Skip to content Skip to footer

Digital Maturity Insight

Which attribution model is right for your business?

by Redweb, 24 June 2020

Read Time: 5 minutes

So now you’ve learnt the base theories of marketing attribution, where to start and why it’s useful, understanding the use and importance of attribution modelling — how conversions are attributed to the different user journey touch points — is the next logical step.

Choosing the right marketing attribution model for your business is absolutely important, but which one is right for you and your business?

With the speedy rise in digital activity and data surrounding the online user journey, deciding on the most dependable attribution model has become more challenging. Not to worry, in this section we help you reach that next step by exploring some of the top-level attribution models out there (See figure 1).

The journey

To demonstrate how each model assigns value, first let’s pretend that you are a customer that has followed a purchase journey from beginning to end:

  • You notice an ad on Instagram showcasing a luxury watch. You click through to the product page on the brand’s website and then leave after browsing the on-page content.
  • Two days later, you see the same ad once more on Instagram. You click through again, this time checking details such as delivery options and the costs associated.
  • The very next day you are served an ad on the Google Display Network, offering a 5% discount on the watch you have been viewing. You click through once again, but decide that even with the discount applied you may have to wait until your next paycheque before committing to the purchase.
  • The following week, pay day comes around and you decide that you want to purchase the watch. You navigate directly to the website by typing the URL into your browser, before finally making the purchase.

Your purchase journey would then appear as such when plotted (See figure 2).

Now let’s shift roles from customer to marketing analyst and see exactly how the value gets assigned under each attribution model.

Last click attribution

The last click (or last interaction) model is one of the most commonly used attribution models and the default option on most analytics platforms (certainly within Google Analytics.) With this model, the ‘credit’ is given entirely to the last touchpoint the user had with your website before they purchased, whilst not accounting for any previous interactions (See figure 3). In this case, direct traffic would be reported as the conversion driver, with the ads on Instagram and Google Display receiving no credit whatsoever.

Benefits of this model

Apart from having it applied by default, you would typically associate using last click when your primary/only goal is to drive action. Therefore, the marketing channel responsible for driving the action will get assigned the entirety of the credit.

Limitations of this model

Even if your goal is to drive action, you’ll still be missing visibility of the value for earlier interactions and the other marketing channels. In our example, would you (as a customer) have ended up buying the watch if you hadn’t have seen it first on Instagram? As a marketer, you would therefore have no insight or reason to invest more into Instagram advertising, when actually it could be the key to achieving significant sales growth.

First click attribution

The first click (or first interaction) model does the opposite of the last click attribution model by giving 100% of the credit to the first interaction (See figure 4). In this instance the first Instagram ad would be given the credit for the purchase of the watch, as it was the first interaction you (the user) had with the company’s marketing initiative.

Benefits of this model

If your goal is to attract new customers to your website, then attributing the eventual sale to the channel that first caught their eye makes logical sense.

Limitations of this model

As once again the full value is attributed to just one source, you lose the visibility of how effective the channel interactions later down the journey.

In our example, you could theorise that if you (as the customer) did not see the discount code via the Google Display ad, you may not have solidified the intention to make a purchase at a later date. Therefore, as a marketer, you may not be inclined to invest more budget into Google Display advertising whilst missing out on an opportunity for overall sales growth.

Last non-direct click attribution

This model is very similar to the last click model, however the last non-direct click (or last non-direct interaction) attribution model gives credit to the penultimate channel for any user journeys that ended with them coming directly to the website (See figure 5).

In our example journey, even though the last interaction before purchasing was via direct traffic, the Google Display ad would be awarded the conversion as it was the last non-direct interaction.

Benefits of this model

As a marketer, you have reduced control over direct traffic. For the most part, these users are already familiar with your website and/or brand at this stage in the journey. It is also unlike the marketing channels at play, where budgets and costs and distinct user targeting are involved.

Therefore, if you are looking to drive action via the marketing platforms you have full control over, attributing the value to the last one featured in the journey is logical.

