Our Insights.

Covering strategic and technical aspects of analytics.

By subscribing, you agree to receive a monthly newsletter from 173Tech.

Analytics, The Catalyst For Maximising Digital Marketing ROI.

June 1, 2020 | By Robin Watteaux.

When people speak about “analytics” and “marketing” together, most likely they are talking about attribution and complex modelling to increase ROI. However there are many other analytics techniques which can increase marketing ROI much more significantly than complex attribution modelling — and usually with less effort.

This article presents a simple yet extremely efficient analytics framework which aims at maximising digital marketing’s long term value. It is composed of five stages to shift your optimisation from CPA towards ROAS / ROI and start using algorithms for optimisation recommendations. Please note that, although this article was written around Direct Response (DR), similar principles can be applied to Brand activity. The chart below displays the framework’s five stages:

Analytics Marketing ROI

Stage 1: CPA Optimisation

Overview:

Action based optimisation (e.g. free trials, purchases, signups, etc.) which is quick and easy to implement but lacks visibility on conversions’ long term value.

Detailed:

The most common marketing conversions rely on sending actions back to individual platforms (e.g. Google, Facebook, etc.). For web, this is done via floodlights (i.e. cookies) whilst mobile tracking uses the implementation of SDKs. This enables the marketing to optimise towards CPA (Cost Per Action). This is action based optimisation (e.g. a user signs up, a customer makes a purchase, etc.). Although such optimisation gets you started quickly, it fails to capture long-term value.

For example, let’s consider you have two campaigns A and B optimising towards signups, with CPA at $8 and $10 respectively. Campaign A appears to be a winner but what if B brings in higher quality users that spend double the amount? CPA optimisation fails to capture customers’ lifetime value (LTV) and therefore leads to suboptimal ROI.

To optimise towards LTV, you need to shift towards behavioural based metrics (e.g. LTV, retention, etc.). This requires a more sophisticated analytics setup as these events are not triggered by simple actions and cannot easily be captured with standard tracking — a user does not click on a button that says his LTV is $100.

Stage 2: ROI Visibility

Overview:

Connecting digital marketing spend and customer data in order to 1) build visibility on long term value and 2) deduplicate conversions.

Detailed:

For ROI / ROAS visibility, you should centralise marketing and customer data in an analytics warehouse where it can be jointly processed (you can find more detail about that topic in this article I wrote). In addition, ensure that you pass acquisition source details to your customer data. In other words, if a user registers or makes a purchase, your backend system needs to capture where this action came from (i.e. media source, campaign, creative, etc.).

Now you are able to join marketing data with your customer journey and behavioural data. You can optimise towards LTV, retention, lower funnel activations, etc. This can be broken down to whichever level you desire (e.g. channel, campaign, adset, creative, geo, etc.) as long as you ensure the relevant information is passed to your system. In addition, the conversions will be deduplicated across platforms during this process.

Web advertisers that capture acquisition sources in their customer data will also be better prepared for the cookieless world. Google has recently announced that they will stop supporting 3rd party tracking (which powers a lot of current web tracking methodologies) in the next two years. This will further widen the measurement gaps created by walled gardens and web advertisers will seriously need to rethink digital tracking in order to keep some performance visibility across channels. In a world where there is no perfect solution and cross-channel visibility is constantly undermined by ever increasing red lines, consolidating attribution on the advertiser’s side is an attractive path.

Stage 3: ROI Optimisation

Overview:

Use data science models to predict target metrics (e.g. LTV) for early optimisation signals.

Detailed:

Stage 2 allowed greater analytics visibility for your marketing activity. However, revenue takes time to build, e.g. a user could have a subscription for years. Your marketing team optimises on-the-fly and cannot wait for that long.

To provide quick feedback on long term ROI, build a data science model to predict your target metrics , e.g. LTV or retention. For example, we have built LTV prediction models based on users’ first 24 hours of activity, allowing the marketing team to optimise towards long term ROI within one day from conversion!

You can also pass predictions as events back to your marketing platforms (i.e. Google, Facebook, etc.) to allow their algorithms to optimise towards long term value.

Stage 4: Automated Optimisation Recommendations

Overview:

Using algorithms to generate optimisation recommendations at any desired frequency to guide marketing budget allocation.

Detailed:

Stage 1, 2 and 3 enable you to shift your optimisation from short term value to long term returns. Next is to automate the optimisation process. The platforms (i.e. Google, Facebook, etc.) have sophisticated algorithms to optimise campaigns, but you still need to optimise across campaigns and channels. At scale, analysing data for the latter can be quite time consuming. This is when algorithms come in extremely handy! They support your team by providing automated cross-channel / campaign budget shift recommendations, and much more. This is especially powerful if you have complex optimisation targets.

For example, if you are a dating app and optimise towards ROI, you might end up acquiring a lot of male users, as male users are more likely to pay. Therefore, you need to balance your ROI target with a health metric such as % female acquired by modelling any deficiencies into monetary values so campaigns can be compared like-for-like. In addition your algorithm can optimise whilst ensuring the overall % female of your marketing activity is over a preset floor.

Algorithms can crunch the data in any way as defined by your business. You can then visualise the algorithm’s recommendations in your reporting tool such as Looker or Tableau. Your digital team can then come in on Mondays already knowing what optimisation to implement thanks to the algorithms!

Stage 5: Automated Buying

Overview:

Automate the buying activity by feeding the algorithms’ recommendations back to the digital marketing platforms (i.e. Google, Facebook, etc.).

Detailed:

Once you have designed your recommendation algorithms, you will have pretty much created an entity that could buy on its own. If you wish to completely automate your digital marketing buying effort, you can start passing back those recommendations to the platforms (i.e. Google, Facebook, etc.) and let the algorithms run the show. For example, your algorithms could automatically upweight or downweight specific campaigns, change bids etc. I will not delve into further details in this stage as this becomes mostly a technical topic.


This is it! All the above stages should help in your journey towards higher long-term marketing returns. Our team at 173Tech has worked on many projects implementing the above and we typically observe an increase of ROI by over 50% in the first month following release. Do not hesitate to reach out if you want to know more!


Subscribe to our monthly newsletter for more insights.

By subscribing, you agree to receive a monthly newsletter from 173Tech.

Get In Touch

Send us a quick message about your data challenges, we are always happy to chat!