Predictive Analytics is an amazing way to surprise your customers and generate more sales. It predicts each customer’s next order date automatically using an “expected date of next order feature.”
Use it as a trigger for the “Repeat order nurture series”. This is a great version of a win-back flow, but one that can be even more effective. Imagine that a store where you’re a regular sent you a message just when you were thinking about buying again. Since much of marketing is about timing, the store would have a real advantage getting you to make a purchase….
Predictive analytics is a tool provided by several ESPs. It estimates results based on a mathematical model which utilizes Bayesian inference. That is, the model improves and learns over time as new data becomes available. It matches data from a single contact with data from the whole list.
Let’s say Customer A buys from you and buys again after 10 days. The model will use this information for predicting Customer A’s next order date. Not only that, it will look into data from your whole list and search for patterns that match with Customer A’s behavior.
Of course, the tool is not perfectly accurate, just like any other statistical model that aims to predict human behavior. Nevertheless, you can’t know your customers accurately either. They’re never exactly who you think they are. There are thousands of them, and they buy for an unthinkable variety of reasons and factors.
That’s where predictive analytics enters the scene: the more repeat buyers you have, the more data you’ll collect. Consequently, the predictive model will be more accurate and make you more revenue per recipient. When your ESP provides you with such powerful mathematics in a simple user experience, there’s no reason why you shouldn’t be testing it.
So how do we adjust that Winback Flow so that it’s based on more than just whether someone is engaged, unengaged, a buyer or a non-buyer? One way is to use the expected date of the next order as the trigger. You can also adjust the time frame, for example triggering the flow a week before the expected date of next order.
Another way to use predictive analysis for segmentation is to create a segment of people with the expected date of the next order within a certain period. It is possible to combine it with your current main segment in different ways.
For example, you can:
- Send the campaign to both segments, adding more people to your sending list.
- Reverse the conditions of the new segment to exclude it from your campaign.
In the first example, you will potentially send the campaign to more people and expand your main segment.
If you’re struggling with deliverability however, I recommend using the latter option. It will exclude people from your main segment that aren’t expected to buy anytime soon.
Predictive models are being developed at a fast pace in several areas of digital marketing, and email marketing is not an exception.
However, in order to be able to use this Magical 8 Ball, you need to have enough data (such as purchases, open and click rates, website activities, etc). To get your data, you need to launch the “Tag First Purchase Date” flow that can be found in Klaviyo flow’s library. We recommend that you do this as soon as you activate your Klaviyo account to start collecting the required data.
Yep, that will give you a FACT-driven Magic 8 Ball. Goodbye, confusing Excel formulas; hello to predictions backed by actual data science algorithms. Once a customer makes one purchase Klaviyo automatically calculates their predicted next purchase date.
Of course, test it thoroughly and build it out slowly, as it’s easy for predictors like this to be wrong, especially if they don’t have enough data. Email accounts that typically have a limited number of repeat purchasers often can’t accrue enough data for predictive models to be accurate. With that caveat in mind, if your list is large enough, it’s definitely worth a try.
To get started you need:
- To launch premade Klaviyo’s flow “Tag First Purchase Date – Enrich Contact Profiles”.
- Create a sequence of emails with nurturing content for repeated buyers.
- Analyze your results in terms of open rates and revenue.
We’d love to know what experiences you’ve had with predictive analytics and similar technologies. Comment below — we’ll compile the results and follow-up with what we learned!
Anna Hrychukh, Rodrigo Santiago Juacaba and Thomas McClintock