be able to have that creativity on a multitude of platforms.

Encourage the entire team to understand core database design, that foundational knowledge such as what’s a table? What’s a row? What’s a primary key? Just understanding how tables fit together, how the data works, and how a touchpoint is related to an opportunity. But also being able to support those “guest-like” models that you would be building outside of Salesforce.

How can an attribution project be implemented across teams?

Sell the benefits of the models and the outcomes as a win for all teams. Focus on the achievements of bringing the data altogether into one platform. Communicate they are an important part of the process, have a voice, and will benefit from the results.

Ensure you have all your fundamentals in place whether or not you are using a platform or trying to get something built in-house. Do your due diligence to ensure everything makes sense with the way that your systems are already in place so that you’re not reinventing the wheel, but having those conversations about whether this is an opportunity to reinvent the wheel in another platform or with another strategy that you are doing at the same time.

Get those in place before you get into the attribution platform because, with multi-touch attribution, you’ve to have a UTM strategy in place. Otherwise, it is not going to work very well for you to be able to group your online and offline engagements together and look at the journey. Therefore, start now by having those foundations there. Evaluating if you’ve got the foundations ready is a good starting place to see if your organization is ready for attribution since it can be easily sold as the fix-all, but it is not. A lot of times, when you don’t have that there and some organizations do struggle to get everything in place. Getting that in place beforehand and then getting the buy-in from many teams so that when it’s implemented, you’re already off to a great start and you can start using it efficiently right away.

How often does dirty data prevent adoption?

Dirty data doesn’t necessarily prevent multi-touch attribution adoption, rather, it prevents the understanding of the adoption. For instance, if you don’t have a good UTM strategy, you can put together your multi-touch attribution platform, though it will take much longer. And you might be looking at the data and get confused by the results and it’s going to take a lot of troubleshooting to go back and answer okay this is what we started out with, do you have any trust in this? Why is that the case?

Let’s look at it from a very detailed level. You provided the answer at the macro level and say okay, our channel distribution looks wrong. So let us get down to the micro level and see why that’s the case. Your UTM strategy could be off or you could be inconsistent. Therefore, you’ve to go back and reevaluate and not having clean data at the start is going to hinder the adoption of attribution and ruin that kind of trust in the platform.  

If people are starting out right away with questions and say, well, this doesn’t look right because of the data at the start. With attribution, there are plenty of misconceptions and distrust around it from diverse teams. So getting ahead requires the data to be correct at the start and to look at that end to ensure that you’ve got the fundamentals in place.

Support processes for other departments

Be a proponent of data democratization and sharing that information. You don’t have to hoard all the data in the reports and insights. Sharing that type of knowledge makes everybody better at the end of the day. When you’re working with sales teams and you say, hey this is what our model is seeing. Let’s change these cadences that are talking to finance about hey, these channels are generating the most ROI, can we add more budget here vs. this other vendor? That kind of discussion makes everyone better when all is said and done.

And when you’re reporting out that information, remember not everybody cares about every model. Some are interested in ROI, some the impact of the business. That is the type of information that you should be sharing with the other organizations in your company.  That is great as far as getting buy-in and ongoing trust. 

Have guardrails because every setup and model has some assumptions baked in. Therefore, consider having processed documentation to ensure there are no list uploads that aren’t happening correctly. And if you set it up, it has a cost and a type, ask yourself what are the rules needed in place when setting up this data in an ongoing way in the CRM. What is the process for what you can do in Salesforce to make more sales?

Lastly, follow the conventions that are going to unlock attribution for these opportunities. It might not seem like it’s attribution-related, but where marketing ops could help is figuring out what those assumptions are in the model and then communicating those up in the documentation.