What is it?
Before we dive deep into Data Driven Attribution (DDA), let’s first clarify what attribution actually is. Attribution modelling is a pretty broad topic, but to sum up, it is a process involving the distribution of conversion credit to various keywords, ad groups and campaigns. To fully understand our users, we need to be able to give credit to these elements as fairly as possible.
DDA is a model which allows us to do exactly that - accurately weighting different touchpoints along the conversion journey depending on their incremental impact. To paint the picture, some of the other ways of attributing conversion credit are broken down as follows:
The default model in SA360 (Search Ads 360 - formerly known as DoubleClick Search or DS), where credit is given to the last click before conversion. Any interactions before the last click are disregarded, as full credit is given to that last click before conversion.
Equal credit is given to every touchpoint. A slightly more accurate model, but limited however, in that it assumes that each touchpoint should be given equal credit, which is not necessarily the case.
When using DDA in Search Ads 360, the model is applied to conversions in the advertiser, giving us a clearer view of how keywords and other biddable elements contribute towards a conversion. These are then weighted, depending on how much or how little impact was seen from each touchpoint. The model is completely tailored to our own data, creating a personalised model for each advertiser, and continually learning.
Using DDA in SA360 therefore, is much more sophisticated, analysing interactions between our campaigns and distributing credit based on the incremental impact an interaction has on the user’s likelihood to convert.
DDA in SA360 also allows us to factor in interactions between Paid Search, GDN, Programmatic Display, Google and Bing, meaning we can gain a much more holistic view on how people are interacting with our brands across the web.
I'm in - how do I set it up?
So, it makes sense why we should use it, but how do you set it up in SA360? The process often looks complicated, but in fact is very simple.
1. Create channel groupings
SA360 uses these to identify the types of interactions you’re interested in tracking in the model. Here, you will categorise your keywords/ad groups/campaigns, based on their position in the conversion funnel, and this can be done in two ways:
Custom label groupings
Create labels and apply them to campaigns, ad groups or keywords, ideally 24 hours before you create your model. As DDA models work with the previous 24 hours’ worth of data, this ensures that once established, the model can work right away. It also helps SA360 to be certain of the association, and makes sure any changes made do not negatively affect the model.
This allows you to categorise your account in a more complex way, with the possibility of up to 15 different channel groupings. This is useful for when you want to create a unique, tailored model, such as if you want to look at how devices are over or under-attributed, or considering DSA and Shopping activity.
The table below suggests ways to group channels by vertical, however you could also try grouping by:
- Shopping campaigns
- Audience specific keywords
- Device groupings
The more you split your channels, the more credit is segmented between interactions along the path, which could be problematic for the model. For this reason, no more than five channel groupings are recommended.
Custom grouping labels need to allow for new activity to fit under a label, as once the model is created, new labels cannot be added in. If you are regularly adding new items to the account, you need to make sure they are added to the model, so creating a couple of blank labels is recommended, so that these new items can be added into the model.
Example channel groupings by vertical:
Auto Channel Grouping
A relatively new feature in SA360, allowing keywords to be grouped automatically by brand or generic
- You’re new to DDA and want to see how it works before creating custom, more tailored channels
- You are constantly adding new keywords and campaigns to the account, as you need to make sure they are added into a label to include in the model, so using auto-grouping may be the better choice
Not so great if:
- You are running Shopping and DSA activity in the account - this does not factor in any DSA or Shopping activity in an account
- You want to report on each custom label that you created
2. Pick your Floodlights
As many as 50 different Floodlights can have their interactions analysed and reported on by SA360. Google Analytics goals can also be used if synced with SA360.
Pick a floodlight depending on its relation to the conversion funnel you want to look at:
- If you want to look at two different types of conversion funnels, best practice would be to create two separate DDA models, and choose the corresponding Floodlight for that funnel
- Be mindful that SA360 will need conversion volumes to be above a certain threshold for the DDA model to work effectively, therefore it may be worth combining two conversions in one, providing the conversion funnels are similar enough
3. Apply your Channel Groupings
- Using the labels you previously created to form custom channels OR, let SA360 do the work with auto-grouping
That’s it! Click save, and within 24 hours your model will be ready to use in the interface. You will be able to see 60 days’ worth of backdated data once the model is ready, meaning you can start using it straight away. As a point to remember, DDA currently only updates once a day, which means when looking at your data, you should avoid including today’s data.
Now What? Creating columns
Now your model is ready, you can start creating your columns to appear in the SA360 interface, allowing you to view your data. Some good places to start are:
- The Floodlight you’re tracking
- Conversion Rate
- Formula columns to track discrepancies between DDA and default attribution models. A handy formula for this is as follows:
This formula can also be replicated for other KPIs you’re interested in, such as CPA or CVR.
Before saving a column, you will see a tick box for including cross environment conversions, the SA360 equivalent of cross-device conversions. If selected, you can report on the model, but not use it in bid strategies. One way to get around this is to create both – one for reporting (including cross-environment) and one for using in bid strategies (not including cross-environment).
Once the columns are set up, you can then create Views in SA360 to look at the DDA and discrepancies between other models.
To let DDA work as effectively as possible, at least 600 conversions of the conversion action you are tracking, and 15,000 clicks are needed over the last 30 days.
If you don’t meet the above requirements we would recommend Linear as the next best model to use!
Why use SA360 DDA over AdWords?
SA360 lets you create custom models with labels, but AdWords does not have this custom functionality.
In SA360, it's possible to view and compare both models in the interface. In AdWords, it's only possible to see one type of conversion at one time, meaning whichever model you choose will be the only one you can view the data in.
SA360 DDA can be included in an SA360 executive report, alongside other models. This gives more visibility on how the models may differ, and makes it much easier to feedback to clients or stakeholders.
Using DDA in Bid Strategies
Once your DDA model is set up, you can start using it in SA360 bid strategies. As mentioned earlier, the column cannot include cross-environment conversions when using in a bid strategy, so keep that in mind! To see some of the great work we’ve done using DDA model in Bid Strategies, click here.
Read this case study to find out how we used data-driven attribution in Search Ads 360 to boost Walks of Italy's ROI by 25% year-on-year.