In all industries (known to us) the acquisition cost of a new customer is much higher compared to expanding by cross-selling to a current customer base. If you know any industry, where this is not the case – we’d love to hear which!
Anyway – as a result of the lower cost of acquisition, Falck has been playing around with cross and upselling and it has been a solid source of revenue growth for many years. So much so, that they run their own Telemarketing(TM) department.
Now don’t go call every single one of your customers expecting them to buy new products every day of the week – that would just be pure madness.
Falck thought of this, and applied logical business rules to the approach, setting reasonable limits like the number of times a person can get contacted within a year.
Falck loves their customers and didn’t want to risk aggravating their customers into leaving, and with the TM response rates declining, they needed a new approach to replace the “one fits all” rules that had been in place for so many years.
Adding more rules was not gonna cut it.
It would have resulted in:
- being a nightmare to manage,
- being impossible to find the right pattern across so much data and
- add even more bias to business decisions.
After all, we’re just humans (unless you’re a machine reading this) – when it comes to data we simply have some limitations that machines don’t. Imagine calculating 128×15×2,3877 on paper, compared to just typing it into the calculator. Of course, it can be done, but the calculator will win on speed every time.
Needless to say, we all need a little help finding the golden nuggets in the giant mines of data.
So how about adding a large dose of machine learning with an Allyy? Using the Sales Audience product the goal was now to:
- Identify customers with the highest chance of a further conversion.
- End outreach to customers, that wouldn’t convert.
With the end of rules built with human bias and the rise of machine learning(ML) through the Allyy.io platform, Falck could see strong results manifest.
With ML models based on a cross-sell campaign that had been running regularly since the previous year, they were able to identify the top 25% of customers to call.
The sales rate in the 25% group is 3x higher than the former campaign results.
As the models are living performance beasts that recalculate often, we were able to see that 10% of the customers dramatically change their probability to buy month to month.
Once integrated the scoring works fully automated and did not require any training from the TM-Team.