How long is customer data really useful for? Legal considerations aside, this question is key to how your communications campaigns perform. Some marketers use databases from the 1950s – and it works! Others think that after six months of silence, you’ve lost the customer. It’s a tricky business. So just how do you make the most of your data over time? And how do you integrate innovations around marketing targeting? Here’s how to set the record straight.
What is your data’s “best-before date”?
For most e-marketers, the rule of thumb is that if there’s been no on-site activity for three to six months, then a contact is inactive – or “profund inactive” if there haven’t been any purchases for 12 months. Although data recency is still important, the rules for sorting active from inactive customers are not always the same. At a time of retargeting and highly targeted segmentation, data may be obsolete after just three days. On the other hand, some information is still useful after three years. So, just how does predictive marketing fit in?
Innovative targeting solutions, like Tinyclues, look beyond using recency as your sole criterion. As a rule, they look at all of your customers as potential targets a priori. It doesn’t matter whether they’re categorized as “active” or “inactive”; they are included in the algorithms’ analyses and may receive a relevant offer, even if they’ve been inactive for several months on your website.
In any case, the more history there is available, the more the algorithms can flex their muscles and work to their full potential. To do this, it’s best to adapt how long you use the data for depending on each industry. Data relevance over time depends on customers’ natural buying cycles, whether that’s three days or three years.
Short buying cycles: this information will self-destruct in 10 seconds…
When business is defined by short buying cycles and regularly updated offers, it’s very tricky for marketers to use such data – all the more so if they don’t use a predictive marketing solution. In those instances, Tinyclues’ solution is here to help. Tinyclues does much more than simple history-based aggregates (number of purchases, clicks, etc.) and is completely self-learning. The AI algorithms can learn to be entirely independent:
- taking into account your customers’ buying volatility
- including how your product offers develop in any way, in all its variety;
- automatically recalculating the most representative variables for buying behaviour on your site each day, without any preconditions or hard-and-fast rules.
To do this, Tinyclues’ technology requires in-depth data. For cultural, cosmetic or textile products, it’s harder to make predictions due to the offer volatility and the diversity of the purchases of each customer. That’s why algorithms need to step back in time (for example 24 months) to characterize your customers in all their offering and behaviour diversity. In the end, Tinyclues enables you to target your customers who are most likely to buy based on implicit signals. This complements the setting up of simple rules which are enough to push food products or consumables (such as ink cartridges) where repurchasing is easy to determine using just two or three occurrences.
Long buying cycles: the time machine
For a vacation, electrical appliance and other less common purchases, buying cycles are much longer. Hence some e-marketers have to travel further back in time to analyse customer behaviour. Current practices often limit data comparisons to just one year to compare with the previous year. It’s time to shake this up: fresh data is still a must-have, but the further you can step back in time, the better you will subtly qualify your customers from the implicit signals linked to their overall behaviour.
If you want to show your customers new and relevant offers, or find another audience for your existing products, your data has to cover a minimum of two buying cycles. And it’s the combination of that historic and most recent data that enables Tinyclues’ algorithms to work their magic. First, they analyse your customers’ behaviour over a lengthy period to build the independent variables. Then, they look at your recent data (last 30 days) to identify recent lookalike buyers using these explanatory variables.
Why should you adopt this method? Because of these two main advantages:
- You broaden the target for a product beyond those who bought it the year before. It’s the end of the Y-1 tunnel effect. For example, looking at long-term history and recent data, Tinyclues could significantly boost the number of bookings made by travel industry customers.
- You differentiate your targets according to buying times or the seasonality of offers. For a trip to the United States, Tinyclues gives the possibility to naturally target early bookers (those who book their holidays six months in advance) and last-minute bookers (those who book the week before they go) depending on the moment when you do the targeting.
The 12-month (or three-month or 24-month) rule has had its day. With AI algorithms and Tinyclues’ deep learning, you can easily factor in buying cycles, seasonality and the implicit signals in your customers’ behaviour. Increase the value of your data for as long as you need to: try Tinyclues’ methods now!