What is attribution, really?
Attribution is all too often placed in the sceptical marketer’s crosshairs. The accusation? The impossibility of 100% attributability. That is, the impossibility of accounting for every single ‘touch’ on the customer journey.
Data privacy regulation, offline touches, even human nature itself, are all hoisted up the anti-attribution flagpole. And with good reason. These all present grey zones in the customer journey that together make it difficult - read impossible - to attribute towards every factor impacting the buying decision.
In 2016 Sergio Maldonado made a passionate case against attribution on chiefmartec, where he systematically plotted all attribution’s shortcomings. Most of which continue to be echoed today.
Unfortunately, these arguments, although compelling, fight an attribution strawman.
So, in this article, we’re asking, and answering, “what is attribution, really?”, how can we define attribution?
In doing so, we’ll also assess whether developments over the last five years have indeed led to the attribution segment forever treading in the “trough of disillusionment” or whether attribution is now confidently heading towards the “plateau of productivity”.
The article is broken into four main sections:
Identify attribution’s perfect imperfections
Debunk the myth of attribution as a method of absolute causality
Delimit attribution to when and how it works best
Assess how attribution has overcome the biggest obstacles standing in its way in 2016
Attribution: perfectly imperfect?
The brunt of critique levelled at attribution centres on its inability to encompass every single factor that has a bearing on the buyer decision.
After all, you will never really know what actually made the customer convert without reading their mind.
And because you can’t do this, the argument follows, why use attribution at all?
Attribution, viewed like this, is certainly not the “Holy Grail”.
But this doesn’t quite mean attribution is the poisoned chalice made out to be either. Instead, attribution, in the right context, can be an ornate cup that is pretty effective in quenching users’ thirst for better performance.
However, when considering the broader scope of marketing performance, especially when it comes to offline channels or factors outside the realm of digital touchpoints, Marketing Mix Modeling (MMM) can be a valuable complement to attribution. MMM allows for the evaluation of both traditional and digital channels' impact on business outcomes, incorporating data from broader marketing efforts like TV, radio, and print, where attribution models may fall short.
Much like René Magritte’s “The Treachery of Images”, the distinction depends very much on the perspective taken when evaluating attribution.
In this way, attribution can either be imperfect or perfectly imperfect depending on the perspective taken when assessing it. To use the pipe analogy, it can be scorned as a complete failure in not being a pipe, or adulated for the brilliant representation of a pipe that it is.
So, if you’re assessing attribution on its ability to take into account every single factor in the buyer journey, it will be found wanting.
Similarly, if you’re expecting attribution to deliver under all and any condition, it will continue treading the trough of disillusionment.
The good thing is that attribution doesn’t, or at least shouldn’t be, claiming to do this.
Debunking the myth of causality
At the core of this critique sits the causality myth, where attribution can only work when there is (absolute) causality, i.e. input A produces output B.
But this is of course impossible, as, once again, we cannot track every factor that has a bearing on the buying decision.
Attribution does not, however, aim to achieve this absolute cause and effect, which makes using this myth as a yardstick misleading.
Instead, attribution models are built to account and compensate for the missing data and causality. With different attribution methods dealing with causality and missing data in different ways.
Yet, even with perfect data, standard mathematical models work with correlations not causation. All common attribution techniques like rule-based, probabilistic, Markov Chains, medical survival-analysis, game-theory or any of the many machine-learning classification techniques, incorporate some form of correlations.
Do variables outside the correlation play a role? Absolutely.
The assessment of how usable the outcomes from the attribution modelling are depends on how well you know your business data. Even though there might be important touchpoints that you could not measure, knowing which things are actually part of the revenue-generating process, and maybe more importantly which are not, is still crucial information to driving the business.
As George Box puts it, “all models are approximations. Essentially, all models are wrong, but some are useful”.
It is always up to the human to interpret and action the model and the data upon which it is built. Attribution is no different. Ultimately, an attribution model is an answer to a question. It is still - for the time being at least - up to the human to ask the right question.
