🎙 Episode 3 on Attribution is here! 🎙

‍

This episode is purposefully geared towards the PPC Newb or Student.

In this brief (seriously, it's only 15 minutes long) Class 101 on Digital Attribution, Kirk & his four returning guests, takes a look at the attribution models that can be used in PPC and when they can/should be used.

This isn't an advanced episode, but you might learn something anyway, and as usual... we hope you enjoy the production quality!

Heck, even better, please share this with those you know just getting into PPC as they contemplate the potential models out there and how to think about using them. It may be just the bite-sized introduction to digital attribution models that a PPC intern needs to hear as they learn their craft from you, the expert.

‍

Listen on Apple Podcasts

Listen on Spotify

Listen on Google Podcasts.
‍

‍

Big thanks to our four guests for these two attribution episodes, and make sure to listen to their full interviews to dig into attribution more deeply:

EPISODE 3 GUESTS

• Amanda Farley - AimClear
• Heather Aeder - Seer Interactive
• Mike Ryan - SMEC (Smarter Ecommerce)
• Nechama Teigman - AdVenture Media

‍

EPISODE 3 Resources

• The Complete Guide to Direct Traffic in Google Analytics by Tom Bennet on Moz
• What’s the best attribution model For PPC? by Aaron Levy on Search Engine Land
• Ponderings of a PPC Professional Book by Kirk Williams

‍

‍

‍

👉 Stay Up on Future Episodes By Subscribing Here 👈

‍

Episode Transcript

Kirk Williams (00:01):

Okay, before going on, I want to pause and give a brief caveat. This episode is geared towards those who are unfamiliar with the basics of attribution models and PPC. If you're more advanced in PPC, this will likely be review for you. So no offense taken for me if you just want to skip this episode. That being said, I'm also of the mindset that even the most advanced person can continue to learn about their craft from someone who thinks just a little bit differently. So even if you know the basics of attribution, I trust you'll find value in at least hearing the viewpoints of the different guests. Caveat over. Let's get to it.

Chris Reeves (00:38):

Welcome to the ZATOWorks PPC Ponderings Podcast, where we discuss the philosophy of PPC and ponder everything related to digital marketing. Our hope is that through these conversations with professionals in the digital marketing space, we can gain a better understanding of what is happening in the digital landscape, and better prepare all us for the future.

Kirk Williams (01:00):

In our last episode, we discuss some philosophical aspects of digital attribution, as well as my own concerns and pitfalls with it. In this episode, we wanted to take a look at how attribution is actually utilized in paid search marketing. We'll spend most of our time thinking through the individual models of attribution and how we should think about using them with paid search or PPC.

Heather Aeder (01:22):

So for me, the way I define attribution is, attribution is kind of the study of understanding how multiple marketing touch points on the way to a conversion impact return on ads [inaudible 00:01:39].

Amanda Farley (01:39):

For attribution in channel. It can really change how you're optimizing for your conversion performance, but on the reporting side, those models are also really important because that'll show you what's influencing what, so you can make smarter marketing decisions.

Kirk Williams (01:54):

Our four attribution guests are back for one more episode.As a reminder, we're talking with Heather Aeder of Seer Interactive, Nechama Teigman of AdVenture Media, Amanda Farley of Aimclear, and Mike Ryan of Smarter Ecommerce or SMEC. If you prefer. We've released their full interviews as bonus episodes. So make sure to catch their extended thoughts in each of those. All that being said, let's keep a couple things in mind from our last episode.

Mike Ryan (02:20):

Like all models are wrong and some are just you more useful than others. And super important to keep in mind with attribution is very applicable here.

Heather Aeder (02:30):

So a hundred percent, yes, attribution can be overvalued.You know, that old hype cycle graphic that has this big bubble coming up in the chart and the period of hype cycle. And then it drops down into the trial of just disillusionment. And then it comes back up more into reality. I would say attribution has gone through that.

