THATBlog

Attribution modeling gets into some complex territory. It can be broken down into its fundamentals with ease, but understanding how they relate can send heads spinning. Is there even such a thing as a best attribution model for eCommerce?

Conversion Paths

Let's first understand what attribution modeling is. When someone makes a purchase on your site, what route did they take to get there? Did they link to content from social media, read the content, click on a link, browse products, add a few to cart, and then purchase? Or did they do a web search where SEO practices guided them to your landing page, where a chat bot helped them quickly narrow down a precise product search?

Attribution modeling does not measure every single page they visit. It measures the conversion that helped them get there on this particular visit. In these two examples, that's the link from social media, and the web search that turned up your landing page. It doesn't involve the browsing or chat bot. A conversion path is about repeat visits or lack thereof, not about tracking how someone browses on a single visit.

Attribution modeling seeks to assign value to how they engaged your site. Was it through social media content? Remarketing? An SEO search that turned up content that they clicked on? Did they directly enter your website's name in the URL?

These are just a few examples. You can probably think of dozens of possible routes that a customer might take before purchasing. Some customers click through content to product and make a purchase quickly. Others will comparison shop, read reviews, and make dozens of return visits before making a final decision.

The goal of attribution modeling is to assign value to the touchpoints of your site that influenced a customer's purchase. Every route they took to engage your site is a touchpoint. These touchpoints together are called a conversion path. The goal is to find which elements of your site – which touchpoints – strongly correlate to purchasing. Along the way, you may also find which elements are weak, extraneous, or need re-working.

Why Use Different Attribution Models?

There are five primary attribution models. Each fits a different circumstance and gives you different information. There is no such thing as one single best attribution model for eCommerce, but there is a best attribution model for your organization.

A small, streamlined site that relies on selling a single, expensive product in each purchase will use a different model than a site that relies on customers buying multiple products each purchase. The conversion path on the first site may be fairly straightforward depending on its marketing model. The conversion path on the second will be a winding road with return visits and an overwhelming number of touchpoints.

A site with an experience that's very guided may not worry about return visits – if every customer is having more or less the same experience, then it's that first click that brought them to the site that counts most.

A site that features tons of content and conversion paths designed for inbound marketing will still value that first click that brought a customer there, but it won't want to overlook all the additional visits that also helped close the customer. Let's look at each of the five attribution models, with pros and cons for each:

First Click Model

This is also called First Touch Attribution. This model gives all of the credit to the very first touchpoint a customer visits. If their first interaction was clicking through to SEO content that showed up in their Google search, that route gets 100% of the credit. If they first click through to a landing page from a social media account, that route gets 100% of the credit.

Pros: This is very effective for measuring what brings customers in to your website. You may find strong correlation to a certain piece of content shared on social media – this informs you to produce more content and share it on the same platform. This can be most useful for a guided site experience, a site that has very few pages beyond content to draw visitors in, or a site that only sells a very few products can rely on First Click models.

Cons: Do you feature a lot of extra elements? Is there no set path through your site that everyone has to follow? Does your content build on itself in a way that reinforces previous content? What if a customer visited several other pages in return visits? Was the customer convinced to purchase by a later touchpoint? First Click can't measure these. It's only useful for valuing what first brought someone to your site.

Last Click Model

This is also called Last Touch Attribution. It gives all the credit to the very last touchpoint before purchase. It can be useful, but it's generally not considered to be as valuable.

Pros: Last Click modeling can help you confirm that customers are finding certain pages useful or informative. It helps you know which routes are helping to reinforce a purchase the customer has decided to make. If you're focused on impulse buyers almost exclusively, Last Click can be a very useful model.

Cons: It doesn't do a good job of telling you what helped the average customer make that purchasing decision in the first place. It doesn't give value to what guided the customer to your site. Ideally, your site makes a strong argument, features intriguing content, or educates in a way that closes the sale earlier than the very last touchpoint a customer hits. Customers will generally talk themselves into a purchase earlier than this, and then re-visit later to make the actual purchase.

Linear Model

Every touchpoint gets equal credit in this model. If there were four touchpoints, each gets 25% of the credit. If there were five, each gets 20% of the credit. You get the idea.