Which brings us to this pithy John Tukey quote:
“An approximate answer to the right problem is worth a good deal more than an exact answer to an approximate problem”
Provided you understand how to ask your attribution model(s) the right questions, the answers will still be gamebreakingly useful for you to drive your business in the right direction and generate revenue - which is ultimately what it’s all about.
Delimiting attribution to when and how it works best
With this said, there are situations where access to data is so limited that even the statistical approximations that underpin attribution become insufficient in generating worthwhile insights. So, to assess whether attribution has moved on from the trough of disillusionment, we need to understand when and how attribution can actually deliver.
Here, the alignment between data sources, business model, and intended use becomes a precondition for attribution actually bearing fruit.
To put it another way, certain conditions allow you to better understand how to ask your attribution model(s) the right questions and get from them those gamebreaker answers. If you want to learn more about what enables deeper insights about your customer behavior and campaign performance, you would have to look into your Marketing Analytics. By analyzing data across multiple touchpoints, marketers can leverage marketing analytics tools to make more informed decisions, ensuring that their attribution efforts are grounded in actionable insights.
Attribution when?
Attribution works best when there is greatest access to Digital Marketing Analytics. Digital Marketing Analytics refers to the collection, measurement, and analysis of data from digital channels like websites, social media, paid ads, email campaigns, and more. It provides the raw data needed to apply attribution models effectively.
After all, nothing which relies on data will ever be able to offer insight that isn’t in the data in the first place. With the exception of statistical methods compensating for missing variables, if it is not in the data, attribution cannot tell you about it.
So for instance, if only 10% of your business’s customer journey takes place digitally, attributing that 10% will not offer much value.
So far, it’s been certain business types, such as B2Bs SaaS with their closed-loop customer journeys, who have been able to benefit most from attribution.
But, the digital transformation and the ceaseless growth in tools is helping more business types migrate their touchpoints digitally. Indeed, the Covid-19 pandemic is having a notable impact here. As this Mckinsey 2020 report highlights, the Covid-19 pandemic has made the digital transformation take “a quantum leap at both the organizational and industry levels.”
Attribution how?
Having access to digital touchpoints obviously doesn’t in itself deliver attribution. Data collection and processing best practice need to be in place to ensure the data gets to the modelling. Similarly, the attribution algorithms themselves need to be tailored to the business data.
Specifically, data needs to be collected from on-site user behaviour tracking (using a CDP or other javascript and UTM mapping), and the tools on your tech stack (CRM, ad platforms, etc.). Then the data processing and cleaning needs to take place, i.e. deduplication, merging, removing unwanted data, etc.
When it comes to the attribution modelling itself, there are also important considerations for the algorithms being put into place. Namely, how you’re defining and modelling the parameters for different activities and metrics, e.g. conversion goals, channel breakdown, etc.
Without these elements - and they will differ depending on the unique business needs - attribution cannot deliver.
This brings us back to our central thesis, that assessing attribution as an ‘all things to all men’ category can only result in disappointment. Instead, assessors and users alike need to understand that enough data needs to be available and adequately processed and modelled for attribution to provide any practical utility.
Evaluation: how has attribution overcome the biggest obstacles standing in its way since 2016
Now that we have adequately qualified attribution, both in what it theoretically aims to achieve and when it’s practically viable, we can assess whether attribution has overcome the obstacles that stood in its way in 2016. Namely, the tightening of data privacy regulation, the cross-device challenge, and the opportunity cost involved.
Lifting the veil of digital privacy
While GDPR, and other data privacy regulation - whether institutional (e.g. CCPA, GDPR) or browser-led (e.g. ITP) have made an impact on which data is collected and how, it has not proven to be the coup de grace they were anticipated to be.
First-party data has become the currency of attribution tools, which even in the permission-based era, continues to bear fruit.
This comes on the back of a number of workarounds that have been put into place over the last five years:
Owning customer data and tracking, by moving away from reliance on ad platforms and towards CDPs.
Placing IDs in browser cookies as well as in local storage, to avoid this being deleted after 24 hours.
Running identify events earlier and more often, gating more content behind email submission or prompting users to log in to your product.