Kirk Williams (02:51):

As we think through which models we should utilize for ourGoogle Ads campaigns, we should keep these caveats in mind. Let's get into the models. What attribution models are available to the Google Ads practitioner?

Heather Aeder (03:02):

So I think the most common models that are available like out of the box and any of these bigger MarTech platforms are going to be first-click last-click or data driven or linear would probably, those are probably the four most common ones. Any that aren't that four are, are probably not widely used. So like U-shaped or time-decay. I don't ever recommend just looking at one model. So the reason that there are multiple models in there is so that you can actually get perspective depending on what you're trying to optimize against.

Kirk Williams (03:36):

So as we consider models, are there any we should avoid?

Nechama Teigman (03:40):

I'm definitely anti last-click and first-click. I don't have a problem being strong on either of those two, because you... By default, you're discounting other one.

Mike Ryan (03:48):

And you'll get a different perspective with last-click than you will with other models. And this has been the default view that people had for a long time, is last-click.

Kirk Williams (03:57):

We talked about this in Episode Two, but let's recap.Last-click attribution means the last source to send the sale gets all of the credit. Remember that illustration we talked about and used in Episode One?

Mike Ryan (04:09):

Strangling the funnel is a marketers, rather grotesque description of describing when a business stops advertising in those initial places, their consumers hang out. Thinking those customers aren't valuable since they don't see the evidence of purchasing in a last-click model.

Kirk Williams (04:23):

Fun fact, but Google analytics uses a last-click non-direct model. That means the last source to send you the sale gets all of the credit unless it's direct traffic. Why the hate for direct traffic? Well, direct traffic has an interesting lineage. Direct traffic is tracked visits to your website that did not come from any other source, referral, publisher, etcetera. For a long time, most people attributed the vast majority of direct traffic as a really valuable bottom funnel traffic source. Like people just typing your website into the search bar and landing directly onto your website because they had nothing more to do except remember your brand and typing your exact URL. Now we know it to be well far more complicated than that. Tom Bennett on Moz talks about HTTPS to HTTP visits, broken tracking code, improper redirection, PDFs, and dark social as the other possibilities for what your direct traffic actually could be. Remember what we determine in the last episode on pitfalls to avoid?

Nechama Teigman (05:24):

But anytime you're looking at data, specifically marketing data, but really any data, it's that it has to make sense. Sometimes that doesn't mean it has to validate your theory, because sometimes our theories are just straight out wrong. But it has to make sense. If it's not making any sense, it doesn't mean that the attribution model's wrong. It doesn't mean that you're wrong. What it means is that you have to do a deeper dive.

Kirk Williams (05:49):

Attribution is tricky and data can be misrepresented or misunderstood pretty easily. So let's keep that in mind during this episode as well, people love to say, "Well, data doesn't lie." Well, that's totally correct, but hey, our understanding of that data can well lie or just be completely wrong.

Heather Aeder (06:08):

So you take an e-commerce brand as an example, they're constantly trying to tie data back to their order management system to be able to track like, oh, this order is in this stage. So it's not just been purchased, but it's been returned. So I actually don't want to count credit towards marketing channels for all my paid search keywords from one ad group are resulted in returns. That's not positive ROI for me. So I think that database component has been around for a long time. And the reason is because it's to connect to data sources beyond just digital sources.

Kirk Williams (06:41):

By the way, interesting news as of late, but GA 4 actually allows you to set a different attribution model as the default model, which is something they've never allowed before in Google Analytics, to my knowledge. Anyway, back to models available to the Google Ads practitioner. What are they?Well, we've already looked at last-click attribution.

Amanda Farley (07:01):

Last-touch, which you know, we like to talk about a lot. I think in PPC because Google for a long time, got all of the credit, because that was default on everything was that they were the last one. They were the deal closer. So they got all the credit.

