• Jervis Koo

Understanding Attribution Models

Most paths to purchase are not linear. There are heaps of ways for people to discover your brand, engage with it, and move further down your sales funnel. Attribution models help marketers gain insights to improve their digital campaigns.

What is attribution modelling?

Attribution modelling is a strategy that allows marketers to analyse and assign credit to marketing touchpoints that occur at the specific steps of the customer journey, from searching for a product online to making a conversion, and every action in between.

Attribution models helps marketers better understand which parts of their marketing effort are driving the most leads to that part of the sales funnel.

With multi-touch modelling, we can distribute credit across multiple touchpoints to see how marketing interactions affect the entire sales funnel.

Why do attribution models matter in digital marketing?

With most marketers running multiple campaigns across multiple channels, it can be challenging to determine which mix of PPC keywords, display ads, landing pages, and SEO are generating leads that move efficiently through the sales funnel and down the conversion path. Using attribution modelling, marketing teams get a bird's eye view of each customer journey from the starting point to the end purchase.

Example Scenario

For example, customer A and customer B are both looking for a product/service your company sells. Customer A converts in 1 step because they came to our website knowing what they wanted and our website had what they were looking for.

Customer B, however, may have a completely different experience getting to that conversion. Let’s say Customer B first clicked on our company’s display ad, then interacted with our brand on social media before visiting the website from an organic search listing, downloaded an ebook and reached out via a live chat a week later before signing up for a demo of our product.

As we can see, customer B’s user journey involved several steps, or touchpoints, over a period of time. This is the most common customer journey we see in today's digital world.

If we only give credit to the last touchpoint before a conversion, we’re missing out on all the touchpoints along the way that influenced our customer’s choice to invest in our brand. This is a perfect example of a multi-touch conversion route and is becoming more common today. By understanding which top-of-funnel efforts are successful, we're able to determine which marketing efforts attract customers in the first place, and guide them deeper into the sales funnel.

Similarly, attribution models that emphasise the last touch before the conversion action give us insight into which marketing channels drive potential buyers to actually convert. When we have an idea of which touchpoint creates conversions, we can make sure we're optimising those channels with the best calls to actions.

But why is it important for us to understand single or last touch versus multi-touch conversion routes? To save us money.

As a digital marketer, we're constantly looking at how and when our ad campaigns create conversion events. Attribution modelling can provide vital insights on our marketing efforts, including what paid search campaigns worked, what campaigns don’t, and where we should be allocating our spend for optimal conversion rates. Attribution modelling gives us the ability to measure marketing campaigns more tactically vs. throwing stuff on the wall to see what sticks approach and leverage return on investment (ROI) data.

Single-touch vs Multi-touch

There are attribution models out there, from AI-based software, Facebook’s own attribution tool, or the standard Google Analytics attribution models. However, they’re generally split into two main varieties: single-touch and multi-touch.

Single-touch models

Single-touch models, as the name suggests, assign credit to one single touchpoint in the customer journey–usually the first (when they initially interacted/became aware of your brand) or the last (when they decided to make a purchase).

The following are a few common single-touch models:

  • First touch model (first-click attribution)

  • Last touch model (last-click attribution)

There's a certain logic here—the first and the last touchpoints are obviously pretty important. If you can’t get people initially interested, then it’ll be hard to lead them along the funnel, however, if you can’t then close the sale, it’s a wasted effort.

But single-click models have their own limitations. The modern consumer journey is increasingly fragmented, with customers quickly jumping from one channel to another. For instance, say your company sells shoes. A customer might see an ad on Instagram, click onto your account, then a few days later go onto your website, before clicking off and only a week later being retargeted by a Facebook ad, clicking on that link, and finally making a purchase.

If you choose to attribute 100% of the credit to any one single touchpoint then you’ll miss the entire consumer buying process. As important as the initial Instagram ad was, was the Facebook ad not just as (if not more) important? In fact, would they have even purchased had they not gone onto your website and browsed your products in more depth?

In fact, would they have even purchased had they not gone onto your website and browsed your products in more depth?

Multi-touch models

This is where multi-touch attribution comes in. Multi-touch attribution aims to assign each touchpoint along the consumer journey with an appropriate amount of value (correlating to its importance in the buying journey).

The following are two common multi-touch models:

  • Linear attribution

  • Time decay attribution

Example scenario: A customer spent bought $100 worth of products from you. In this example, you might decide that the initial touchpoint (Instagram ad) was worth $30. Them clicking onto your Instagram account was worth $20, visiting your site a few days later another $20, and the final retargeted Facebook ad which directly led to the sale was worth $30. Some other examples of multi-touch models include the linear model, time decay model, position-based, or custom rule-based models. Some multi-touch models have pre-determined mathematical equations that dictate how much credit to assign to each touchpoint, whilst custom multi-touch attribution models give you the ability to fine-tune this according to a myriad of factors (prior experience being the main one). In all, it’s clear that multi-touch attribution is the more comprehensive model, and offers marketers significantly richer insights than single-click attribution models.

Single Touch Attribution Models

1) First Touch Model

Also referred to as first-click attribution, this model gives all of the credit to the very first interaction your business had with a customer before they convert i.e. make a purchase. Use this model to see what catches new customers’ attention at the top of the funnel.

This attribution model identifies the unsung heroes of your marketing. Maybe your PPC campaigns don’t directly create sales, but they lead to the kind of awareness and engagement with future customers that does eventually generate revenue. If you only count what’s at the end of the funnel, you might cut your PPC efforts and see a drop in sales accordingly. This attribution model is especially important for businesses that have longer sales cycles and are working to create brand advocates, rather than simply one time purchases.