Operating less across domains, that is, opting for yourdomain.com/landingpage instead of the sub.yourdomain.com, and where unavoidable, making sure that the ID is forwarded from one domain to another.
Most attribution tools now enable this tracking and identification out-of-the-box.
Cross-device - a challenge overcome?
Although, as this recent emarketer report shows, the cross-device challenge continues to rank high in marketers’ concerns, it has come a long way since 2016.
Here too we can find solace in the explosion of not only CDPs and DMPs, but also CRM and automation tools. These tools are helping suppress the cross-device challenge through advanced tracking and identification processes, such as UTMs, form and other events identifys.
By pulling data from across sources and tools in the ecosystem - from start to end of the journey and beyond - attribution tools are well placed to ensure the cross-device challenge is as limited as possible.
What is attribution when it comes to opportunity cost?
But even where the business is able to gather enough of the right data to draw useful outcomes, is all this effort too much? What’s the opportunity cost for attribution today?
The central contention here is that if you implement attribution you forgo other investments, i.e. you incur an opportunity cost.
If we assume that the opportunity cost assessment is being made by a business with significant access to data - such as B2B SaaS - attribution today should be a no-brainer.
Over the last five years, the cost and time involved in setting up an off-the-shelf attribution tool has been reduced significantly. Plug-and-play attribution tools with freemium options make attribution readily accessible and testable.
Of course, even with this offering, once we’re outside the archetypal attribution benefactor (B2B SaaS), the opportunity cost becomes greater. But it does so in a subjective manner, where the unique circumstances of the business - again, think data and intended use - define the potential benefit.
Additionally, we see the opportunity cost significantly spike where the option to build a tool in-house is being considered. In this case, the time and effort needed from an engineering team to arrive at a solution on par or better than off-the-shelf options is significant.
Evaluation: plateau of productivity?
Since 2016 attribution has continued to feature prominently in digital companies’ setup. And, as this emarketer chart shows, the trend is set to continue. You can find the full report here.
Yet, according to Gartner, multi-touch attribution is now firmly treading the trough of disillusionment and is some 2 to 5 years away from the plateau. This paints an important picture and one which reconciles with our analysis above. Attribution’s descent from the peak of inflated expectations no doubt reflects the fledgling realisation of attribution as a correlative methodology whose utility is delimited by data availability and processing.
It will only be through altering our thinking of attribution as a method of absolutes, and understanding where its application is best suited that we can start on the slope of enlightenment. Otherwise, it’ll continue setting the user up for disappointment and disillusionment. Much like Magritte’s pipe illustration, attribution will prove woefully inadequate for smoking tobacco.
Marketing attribution software list
Here you have a list with some marketing attribution software options you can consider:
Dreamdata: a multi-touch attribution platform that helps businesses connect revenue to their marketing efforts.
Measure and analyze performance across all your revenue-generating activities with the Performance Feature. Maximize effectiveness, improve ROI, and cut waste.
2. Bizible: a B2B marketing attribution software that helps marketers track their customer's journey and measure the impact of their marketing campaigns.
3. Marketo: a marketing automation software that provides insights into which marketing channels are driving the most revenue for your business.
4. Impact: a marketing attribution software that helps businesses track and measure the effectiveness of their influencer marketing campaigns.
5. Wicked Reports: a marketing attribution software that tracks customer behavior across multiple channels and provides insights on which channels are driving the most revenue.
6. Attribution: a marketing analytics platform that provides insights into which marketing channels are most effective and how they are contributing to revenue growth.
7.TapClicks: a marketing analytics platform that helps businesses connect marketing data across multiple channels and make data-driven decisions.
8. Full Circle Insights: a marketing attribution software that provides insights into which marketing campaigns are driving revenue and how to optimize them.
9. BrightFunnel: a marketing attribution software that helps businesses understand the impact of their marketing efforts on revenue and pipeline growth.
10. Convertro: a multi-touch attribution platform that helps businesses track and measure the impact of their marketing campaigns across multiple channels
Related articles
https://dreamdata.io/blog/solving-the-marketing-attribution-puzzle
