Solving attribution and the limitations of ROAS

I’m on another attribution project this week which admittedly, feels more complex than most, while at the same time, not being so complex at all.

 

I found myself reflecting on why that was exactly, and I’m coming to realise the biggest battle attribution faces, is in smashing apart the high level “hard problem” and reframing it back to the person who is asking the question.  

 

I’ve said it before and I’ll say it again.

 

Attribution is not a destination. 

 

It’s not an end state. It’s an infinite game and it’s the noun itself that tricks us into believing we could ever find a simple solution to a complex set of questions. 

 

So how do we smash it up and break it down? I’m glad you asked. 

 

I now have a decision framework for this.

 

1) Stakeholder, Question & Decision: 

 

This means; who is asking, what are they asking and what decision needs to be made. 

 

To the purists, the true definition of attribution is the outcome of spend which is to be distributed by one dependent variable (usually sales or the equivalent). Channels get their cut based on the business rules that we think make the most sense for our customer and market, and off we go. 

 

Those that talk to marketers as I do though, know that this simply does not meet the need. Marketers need granularity. Specifically, they need to cut the same insight by campaign type, audience targeting (e.g. winback customer campaigns), promotions/offers, business-as-usual performance activity, the customer lifecycle and their lifetime value, new customers acquired, retained and more. In fact, a few weeks ago I offered a framework to organise this. It’s called the OPTICS framework. Read up on that here. 

 

Step 1 in assessing the “hard problem of attribution” is to isolate the specific question we are answering. I’m cheating a little bit, as it’s through asking the question of the individual who requires the report that you get an inside look into their strategic priorities. 

 

Here are a few example questions I hear all the time. 

 

As a Marketing Manager, how should I reallocate my existing budget across above the line and digital to make the most efficient use of the current budget? 

 

As a Channel Manager, how can I understand how much more life my channel has, up until the point of diminishing returns to further grow the impact my channel has on our funnel? 

 

As a Marketing Manager, how do I guide my agency to distribute budget across digital channels and the campaigns they are deploying within them (e.g. acquisition vs retargeting)? 

 

We need to get more specific with the question we are asking the the decision to be made. Then we can find the right methodology to suit based on this and the information to follow. 

 

2) ROI on information 

 

This one requires a bit of guess work but in my experience it really does help to hone in on which attribution framework makes the most sense for an organisation to adopt (or at least where to start and build from). As I already highlighted, there will be no shortage of questions. It will never stop. 

 

Sitting with marketing managers as an analyst feels kind of similar to sitting with that great aunt you haven’t seen all year at Christmas (the questions never stop). 

 

 

This stage is about asking your stakeholder to think commercially about the return they hope to achieve from gaining the answer to their question. Keep in mind, we will soon move towards understanding the feasibility and therefore complexity and cost associated with GETTING that answer which will further contribute, but for now this is about asking the marketer to think about what specific actions they would take and how that might support improved results. 

 

This is the kind of answer I am looking for: 

 

“Digital currently makes up 80% of our total spend. We spend very little above the line at the moment, so even visibility into our digital channel performance and reallocation of that spend alone could yield us a 15 – 20% improvement on ROAS.” 

 

Keep digging (and supporting) until you understand what specific area of their strategy and efforts will yield the highest return. If all else fails, get in there and perform a discovery to understand the distribution of spend and conversion. 

 

 

 

3) Data & tech feasibility 

 

Now you know where the focus NEEDS to be in order to yield results that will actually move the needle, you can move into a high level data and tech feasibility. 

 

This means understanding what is being tracked already and what technology is at our disposal without any further investment. 

 

To further contextualise, here are some things I’ve seen: 

 

• Some organisations are rich with orderly historical data. They’ve been with the same media agency for years and running a media mix model (MMM) is the natural next step.
• Some startups, non for profits (who often experience many pro-bono, changing agencies) or low maturity organisations aren’t even collecting the transaction or conversion data that would be required to run a meaningful model of any kind
• Some are spending so much in paid digital media, even getting conversions sent back into platform for walled-garden media attribution for auto-optimisation is the fastest win and lowest effort solution without boiling the ocean

 

 

 

 

4) Skills to maintain

 

The final and arguably most important element is ensuring the team you are working with have the capability to adopt and maintain whatever you support them with. This means not only ensuring they have the hard skills (whether upskilling on digital analytics tools, access to digital analysts or engineers) or the soft skills and data literacy needed to simply interpret the data and understand how best to action a decision. 

 

Without understanding and matching this, regardless of the solution you choose, it won’t support them to achieve that ROI at all, so match that maturity. 

 

 

 

If you would value a future post on mapping specific attribution questions and decisions to be made to the best fit attribution methodology – “hit reply” and let me know. I’ll include it in a future newsletter. 

 

 

 

Literacy foundations

 

 

 

Return on Advertising Spend (ROAS)

 

 

 

 

 

Return on Advertising Spend (ROAS) is a metric used largely across digital channels to measure the effectiveness of advertising campaigns. It calculates the revenue or conversions generated by an advertising campaign divided by the spend that was incurred for that campaign to run. 

 

As a metric used from the perspective of a channel owner on channel performance, ROAS provides insight into how well constructed the channel strategy is at driving the intended action. 

 

This however is also its limitation, as to calculate this, spend is limited to that of the channel that is being measured and the attribution method that offers a single channel a fixed allocation of credit for driving revenue. The final calculated metric is one that indicates performance in review (historical, look back) rather than a forward looking (predictive) forecast. 

 

As a high level guiding metric, ROAS can provide an indicator that at the very least, a channel is meeting a threshold that ensures that it on the whole, is contributing to a positive outcome for the investment being made. However, it is due to this that it can run the risk of misleading the observer.

 

 

 

Limitations of ROAS

 

• Does not provide insight into the distribution of spend or the return. For example, there may be few customers within the whole target set that drive the majority share of return with high basket sizes. On the contrary, there may be many customers who contribute small amounts to the bottom line. Therefore, ROAS alone does not provide an indication of the efficiency of the campaign set up, or the cohorts of users who are engaging in the channel when used as a metric to compare channels.
• It only provides a report card metric of the historical performance of a single channel and therefore, is not a suitable metric to compare omni-channel performance. If marketers are utilising multiple channels and assuming it requires multiple touches and messages to drive an individual to convert, a more useful metric is to understand the total number of touchpoints (frequency and clicks) alongside a view of how a channel supported an action that may have occurred via another channel, or drove the final action.
• It requires a single attribution model to be chosen. This in itself is not necessarily a limitation or concern, however the risk comes when multiple channels each report ROAS and each metric is used as a comparative metric to assess channel performance. Channels or channel owners can often use ROAS as a method of proving that all efforts deployed on a channel are effective and therefore, can choose a generous attribution model to attribute revenue. For example, Instagram advertising may report on any ads that drove revenue up to 7 days post a click of a sponsored ad or post, and up to 1 day post viewing an ad or sponsored post. If this is compared to the ROAS that might be calculated using Google Analytics, the attribution model may only account for last click and no view-through resulting in the interpretation being overly biasing one channel and the risk of double counting revenue.

 

ROAS is most commonly used as a defensive metric by channel owners to prove that a baseline threshold to continue efforts has been reached as on the whole, a channel can be seen to be contributing more than the total investment needed to activate on that channel. 

 

Used as a baseline threshold to monitor, ROAS can be a useful guiding metric however offers little insight into the true effectiveness of channel strategies and the campaigns deployed within them. 

 

 

 

To calculate Return on Advertising Spend, take the total revenue that can be attributed to advertising based on the attribution model of choice, divided by the cost of advertising. 

---

 

The Lates