Kirk Williams (07:13):

In my opinion, first-click falls into a similar category as last-click. Instead of the last source, it assigns all credit for that sale to the first source to send traffic. So you're still removing all potential value from any source, except just the one, in this case, first rather than last-click. I think there are still weaknesses in doing so. One thing to consider is whether to use multiple tools to analyze how individual campaigns are performing based on the attribution model that fits their place in the user funnel. Hm? Well think of it like this, brand and remarking traffic tend to be bottom of the user funnel. So how are they going to perform in a first-click model? Not well, but I mean that's not their fault. That's a funnel model mismatch. They're being asked to perform in a way outside of their nature. So consider this. What if you analyze your upper funnel traffic with a first-click model and your bottom funnel traffic with a last-click model?

Nechama Teigman (08:07):

I always find it helpful to look at a last-click model and a first-click model versus a model for whatever giving full credit to whatever platform I'm interested in. So for example, earlier today I was looking at a client's account and we're looking at the last-click versus the first-click versus giving a hundred percent of credit. And that really gives you like a much better understanding of the customer's journey.

Kirk Williams (08:31):

Remember, according to our interviewees, keeping an eye on multiple models is crucial to get a better understanding of how people are purchasing from your store and what channels they're using.

Mike Ryan (08:41):

A misunderstanding of attribution is maybe that this is actually reflecting objective reality. That can't be the case entirely. You know, attribution is always going to be just one view of reality and these different models can have value in terms of offering you different perspectives.

Kirk Williams (09:00):

What about the linear model?

Amanda Farley (09:01):

Linear that's when everything is equal across the board.

Kirk Williams (09:05):

In this model, every visit is assigned the same credit.The benefit to this model is that we're not attempting to determine intent, like we discussed last time. We're just assigning completely equivalent values based on how many visits there were in that buying journey. Truthfully, I don't think I've ever met anyone who actually prefers this model. I mean, someone might be out there, but like I've never met them. Most advertisers acknowledge that a first and last-touch probably have additional value that the middle touchpoints don't have in a user journey or in some ways, at least different touchpoints have different values rather than just all of them being equal. So this one hasn't really caught on, at least in my network.

Kirk Williams (09:47):

Time-decay is an intriguing model since it allows for increased value, as a person gets closer to the point of sale. But then again, we have the issue with these values being completely made up. Some actually argue that the reverse should occur. The closer a person gets to sale, the less the value that should be assigned since a person is becoming nurtured in the funnel without the advertising itself actually becoming more effective. Like your sales process is working, that doesn't necessarily mean each ad is getting more and more strong in quality. Regardless, I rarely meet anyone utilizing this model even. What about U-shaped or position based?

Amanda Farley (10:25):

And then also position based where you give more credit to the first or the last-touch of the journey.

Kirk Williams (10:30):

If I can't use data driven attribution or DDA, I tend to prefer this model, the U-shape or position based as it weights the first and last-clicks heavier than those middle sessions. And then it divides the middle session credit up amongst themselves equally.

Nechama Teigman (10:44):

So being that it is a new business, the first touchpoint's going to be incredibly important, right? Because we're educating people about the existence of this business versus, and the last click is always important because without it, you end up nowhere. And then the middle clicks are also important, but probably less so. So I would default to a position based a model in that situation. However, whenever I'm looking at data kind of like moving forward, I'm always using model comparison tool all because I don't think that there's any one model, which is perfect at all.

Kirk Williams (11:20):

Okay. So what is the preferred model? Well, according to our guests, it's inevitably an algorithmic based model.

Heather Aeder (11:27):

Data driven is where you're taking a statistical model and layering that in. They're measuring things like incrementality as an input into that model for are predicting how much credit a touchpoint should get. They are also looking at things like recency and frequency as part of that model. And then some tools even look at outside or external data sets as input into that model as well. It could be something like seasonality.

Kirk Williams (11:55):

Data driven attribution or DDA in Google started out requiring a lot of conversions. So as to build a model. But Google's been lowering the required count. I'm not fully sure as to why.

