Marketing Measurement – a quick guide to getting it right!

We’ve all heard the well-worn clichés about the difficulty of measuring marketing effectiveness. As tired as they are, there are valid reasons for their persistence – determining what works, why, and who for can be complex. But does it have to be? 

We still find that most organisations are relying on a ‘last click’ attribution approach to Digital & CRM measurement, and with good reason – most ‘off the shelf’ software platforms default to it, it is relatively easy to establish, and it can provide results, where previously there weren’t any! That said, it’s widely accepted that this lacks sophistication and accuracy, resulting in ‘best guess’ outputs that will over or understate the true impact of channel contribution. 

 So, how can you more accurately track the performance of your marketing, leaving you with actionable insights that can drive improved ROI?  

Our tips for enhancing your marketing measurement 

Moving to multi-touch attribution, considering all digital and direct channels, is typically the first step we recommend and is an area where we have had significant impact with clients. Having evaluated multiple approaches, we currently favour Markov Chain modelling, which calculates probabilities between successive channel interactions. This attributes impact to each channel, whilst also accounting for the sequencing of contacts between different channels. 

Another approach we often recommend is to use Shapley Values to work out the ‘co-operation’ that can take place between channels. This can help us to calculate transition probabilities (to determine the paths taken between channels) and the impact of channels in combinations.  

Once this level of capability and insight is established, we recommend that Channel Incrementality is established to further evidence the isolated contribution of each channel. This is achieved through a controlled testing approach, with channels stopped completely for short periods. We can advise on lean approaches here to minimise the time that channels are out of action. 

In addition to the Attribution approach, we also find great value in also considering the impact of ATL channels and wider econometric factors. We have seen influences such as competitor activity, economic factors (such as cost of living & employment data), weather and global events (Covid & Brexit) all have a significant and measurable impact on sales. The key is to identify data that has significant variability (e.g., things that change regularly over time as anything too static can be all but impossible to infer impact from). Our recommended approach is to combine the Media Mix Modelling with Multi Touch Attribution to provide a true 360-degree view of performance. 

This MMM approach is used to infer not only the size of the effect of each channel on sales, but also, where data permits, to determine the time decay of each channel – this defines the delay between any marketing channel activity, and its knock-on effect on sales.  This may often differ significantly between channels. For these we typically apply bespoke Bayesian models which incorporate business knowledge to inform probabilities for any unknown parameters. 

An additional approach we have seen add significant value is to consider ‘Segmented Attribution’ – moving from a channel or ‘overall marketing’ level focus, to a view of performance at the consumer segment level. This allows us to determine not only how effective a channel is, but also who it is effective for. This enables organisations to target the channel activity at the segments that will return the optimum engagement. 

Finally, once all of this capability is established, we recommend that you not only use this to explain what has happened in the past, but also to create performance forecasts for future activity. Ongoing tracking and optimisation can keep this accurate as the business evolves, so that attribution becomes a key asset informing ongoing marketing decision-making and investment planning. 

It’s still a complex and evolving discipline, but we’re helping organisations of all shapes and sizes, across all sectors, to understand what is working (or isn’t) and why, and to make better decisions as a result. If you’d like an informal chat about how we can help you, get in touch & one of our data science experts will be in touch to discuss your specific needs.