009. Attribution Models: What You Need to Know

Hello everyone, this is Emilio Garcia from boundify. Really excited for our next session of Boundify Live. We usually set up these meetings through zoom and people joining if they want to, and we discuss different topics. Today is a little bit different because I decided to stream this event into LinkedIn too. So I hope some of you will be able to join just directly from LinkedIn. If you have any comments or questions during the transmission, please just put it on the comments and I will try at the end to answer anything related to demand generation, lead generation, inbound marketing, sales enablement. Those are the topics that we usually discuss. And for those of you that cannot join today, the recording will be available both in video on YouTube and as a podcast in our show Bountify show. So many options out there for you to continue communicating with us. And today I'm happy to talk about a very interesting topic. Today I want to focus on something called attribution models or marketing attribution models. So obviously the first question will be what is exactly marketing? What is exactly the attribution models and how do you use them on marketing? So for anyone that is especially, you know, performing any marketing efforts, there's usually some need for measurement, for trying to understand what will be the performance, what is the performance of a campaign. And there's different ways that you can attribute the performance of a campaign to whatever metric you're trying to improve. So attribution models is just a fancy way to tie performance to specific actions on your marketing efforts. There are many different attribution models, and different platforms have different ways of measuring it. One of the first things that I can tell you, for example, is that you usually, if you pick an advertising platform as your reference, you will lack some information putting, for example, imagine that you are using Google Ads as your reference for attribution. So Google will be able, through their tracking code, to see all the interactions that happen through the advertising, but it will attribute to itself any conversions that you are measuring, even if, for example, you are having also you have running another effort on Facebook or any other social media platform. So they will compete for the attribution because each of those platforms will only be able to see their own silo. So that's one of the challenges of trying to work out attributions per platform is they cannot see the rest. So usually the attribution models that work the best are the ones that can see the whole, you know, every single channel, not just one channel in particular. So in order to do that, for example, in this occasion, I will use this reference Google Analytics, obviously there are many, many web analytic tools out there. I will also share a little bit about HubSpot, which is another platform that you can use to kind of like measure attributions across the funnel, not only for contacts, but also for revenue, for opportunities. But, but, but Google Analytics will be best because it's wildly used and obviously it's free. So those are great incentives to start there. And once you have that narrowed down that you can jump and tweak your attribution models for the rest of the platforms if it makes sense. So I will start sharing my screen just to kind of like have something to guide us. Has reference here. Let me do that. I will try to share one specific tab. So I have it here. All right, so I have here some of the models available. So the first one, the one that usually is that most marketers use for the file for default, even if they are not aware of that. I have seen this happening many, many times that sometimes you're not aware that you have that model available and you're just still using it. It's last interaction or last touch. That's what it's called. So in this scenario you are attributing whatever conversion you are measuring to the last event that lead to it. Imagine for example, you have, for example, your conversion is someone filling a form. So if you have multiple channels, advertising and some SEO efforts and some email marketing, the last action that the user took before filling that form, that will be the one that will take the recognition of the conversion is usually the default for platforms like Google Ads. We'll call it Last click. It's pretty common. Obviously what lacks on this model is that you ignore the contribution of the rest of the channels. When is the best time to use this one? Well, when you have just one channel that you're pushing hard, bringing leads or bringing revenue, so you can pretty much say that all the effort comes from that channel. In those cases makes sense to use last click or last attribution because you don't have competing platforms trying to get attribution for the, for the success. However, if you have multiple channels, especially if you have remarketing, then you will have this situation where maybe a channel is contributing through the pad, but you will not recognize that contribution because you are just assigning it to the last platform that did the effort. Slightly variation of that is the last non direct click model attribution. In this case you exclude the last direct click, which is when someone visits the website directly and then you attribute it to whatever channel came before that one. Imagine for example that someone searched for you organically and then they visit the website a few days later. They just type the website URL directly. Then the attribution will go for SEO or for organic versus going for direct. This is slightly helpful in this case. You are taking away any. You're trying to see if any other channel is actually contributing. Because direct in a sense is not really a channel. It's just someone getting directly is the catch all term for anyone that just visit your website. But still there's the opportunity of missing whatever else contribute to the conversion. The other model that is quite popular is the first or not quite popular, but can be used is the first interaction model. It's a flip on the idea. In this case you say, well I will not attribute the conversion to the last click, I will attribute it to the first one. You will use this when you have scenarios where what matters is what channel brings the attention of the user for the first time. When you're trying to build awareness, for example, that will be a good example. In those cases, even if you have remarketing efforts or you nurture that prospect across many, many days, what interests you is what happened at the beginning. And you will attribute that conversion to the first channel that touches the visitor. So far all of them have this problem or trouble that is either the last or the first, but there's nothing in between. So that's when the linear attribution models come saying this is the first kind of like simple version that you have to have multi touch or multi channel attribution. It's simple, but it's at least better than having just last Stanford and it's very simple. The whole idea is that you will distribute the attribution between all the channels that contribute. So for example, imagine if you have a particular path where you have paid, search, social, email and direct, like in the example that we're seeing right now. In those cases, each of them will get 25% of the attribution. Not bad. But some people