Heather Aeder (12:05):

I do think that's potentially where it comes in, the predicted piece comes into play is filling the gaps of where we're missing information.And that's one of the primary reasons Google's rolling out GA 4, is they are incorporating a component of that is the ability to be predictive on saying like, yeah, these are the same people and tying the journey together. Right? So making assumptions and modelling out that like yeah, person A's journey started here. Oh. And we're predicting that this is actually person A over here and tying together the journey. So I do see some of that coming in there and that's really going to be highlighted in GA 4.

Kirk Williams (12:49):

So there you have it. At least as it comes to PPC and paid search, those are the attribution models we have to work with at this time.What about other aspects of attribution? Like look back windows to consider.

Mike Ryan (12:59):

You might look at a 365 day look-back to understand the big picture. What is my conversion rate over time? And in a rather stable way.You can get a very stable average with that. If you are looking at the last seven days, for example, it's going to fluctuate a lot more and there are advantages to both things. If you look at last 365 days, you can get a broad sense of seasonality, for example. If you look back at last seven days, you can get a picture on trends and things that are merging. What's happening now currently.

Kirk Williams (13:34):

Okay. What about your time lag? What impact does time to purchase play in an attribution model?

Heather Aeder (13:39):

So for example, if you're a brand and your typical time to conversion from first marketing touchpoint to conversion is less than a day;attribution's not worth it. Because you're probably only going to have one touchpoint, maybe two on the way to the conversion.

Kirk Williams (13:56):

Finally, when should you consider changing your attribution model or if you've decided to change it, anything that you should consider before changing that, or while you're changing it?

Amanda Farley (14:06):

I ask "Where I'm allowed to test." And then I like to start in like a sandbox environment and not mess with what's working.

Mike Ryan (14:14):

If you've been measuring in a certain way for the last year, let's say you want to predict what a bid should look like. You're basically going to try to predict the conversion rates. You're going to try to predict the average order value and you'll want to do this on the basis of historical data. When yeah, the conversion that you get assigned is changing due to attribution. And you make a big cut like that, the look back window that you're looking at has been measuring one way and you're measuring a new way. Can you make valid bids and predictions on this new model, this new way of looking at things?

Nechama Teigman (14:47):

So I think a lot of the default I've heard, especially from Google side is like two weeks is a time period. I like to think of it more at more so in terms of clicks. I know people like to talk about spend like, okay, once you spend $10,000 you're okay. Or whatever it is. But I don't think that's makes sense. Because you could be spending $20 on a click or 50 cents on a click. We're talking about a volume game here. So if a client has a higher amount of clicks, so they're a client that gets more volume to their site from our campaigns, then I'm usually comfortable saying within two to three weeks, we should be able to move on.

Nechama Teigman (15:25):

If it's a client that gets lower, a smaller amount of clicks to their site, it could take a couple months. And that's also part of the...When we're making the decision to switch over, that's part of it. If it's going to take six months to get anything there. And also when there's so little data that it's going to take six months, that's a problem because it's not that it's going to take six months. It's also that Google tends to value more recent data, more so than older data. So when you tell Google to switch the attribution model and you're not going to have enough data for six months, it's probably not worth it to even switch if we're being honest because it's never going to actually have enough data to do a good job for you.

Kirk Williams (16:10):

I hope this has been helpful to you. Just some practical tips about considering attribution models in your Google Ads account. Whatever you've picked up over the past two episodes, I do hope you don't forget to use data as a wise counselor, but not as the end all that explains every aspect of your customer journey. And as always, may the auctions be ever in your favor.

Chris Reeves (16:39):

This podcast was recorded in the ZATOWorks PublishingStudio in Billings, Montana. ZATOWorks Publishing is a subsidiary of ZATO, a paid search agency focused on e-commerce brands owned by Kirk Williams. We look forward to seeing you again next time.

‍