Lift Curves, A Statistical Standard

The purpose of the lift curve is clear: it lets the marketer visualize the increase in potential revenue as compared to a random selection of the database.

Data Operations Director

Time to read3 min read
October 3, 2016

In a context where marketers are struggling to achieve and maintain high deliverability, and where customer loyalty is increasingly fragile, the intelligent management of marketing pressure, whilst at the same optimizing ROI, is becoming a real make-or-break issue.

What is the best way of managing this pressure?

The best strategy to adopt in terms of pressure is to combine “full base” campaigns (sent to the entire customer base) with “targeted” campaigns (each sent to 5%, 10%, 20% or 30% of your base). Since targeted campaigns tend to be very well received, they enable you to grow your CRM channel revenue without detriment to your base.

This is precisely what Tinyclues is all about. The machine-learning algorithms developed by Tinyclues makes it possible to find the customers with the highest propensity to buy each product being promoted – therefore targeting campaigns correctly – enabling marketers to evenly distribute marketing pressure. In concrete terms, this distribution of pressure is effected via a tool that is little known amongst marketers: the lift curve. Let us explain!

Understanding lift curves

In the image below, the lift curve is represented by the blue line.

Lift curve

  • The horizontal axis shows the volume of the customer base that will be contacted during a particular marketing campaign (between 0% and 100%).
  • The vertical axis shows the potential revenue from the marketing campaign, expressed as a percentage (between 0% and 100%).
  • The diagonal line represents the results of this campaign if targets are selected at random. For example, if the base is exploited to its maximum (100%), this would result in a return on investment of 100%. Likewise, if you contact a random 20% of the population, you will, in theory, achieve 20% of the sales potential of that campaign.
  • The lift curve represents the potential additional revenue generated relative to the above-mentioned, as a result of Tinyclues’ algorithmic analysis. Thus, through having determined the best X% of the population for your product, you actually achieve Y% of potential. In the example above, you can achieve 78% of potential revenue by targeting 8% of your database. In a sense this is an extension of the Pareto principle, which states that 80% of your turnover comes from 20% of your customers. In this case, Tinyclues helps you to find those 20%.

Making use of the lift curve

As you can see, the lift curve remains constantly above the diagonal line. The purpose of the lift curve is clear: it lets the marketer visualize the increase in potential revenue as compared to a random selection of the customer database.

For optimum usage, it is better not to go beyond the tipping point, shown here as the red dot. In fact, the tipping point marks the limit above which an increase in volume will adversely affect your campaign stats (opening rate, click rate, response rate). It is therefore recommended to stay closer to the blue dot. The concept of targeting is generally most meaningful for volumes between approximately 5% up to 25-30% of the total database.

Lift curves for several simultaneous campaigns

This raises another question for marketers who want to run multiple targeted campaigns and use this chart for each of them: how can we make sure we do not communicate with the same customers several times, consequently falling into the trap of exerting excessive marketing pressure? If a marketer releases five targeted campaigns on a daily basis, will he accept that some customers may receive two, three, four or even five campaigns on the same day? That is where another of Tinyclues’ functionalities comes into play: it is possible to specify a maximum pressure per customer and for a specific period of time.

No matter what point is selected on the lift curve, there is always one constant: the turnover generated by a targeted campaign will always remain lower than for a campaign sent to 100% of a subscriber base. For that reason, the real interest of using lift curves hinges on multiplying the number of targeted campaigns. This is the ultimate answer to increasing the revenue generated by your CRM channel whilst at the same time strengthening customer engagement.

You might also like