Valuing Customer Actions – the Key to Paid Marketing Success

1. Understand how customer actions contribute to long term value.

This will be a context-specific process, as every business will have a particular set of customer actions that deliver long term value. A retailer, for instance, might attribute widely differing value to the sale of the same product depending on whether the purchaser is a new customer, a reactivated customer or an existing customer. Even existing customer purchases will usually have a long-term positive effect beyond the immediate contribution margin, as each additional purchase fosters habituation. Marketeers should also consider actions apart from an immediate purchase, such as an app install or signup for marketing emails or a subscription program.

2. Recognise that paid marketing campaigns often generate a combination of actions.

These could be “x” purchases from new customers; “y” purchases from lapsed customers; “z” secondary purchases from existing customers. To calculate the total value associated with a session requires the number of each of these customer actions to be tracked and then multiplied by the value of each action.

3. Calculate the value of each type of action.

Determining the value of these actions can be approached in various ways and with different degrees of sophistication and precision. A basic approach is to compare a 12-month lifetime value (LTV) of a cohort that includes the action in question and then apply a rule-of-thumb discount, say 50%, to account for self-selection bias. A more sophisticated approach might use a machine learning model to identify “identical twins” based on a wide variety of factors, excluding the action in question, and measure the difference in LTV over 12 months.

4. Apply attribution modelling to determine a channel’s impact.

Multi-touch attribution (MTA) models are generally considered the preferred approach for allocating value across various contributing channels. However, not only do such models involve a high degree of complexity and cost to implement properly, but they are also heavily dependent on the ability to track clicks and they therefore come with inherent limitations in today’s digital marketing ecosystem. Initially it may make sense to stick with basic ‘last touch’ attribution, eventually moving to multi-touch attribution supplemented by incrementality testing and MMM (Marketing or Media Mix Modelling) to help overcome MTA’s inherent limitations.

5. Establish the relevant target for the specific channel to maximize the total value generated.

To properly identify the value-maximising target return on advertising spending (TROAS) setting will require an ongoing series of elasticity tests. This is because the change in value and cost associated with different TROAS settings is impossible to predict and requires ongoing testing to figure out whether an upwards or downwards shift may increase the overall net value created (Total Value – Total Spend). To make matters more complex, this Value:Spend curve will shift over time based on a wide variety of factors including seasonality and competitive pressure, which means it’s important to conduct new tests throughout the year, and also to look back at historical results to understand when and what changes are most likely to occur ahead of time.