Limitations of this model

Just like with the last click and first click models, this is another example of single-touch attribution and therefore provides no visibility of value to the other channels that occurred elsewhere in the journey.

Linear attribution

The linear attribution model is the first in the list that demonstrates multi-touch attribution, meaning that it ‘splits’ the conversion amongst the channels to show value where due. This particular model gives equal credit to every touchpoint that the user interacted with prior to converting. For instance, if there were 10 touch points, each of them would receive 10% of the credit (See figure 6).

In our example, we would see each interaction be ‘awarded’ 25% of one whole purchase. In fact, as the Instagram ad was interacted with twice in the journey, the attribution report would show Instagram advertising as 50% ‘responsible’ for driving that purchase.

Benefits of this model

If you wish to begin using multi-touch attribution and be able to quickly get up-and-running with optimising your marketing experience across the entire breadth of your activity, going with a Linear model is a good way to go.

Limitations of this model

Whilst you’ll receive a better picture of each of the channels in the mix, it won’t provide any indication of which channel is most effective at driving the end result. In our example, Instagram received a 50% share of the conversion because it appeared twice in a journey that was four interactions long. The Google Display ad may ‘deserve’ the majority share however, as it was at this point in the journey that the intention to buy was solidified.

Time decay attribution

Like the linear model, the time decay attribution model provides credit to each and every channel that drives a customer closer to the actual conversion, however the value assignment will be weighted according to the how recent the channel featured in relation to the conversion interaction (See figure 7). In our example, the direct interaction would receive the largest amount of the conversion ‘value’ as it was the channel that drove the conversion. The other channels then get a smaller amount, which diminishes as we go back in time to the first interaction.

Incidentally when we look at the attribution report, Instagram gets the same amount of credit for the conversion as Google Display, due to it appearing twice at the beginning of the journey.

It is important to consider that the credit that each channel will be assigned in every single user journey will depend on the amount of interactions and the time in-between them.

Benefits of this model

Time decay essentially gives you ‘the best of both worlds’ of the last click and linear models. Focus is put on the last interaction that drove the conversion, whilst providing visibility of the steps in the journey that came before. It is a popular choice for businesses with long sales cycles, especially for high-value B2B services or products.

Limitations of this model

Even though more visibility is granted, as the first interaction is receiving the lowest amount of credit for the purchase, it’s true value may be being suppressed. Think about our example journey – would you have bought the watch if you hadn’t have seen the first ad on Instagram?

Position-based attribution

The position-based attribution model combines the features of the linear, last-click and first-click models. This model gives the most credit to the first and the last touchpoint, which both receive 40%, whilst the touch points in-between receive an equal piece of the remaining 20% pie (See figure 8).

In our example, the first interaction (Instagram ad) and the last interaction (direct) each get a 40% share, with the two interactions in the middle both receiving 10%.

The acquisition report would therefore show the following:

Benefits of this model

You could say that this model combines the benefits of the three other models:

  • Weighting on the first interaction, how new customers came to be aware of your business
  • Weighting on the last interaction, the channel that drove the action
  • Visibility of value for all of the interactions falling between

Therefore, if your goals include attracting new customers, driving action and achieving growth simultaneously (for instance, if what your business offers is a single-use service or product) this model will fit the requirement.

Limitations of this model

As detailed in some of the other multi-touch models, your most effective channel may lie in the middle-interactions. From our example, if the Google Display interaction was the true motivator, under this model it would only be getting one quarter of the credit that direct traffic receives. As marketers, we would therefore have a dataset that wouldn’t necessarily support the idea of increasing the Google Display advertising budget in exchange for a better return.

Custom attribution

A custom attribution model is, as the name suggests, a model which is tailor-made by you. It is one which is deliberately designed & utilised to custom-fit the customer journey of your business, whatever that may be. Therefore, how conversion credit is assigned is based on a set of rules that you define entirely.