2) Last touch

The qualified lead model, also known as last touch attribution, gives all of the credit to the last interaction your business had with a customer before they convert i.e. make a purchase. It’s also referred to as “last-click attribution” or “last-interaction.” This model is the default on most platforms, including Google Analytics.

For example, if a customer visited a blog post you wrote before calling your company number on that page, the blog post would be their “last touch”, or the touchpoint that created the raw lead.

The qualified lead model approach has been used in marketing for a long time, but it’s starting to fade as the most effective way to analyze your marketing–especially when there are so many other touchpoints involved in a consumer’s buying decision. However, this model lets you know which marketing channels or campaigns drive your customers to action, so you can optimise campaigns towards those important moments. This model is particularly effective for understanding shorter sales cycles where the buying process is quick and the conversion action is often immediate. 3) Lead Creation Model

Lead-creation-touch models put a lot of emphasis on the distinction between prospects and leads. In fact, lead creation touch attribution models give 100% of the credit to the milestone when a prospect becomes a qualifiable lead so you can pinpoint and understand the critical touch point at which a prospect voluntarily enters their personal information into a form for a webinar, newsletter, etc. and becomes a lead.

Multi-Touch Attribution Models

1) Linear attribution

This model involves dividing the credit equally amongst all of the touchpoints in a conversion path or customer journey. If there were 3 touchpoints in the consumer journey, each of those points would get 33.33% of the overall credit for that sale.

2) Time decay

This model is similar to the “linear attribution” model where every touchpoint is given some credit for a conversion, however, the more recent touchpoints are given more of the pie, and the least recent interactions get a smaller piece.

Choosing the right attribution model for your business

No two brands have the same marketing strategy so you need to be selective when choosing the appropriate attribution model.

If you’re selling an incredibly high-value product—for example, a vehicle—the sales cycle is naturally longer vs selling shoes. The marketing strategy to sell vehicles will be far more complex, combining multiple campaigns, both online and offline to drive prospects to a sale.

To measure this, you’ll need a comprehensive attribution model that shows how prospects are interacting with different campaigns throughout the buying journey;

  • How they first become aware of your brand

  • Why they dropped off

  • What made them return

  • What made them purchase

When choosing an attribution model, the first step is to decide on what is you are trying to achieve. Are you looking to have oversight into how different campaigns work with each other to push the customer down the buying funnel?

For example, you might simply want to measure how effective certain channels are at acquiring prospects, before coming back to your website to purchase after lapsing by seeing your retargeting ad.

Alternatively, you might want to deep dive into the details of your marketing campaigns. Not all touchpoints necessarily contribute equally towards a customer's purchase decision, but it can be valuable to find out which campaigns worked the best. Doing this will help you identify which touchpoints are more valuable in comparison to the others.

The most advanced attribution models often require fine-tuning how much credit you give to each touchpoint—this isn’t always easily worked out and requires data-scientists. For smaller organisations, a good place to start would be using common out of the box models that come with your tools and improve on the model as the business grows.

Account for online and offline touchpoints

Most attribution models are great at looking at online touchpoints, but don’t take into account any offline interactions a customer have with your brand.

Something a consumer sees online may prompt an in-store visit or vice versa. Prospects might have read your blog, which prompted them to attend a conference, which then convinced them to buy from your company.

Blending offline and online touchpoints can be tricky, but failing to do so could give you a limited and inaccurate view of your consumer journey, however, there are potential solutions:

1. Measuring foot traffic

Many attribution companies measure foot traffic by using beacon technology. These radio transmitters attach to Bluetooth-enabled devices, like smartphones, to capture location data. If someone saw your campaign on their phone and visits your store, you then have a way of knowing that their online activity affected their offline behaviour.

2. Point-of-sale surveys

Point-of-sale surveys is an option, though they’re not usually very popular with customers. Briefly asking where they came across your company might reveal some engagement with online campaigns, though this method can be fairly imprecise–sometimes people either don’t remember which specific campaign or channel introduced them to your brand, or they may not really want to engage.

3. Call Tracking

Call tracking is one highly effective means of accurately attributing both online and offline touchpoints. For instance, your company launches a series of radio ads, puts up billboards, and sends out a direct mail campaign. If each medium has its own phone number attached, someone who calls having seen the billboard will call a different number as opposed to someone who heard the radio ad. Using this method will help quickly identify the effectiveness of each campaign.

Evaluating your current attribution model takes time

Your attribution model isn’t there just as a nice-to-have. It’s supposed to help you improve your marketing strategy on an ongoing basis. If you change your marketing strategy multiple times according to your attribution model but don’t see a tangible change in results (i.e. overall sales), then it’s time to find a different attribution model.

Any time you make a change to your marketing strategy, give it some time before assessing whether or not you need to make another. Your prospects’ decision to convert depends on multiple factors–not just your marketing strategy. Therefore, make sure you give each change at least a few months before changing again (unless you see a sudden and otherwise inexplicable drop in sales).

Which attribution model should I use?

This depends on how complicated your average consumer journey is. If you’re a larger organisation, you might have multiple ads running across different channels: digital, social, TV, out-of-home, radio, etc.

This means that many customers will interact with your brand several times before purchasing. In this case, it’s worthwhile choosing a comprehensive multi-touch model that takes this complicated journey into account.

However, if you’re a small company that markets on fewer channels: digital and social, investing in a highly complex attribution model might be counterproductive.