In last week’s blog we tackled the concept The Diversity Paradigm, what it is and why it matters. This week’s blog continues on that theme and focuses on how you leverage diversity to create incremental revenue. To do that, we must balance the equation. We must diversify digital stimulus until its status rivals the diversity of customer behavior. When we do this successfully, which takes time and iteration, we see response rates increase first and then the optimization of revenue or margin second. We drive this diversity through the exploration of what features of digital stimulus matter. This is a process unto itself know as feature engineering. So, if we begin with 5 different kinds of digital stimulus and we believe that there are 150 kinds of customer behaviors then our path begins with a question of what features can drive variation from 5 to say 15. We proceed in phases of iteration so that we can learn with each iteration via machine learning what features are meaningful. For example, let’s say we hypothesize that time variants will drive response. So, instead of sending digital stimulus on a Friday AM to all of our customers we create a strata of time or daypart variants; AM, PM, Weekdays, Weekends, etc.
This simple start could yield an exponential increase in diversity of digital stimulus but fundamentally these variants are limited to time or daypart differences which may end up being irrelevant to customer behavior or customer interest.
Thus, we must have some robust capabilities that can execute and measure these variants which are relatively absent of human labor via intelligent machine automation.
Much of the original hypothesis in the early iterations on what features matter should be driven by human domain expertise that is likely informed by empathy for customer behavior in the context of the digital stimulus’ call to action.
For example, if the digital stimulus is asking the customer to take action within a limited time period then perhaps the time variant features are indeed interesting. In contrast, if the call to action is educational and not necessarily tied to immediate conversion then perhaps the message and its tone is a far better feature to begin with.
These phases of iteration are best leveraged in “real-time micro trigger” campaigns because with intelligent machine automation we can close the loop on insight latency in days vs. months. Batch campaigns that go out say weekly, in contrast, would be the worst place to test diversity because even with advanced capabilities it would take months to close the loop on insight latency.
So ultimately if you want to create incremental sales, you need diversity, and the key reminder is that diversity for the sake of diversity is better than not having it. Look for where you can move the needle on customer metrics at the customer level.