It makes a lot of sense: A new visitor signs up to your website or mobile app, and you’d like to pace them through their “welcome” period. Maybe you’d like to tell the story of your company vision over a series of emails. Or you’d like to give them an introductory discount on the day they signed up, and then a gentle reminder a few days later. Maybe even escalate the discount if they don’t respond. And add a few other series glued to the end to accommodate for any post-purchase or churn activity.
This is what’s generally accepted as customer “journeys” or lifecycles, and solutions in the market provide many similar-looking flowchart tools with a dizzying array of settings. Here is the problem: your customers rarely follow such static and linear paths. They react in unexpected ways and can move in any number of directions. To try and capture every user in a rule-based series is impossible, and will often put the wrong message in front of a large percentage of users.
Fig 1: A market lifecycle flowchart tool. How many users potentially get the wrong message at point A? Do they drop off completely from that one error? Or how many people got a message too quickly at point B and churn? By point C, are we confident the state-of-mind of these users matches the message we’re putting in front of them?
A good example of the nuanced states of different customers comes from this excellent article from McKinsey and Co. It starts with a new customer evaluating a set of brands, yours included, at the beginning of their journey. As they evaluate and gather more info, brands start to get eliminated. Then the purchase decision moment comes, and then is where your brand and product need to win. We’d like to take this one step further, and posit that a customer could already be past some initial set of brand eliminations when you first see them, or are possibly already at their moment of purchase. No rule-based system will catch that, and a static Welcome series woefully under-serves these customers.
And how long does it take to manually segment all your users and map out these flowcharts? Can you ensure coverage for all lifecycle stages and customers if content creation for one small part of a lifecycle is taking weeks of time? Would a push notification have worked better in the flow somewhere? Should a cohort wait 3 days or 7 before their next message? And for all these rules and mappings, are errors being introduced? If someone else takes over a heavily complex flowchart, will they break it if they try to modify any steps?
These rule-based series are simply inefficient and unscalable. The best you can do is lay the tracks for a handful of “common” journeys, and hope your customers follow. But this does not adapt to changing customer behaviors, and cannot capture all possible scenarios. Even the best flowchart tools cannot personalize to the precision that is required to manage a business at scale. A restrictive flowchart pigeonholing customers into static paths simply does not reflect reality, and can do more harm than good.
So what is the trade-off for all this complexity and pigeon-holing? We took a close look at series data from several eCommerce stores, and here’s what we found:
For most campaigns, “stage 1” is almost always the highest performing, as it should be. That first touch will conceivably be the point of highest engagement for most users, whether it’s the first message they receive from your business, or the first note about winning them back. But all the subsequent stages tend to be flat or worse, sometimes much worse.
This Winback Series 1 is fairly typical of what you can expect in terms of performance: a declining trend after stage 1 as users see the same contextually similar messages one after the other. Could Artificial Intelligence have put something else there instead? Well, too late, as these users have been set on an inflexible path they have no choice but to follow.
Nice spike in performance in stage 3 and 4 for Winback Series 2 — was it because of the order or the timing, or something about the messaging? Nope: it was because these two stages offered a big discount, while the others did not. So a specific series of emails doesn’t seem effective here at all either, with performance boost only because a discount was offered. Maybe just simply a more relevant message could’ve gotten that lift and saved you a discount.
So what is the alternative? The only way to instantly react to every user in lockstep with their behavioral changes, and to predict what content is most relevant to their current state of mind at scale is with Artificial Intelligence. Big advances in this field of study have been paving the way for technology companies to bring some futuristic innovations to market.
With the right technology, you can predict how much each and every person will spend in the future, what their actual probability to purchase is, and how long they will be a customer before they churn. No more guessing. How sensitive is each person to price? What hidden affinities do people have with certain products or categories? What extrapolation can be made from certain demographical patterns and behaviors? Based on different types of activity, how frequently should you reach each person with a message (or how infrequently)? When should you insert a mobile notification? Will they have a higher probability of responding at a certain time of day?
You would need an army of data scientists managing these models and the massive amounts of data to make this work. They would need to constantly tune these models and research what advances are happening in the market, and decide what can or can’t be applied effectively. Their models must be programmed to constantly learn and adapt, and then learn some more. The data scientists can go to sleep (sometimes), but their models have to be always on, always optimizing.
Exhibit A: Retention Science data scientists, not sleeping.
And only with Artificial Intelligence can predictive models be made actionable and automated. The Artificial Intelligence’s job is to predict the state-of-mind of each user and assemble, in real time, the “right message at the right time to the right customer.” A single customer may have a 6% chance of converting with a 30% discount in the sports category two days from now at 10AM.
But what if you blasted him with a generic promotional email instead on that day, which only had a 1% probability to convert? How often is that happening, and to how many people? What if that day was also his birthday, and thatmessage would’ve converted at 9%? Artificial Intelligence can triage all these messages and predict which one would perform best, personalized to every customer, and do the send for you. Can your flowchart tools do that?
Fig 2: A real-life example of a customer base flowing through different lifecycle stages, managed by Artificial Intelligence.
Most importantly, Artificial Intelligence-powered marketing provides marketers with the much-needed bandwidth and opportunity to focus on honing overall brand strategy, perfecting content, and developing new ideas on how to reach your customers.
Sounds great in theory, but what about in practice? Luckily, we happen to be pretty knowledgeable on the subject: Our beta partners on our own Artificial Intelligence technology have seen on average a 45% increase in conversion rate, a 66% increase in open rate, and a 57% increase in click rate. And we know this performance will only improve over time, as the machines keep learning and optimizing.
The future of marketing technology has landed, and it’s a game-changer in every way. Artificial Intelligence eliminates manual and error-prone pre-planning, and replaces it with dynamic, automatic adaptation. Learn more about Cortex here.
About the Author
Derek Kwan is a builder of products for Retention Science, crafting data science into strategic tools for marketers. With 15+ years of experience in the marketing and ad tech industries, he previously led product innovation at Yahoo and YP. Derek also trains his game theory skills in his spare time as a poker player.