Pros: This approach can help you notice major touchpoints that are working very well. Perhaps there are midpoints that work as a hub to drive people toward making purchasing decisions. Perhaps there's a page they used that works very well for helping people find products very efficiently. Many people bookmark product pages for later visits. First Click and Last Click models would completely miss the value of these middle visits. Linear finds value in midpoints that usually make up the bulk of a conversion path.

Cons: The Linear model can, however, overvalue more minor touchpoints. The first and last clicks are both important. Many midpoint routes are important, too – but many aren't. The ones that are extraneous or distracting may get overvalued, and you may not be able to tell the difference. This can then cause you to place resources into overvalued routes that were simply represented too strongly in the model.

Positional Model

This model values each touchpoint on a conversion path, but not equally. The usual Positional model gives the first click 40% value, the last click 40% value, and the remaining 20% is divided like so many scraps to the midpoints.

Pros: This can be very effective for streamlined sites that sell only a few products. You're essentially prioritizing what attracts the customer, what reinforces or closes the purchasing decision, and not worrying too much about those middle visits.

The weighting is also a recommended place to start getting into deeper analysis. You have flexibility to change how values are weighted as you see fit. This can work well when you have precise questions about your marketing you want to assess and you know how weighting can provide an answer.

Cons: For sites that have a great deal of content, products, pages, and invite the customer to re-visit, share, and keep adding items to their purchase, this attribution model can quickly become uninformative. As brands become more community-oriented in building their customer base, re-visits become much more frequent. The chief touchpoints that are closing sales in this circumstance are all midpoints that are severely undervalued.

That weighting being flexible can also work against you if you toy with it or obsess over it to the point of distraction. If you don't have a precise question about the data, you may keep adjusting the weighting of different points until you see what you want to see.

Time Decay Model

This model applies an algorithm that values most highly the touchpoint closest in time to the point of purchase. As you go back through the conversion path, each touchpoint gets less and less weight until the first click receives the least.

Pros: Every step is given value in relation to the purchase itself. This model also favors the later midpoints where the customer is really becoming convinced to make a purchase. It's the most favored model, although many organizations make their own tweaks.

Cons: Last click may be overvalued and first click is generally undervalued in this model. It'll give you a great idea on what's closing purchases, but not always what attracts a customer in the first place. At the same time, there are other tools for that.

The Best Attribution Model for eCommerce

You can see that there is no single answer. Some organizations may measure using multiple models. They may blend a Time Decay and Positional model together to compensate for potentially undervaluing First Click and overvaluing Last Click.

Or they may use Time Decay because it helps them understand how conversions are being closed, and also keep an eye on First Click because this informs them what's drawing customers in. Many organizations practice a bit of trial-and-error before they start finding useful correlations. Like any exact science, it takes some work and patience to figure out how to apply it in different situations.

An attribution model is absolutely worth using. Even if you're not using the best model for your circumstance, it's going to give you more information than if you weren't using any attribution model at all.

Analytics That Go Further

Beyond this, you can speed up the process by working with a marketing agency. When they've worked with similar companies or with businesses that rely on websites that have similar set-ups, they can focus in on the best attribution model much more quickly. After all, they're doing this for countless other organizations. That gives them a much larger view of how and in what circumstances each individual attribution model becomes most useful.

Remember that most first-time visitors won't buy, and that assisted conversions will make up the majority of your sales. Attribution models can be a lot more dynamic once you get a hold on using them effectively. For instance, they can be segmented to measure what's effective with return customers vs. first-time purchasers. You can identify what's effective for customers who linger their way through multiple visits vs. those who know what they want and zoom straight through to conversion.

Returning customers are much, much more valuable than the average customer. You can assign attribution models to measure what's effective with them – it may be very different than what's effective for less frequent customers. A combination of attribution model, plugins, and analytics can begin to reveal far deeper information that enables your business to identify opportunities that would otherwise be overlooked.

It's not just about which attribution model is the best for your organization. That's just the first step. You want to make sure you've figured out how to make that useful before you dive into more complex territory. Just be aware that there are even more complex and revealing analytics that can be built on top of an effective implementation of an attribution model.