The Google Analytics attribution modeling tool has been around for a long time. Around 2014 a similar tool was introduced to Google AdWords as well.

However, up until this point you could only analyse the impact of different models but you were stuck with last-click attribution.

Since May 2016 you can actually change the AdWords attribution model that your conversion tracking is based on.

So what are the different attribution models? And what’s the best AdWords attribution model?

The Five Different AdWords Attribution Models

There are five attribution models to choose from:

Last Click (default), First Click, Linear, Time Decay, Position-Based.

Below is a screenshot from the AdWords interface showing how a conversion is credited under the different attribution models:

AdWords Attribution Models

For an in-depth explanation of attribution models, read this article.

There’s a sixth option called data-driven attribution if you meet the minimum conversion requirements.

In order to use data-driven attribution for a conversion action it needs to have at least 800 conversions and its search campaigns need to have generated more than 20,000 clicks in the last 30 days.

Note: If you change your attribution model it will only affect clicks on the Google Search Network and Shopping Ads. All interactions on the Display Network are still tracked based on last-click attribution.

How Each Attribution Model Tracks Conversions

The typical research and buying process online involves several searches, ad clicks and visits to a website before the actual purchase. The longer the buying cycle the more interactions happen before a conversion.

The challenge with picking an attribution model is how to value all of the clicks (interactions) that happen before the conversion.

Last click attribution gives 100% of the credit to the last click in the buying cycle.

First click attribution gives 100% of the credit to the first click in the buying cycle.

Linear attribution gives each click in the buying cycle the same amount of credit (e.g. if there are five ad clicks before a conversion, each of them would receive 20% credit).

Time decay attribution also distributes the conversion credit over all of the ad clicks, however clicks closer to the conversion receive more credit than earlier clicks in the buying cycle.

Position-based attribution gives 40% of the credit to both the first and last click. The remaining 20% are distributed evenly over all of the clicks that happen in between the first and last click.

AdWords Attribution Models and Campaign Performance

I have put together a case study to illustrate the performance implications of each attribution model for two of my AdWords campaigns.

One campaign is targeting generic search terms and the other campaign is targeting brand keywords.

The table shows the differences in conversion performance under each attribution model.

Last click attribution shows the lowest number of conversions in the generic campaign and the highest number of conversions in the brand campaign. This is because people search for generic keywords in the earlier stages of the buying cycle whereas brand keywords are used close to the conversion.

Obviously first click attribution must show the opposite trend. The generic campaign shows the highest number of conversions whereas the brand campaign shows the lowest number of conversions.

The conversion numbers for linear attribution are in between first click and last click attribution.

Time decay attribution reports similar conversion numbers as last click attribution. This is because the conversion credit is heavily skewed towards the end of the buying cycle (i.e. the last clicks receive more credit).

Position-based attribution reports conversion numbers that are roughly in between last click and first click. Since 80% of the credit goes to the first and last click conversion numbers are heavily skewed towards the average of these two models.

AdWords Attribution Models and Keyword Performance

The illustration above shows how overall campaign performance changes based on different attribution models.

Let’s now look at how the different attribution models affect keyword performance within the campaigns.

Last click attribution tends to favour specific, “buying” keywords which are very close to the conversion (for example brand keywords or product names). Keywords which are at the beginning of the buying cycle (broader, more general keywords) will receive less credit.

First click attribution will show the opposite trend: lower cost per conversion figures for keywords at the beginning of the buying cycle and higher cost per conversion figures more “buying” keywords close to the transaction.

Linear attribution will give equal credit to all ad clicks which happened before a conversion. If five keywords received an ad click, linear attribution would allocate “0.2 conversions” to each keyword. This can be useful for bid management if you want to make sure that all keywords which led to a conversion receive credit.

Time decay attribution will give credit to all ad clicks before a conversion, however clicks closer to the conversion will get significantly more credit than clicks that happened earlier in the buying cycle.

Position-based attribution will give 80% of the credit to the first and last click in the conversion path. All other clicks share the remaining 20% conversion credit.

Implications of AdWords vs Analytics Conversion Tracking

There are two main ways to track performance in Google AdWords:

1. Google AdWords conversion tracking
2. Importing goals or transactions from Google Analytics into Google AdWords

All of the advice in this article assumes that you are using Google AdWords conversion tracking (instead of the Analytics import).

There are significant differences between AdWords conversion tracking and importing Analytics goals/transactions.

Importing Analytics goals/transactions into AdWords will typically result in much lower conversion numbers than using AdWords conversion tracking (for the exact same conversion action!).

Here’s why:

Both tracking systems use last click attribution, however AdWords only takes into account AdWords traffic while Google Analytics considers all traffic sources.

If the last (non-direct) click before a conversion is not from AdWords (for example referral or organic), your Google Analytics import will not show a conversion.

So with respect to AdWords optimisation, it only makes sense to import Analytics goals if your approach is advanced enough to optimise your traffic acquisition across all channels.

What’s The Best AdWords Attribution Model?

There really isn’t an easy answer to this.

From my perspective, both the first and last interaction in the buying cycle should receive special credit.

The first interaction is how people find out about your business. The last click before a conversion is the one that seals the deal.

If you are running a brand campaign in AdWords it will often receive the last click before a conversion. In that case the second-to-last click is typically more important because it ultimately led to the search for your brand.

Both last and first click attribution don’t give credit to clicks that happen in the middle of the buying cycle. If these clicks really didn’t have any value, why do they need to happen to generate a sale in the first place?

Linear attribution is equally wrong by giving the same amount of credit to every click.

In my opinion both time decay attribution and position-based attribution are the best options.

Overall I’m slightly in favour of position-based attribution since time decay attribution gives very little credit to the first click in the buying cycle.