might argue that maybe the first touch or the last touch or will be the ones that contribute the most to the, to the effort. And what is in between is not as important. And you don't have a say on the percentages of course. They are just, you know, distribute equally across the events. So in those cases, this particular model, which is usually the one that I rely more into the position based one, just because of the way that we try to generate demand, we pay close attention to whatever trigger a client or a prospect in the Beginning or whatever. Finally convince, make them to take the leap and fill the form or make the purchase. I rely heavily on position base when I lack access to another one that is based on data. So position base, as he says, is it gives more weight or credit to the first point and to the last touch. So in those cases, whatever's in the middle gets the rest of the credit. And again, it's a good compromise between last and first and just giving some credit to whatever is in the middle. Like for example, if you have display, which is usually in the middle of a conversion, a long conversion path, they will get some contribution, you will get some credit there, but you will rely heavily on the ones that started or finished the conversion path. Finally, you have time decay. Time decay is interesting and I think it was more important or more relevant in the past that is going to be in the future. Because the whole idea of time decay is that you start discounting the first touches in an exponential way or logarithmic way, you'll start to count less and less as time pass. And the whole idea is that at least the user will remember or whatever happened more recently in theory has more effect on what the user experimented. So you give a higher weight to whatever channel influence the user more recently. That's the whole idea. Again, a little bit of a compromise is like changing the slope on linear. And so those are the ones that most people usually get exposed to. There's another one that comes from Google and to the best of my understanding it's available at least on Google app. But you require data and it's called, you know, it's driven by data. In this case it's. It will calculate what are the right or the appropriate percentages based on the conversions that you have on your analytics and it will give you that tweak version of the percentages instead of being fixated on either linear 40 and 40 or just time decay. And you can use those scenarios for you. Now, at the end of the day, obviously there's no right or wrong on what model to use. It has to make sense for you. But some guidelines that I can share with you that are simpler to follow are if you for example, are just starting and you just have one channel doing some efforts to bring awareness, traffic and conversions to your website, then usually first or last touch is a good start because you don't have competing platforms between your efforts. The moment that you start having multiple platforms bringing traffic driving awareness to your website, especially if you're doing remarketing, it will be wise to Move into a model that gets some attribution to multiple channels. You might start with linear. If you're heavy on paid, you might, you might use something like position based which is usually what paid usually affects either the beginning or the or the end in most scenarios. Or you can use something like time decay. If you are heavy on for example loan pads with content, that will be a better strategy. And obviously if you have something that is data intensive, you can take advantage of the of the database model if it's something available to your account. But there starts simple, right? Start with. At least in my experience, the largest problem, the biggest problem is not choosing the right model for attribute conversions. Usually what I see is that the account doesn't have conversion tracking or if it have one, it's not being properly set up. So first of all fix the conversion right measure conversions and then after you have accumulated a lot of data, then you can dive or getting to the more complex idea of attribution, one last thing that I want to do is just to give you a comparison is to show you some of the attribution models that a different tool uses. In this case is HubSpot. When you are doing advertising and using HubSpot for trying to connect the information, they have a very interesting way of, of tying up the metrics because in this case they use for example the form submission as the reference point. The names are different because not sarily the form submission will be the last activity that someone will do. It will be equivalent to a last click. But you have that available. But you also have the first website session that is kind of similar to first touch, right? In this sense you attribute to the first session that you have even if that didn't end up being a form. So that will be like equivalent to first and then you have the first influence, this one. All form submissions, where it's a little bit like linear, where if you have multiple forms because you have multiple pieces of content and you want to see how those pieces contribute to the path, then you can give some credit to those forms being filled. You can see for example here just for a contrast, that depending on the tool, you will have different tools and different ways of measuring the same events. And that will be the second most important piece of advice or reference that I can share with you. One, focus on starting conversations and then you go into more complex scenarios. Second, be very careful what will be your source of truth or the system of truth. Because different platforms will report the same information in different ways. Depending on your attribution model, your conversion windows and if they are seeing the whole channels or all of channels instead of just watching their own track, right? Sometimes what happens in my experience is for example that people will get fixated on Google Ads is not reporting the same amount of conversions or the same timeline of conversions that Facebook and Google Analytics and then you try to match that to HubSpot. That will be very hard because you will have different rules on all the systems even if you intentionally try to match them. So a better approach will be choose one tool that will be your reference or source of truth and the rest you will use as approximations or leading indicators for the things that matter, right? For example, if you choose your CRM or automation platform to be the one that will tell you how you will distribute the leads, then you can use something like Google Analytics or the platforms to inform you about more specific topics that you can understand. There are but not necessarily trying to match them 100% because there's going to be always some discrepancy. That's the topic that I have for today. Let me go and see very quickly if we have any comments here. I think no, not right now. Which is okay. Just happy to have this conversation today and hopefully you found value on this episode. So again, my name is Emilio Garcia. If you want to watch this episode completely. If you're just getting into the live session, remember it will be available on our YouTube channel and also in our podcast. I will be sharing snippets of some of the episodes across time on LinkedIn and hopefully we will be able to talk next session next week. Have a great day everyone.