Benefits of this model

When set up correctly and properly fine-tuned, a model that considers the intrinsic details of your business’s own purchase journey will provide the highest level of accuracy when it comes to assigning conversion credit across your marketing channels.

You will also be able to factor in aspects that the default models do not consider, such as weighting conversion value based on how engaged a user was with your website on each visit, or which channels drove a user to completing a pre-purchase milestone like booking a consultation or test-drive, for example.

Limitations of this model

As custom attribution modelling isn’t available in some of the most widely-used ad platforms (such as Google Ads, Microsoft Advertising, Facebook etc.) you may have to invest in a dedicated attribution platform to properly collect and present your data. There’s also a challenge in that creating a custom attribution model takes a great deal of time and expertise to accomplish.

You also need to ensure that your data collection is to a very high level of accuracy, and has been at that level for quite some time. Any misreadings or unintentional inflation in conversion data will dampen the effectiveness of your carefully planned out custom model.

Data driven attribution (DDA)

Data Driven Attribution (often referred to as DDA) differs noticeably from the rest of the attribution models on this list and is arguably one of the more effective methods of multi touch attribution. Instead of conversion credit being assigned to channels based on a set of rules, DDA uses machine learning to analyse all of your customer journeys. It then determines how effective a channel has been at driving a conversion by analysing the journeys where it has and hasn’t appeared at certain points along the way (See figure 9).

Going back to our original example of the watch purchase, under a DDA model all of the purchase journeys would be examined. To determine the value that gets assigned to Google Display as a channel, all of the user journeys where it appeared would be compared to those where it didn’t, observing the likelihood of purchase for each. If it is determined that a user is more likely to convert when Google Display has featured mid-journey, it will get a heavier weighting of conversion credit.

This functionality makes it a very useful multi-channel attribution model, bringing some of the benefits of custom attribution without the technical difficulty.

Benefits of this model

As it is driven by a machine learning algorithm, no rules need to be defined and created. It is also based on data generated by your own activity, and is therefore custom-fit to your own business. It is also readily available for use within the Google Ads platform with minimal setup required.

It also works very well within Google Ads when combined with a well-trained smart bidding solution.

Limitations of this model

To be able to use this model, you will need to hit a data threshold first. Within Google Ads you’ll need at least 15,000 clicks, along with 600 conversions, per conversion type, within a 30-day timeframe. This isn’t good news for low traffic websites or smaller, low budget advertising accounts.

It also is not a feature of the standard Google Analytics, appearing only in Google Analytics 360 at the time of writing. Therefore, you may need to invest in an alternative attribution platform to apply this model to your entire range of marketing activity.

B2B vs B2C

A question that commonly occurs is whether attribution modelling should differ based on whether your business supplies products/services on a B2B (business-to-business) or B2C (business-to-consumer) basis. There are in fact no set rules, but there are certainly some considerations that should be front of mind:

Time lag to conversion: how long it typically takes a user to move from first interaction to purchase.

Milestones/micro conversions: does the journey to purchase involve a free trial or a consultation before a purchase is made? And if so, how should the channels that drive these pre-end goal actions be weighted?

Repeat purchase likelihood: is a user likely to purchase your product or service regularly, occasionally or just the once?

Once realised, you should have a clearer picture of which model will prove to be most useful.

The bottom line

The truth is this; there is no ‘right’ or ‘wrong’ attribution model for your business. Ultimately to help you make that final decision, ask yourself the following question:

“What is the end goal of my marketing campaign and which model will make the most effective measuring gauge?”

Know that each model has its pros and cons and each work well in their own right, and in different situations.

You may however find that due to the complexities of your business, your marketing activity, or the sheer volume of data that you capture, there is added difficulty to this learning process. The third article in this series explores how to tackle these challenges, along with how to identify opportunities for growth.

Photo by Carlos Muza on Unsplash

Take our simple Digital Maturity Assessment to see where your business should focus next.