The Urgency behind Customer Anomalies

Anomalies are the new path to revenue - do you know yours??

In the past few years there has been a proliferation of roles created in companies about understanding the customer - Customer Experience, Customer Advocacy, Customer Success, and of course the continuous journey of all of your sales, marketing and product teams to deliver results based on this knowledge.?Yet, despite a need to gain more customers, more market share and differentiation, most of a company's spend and effort has been on well-known segments, behaviours and tried and tested approaches. It makes sense, there’s a reason for the 80/20 rule.? But it's time to get savvier about our approach.? Many of our customers budgets have declined during Covid and have not yet stabilized to pre-pandemic levels or optimism. Conversations need to pivot to targeting the best set of customers based on our own limited capacity. We need to identify and grow micro-segments to create new revenue streams. We need to know how to shift operations precisely to improve our margins.?

This is where most organizations are struggling.? Executives can look at their dashboards and reports see how their key business metrics are trending.? During discussions, they can provide high level, reasonable reasons for this behaviour or cite well known market conditions and issues. But when pressed, teams don’t know exactly what precise combinations of actions will improve their numbers and their data teams often don’t have the answers either.?Why? There are simply too many factors to consider.?Think about it this way.?Let’s say you have 10 different customer segments, a handful of geographies, different income groups (whether B2B or B2C), multiple products, and further combinations of shipping modes, risk profiles, payment terms, etc.?

By now, you and your team should be able to see how those individual factors are trending.?Maybe air shipping is up and sales in the Midwest are down.?But that’s still not telling us how to address these issues.?We might have some hypotheses, or our own well-honed intuition based on years of experience, but is it enough to implement a specific plan with associated funding and know it’s going to get you the best possible return??

The scenario above is common to every industry.?Every single day we’re being asked to make complex decisions based on understanding individual customer and business variables in isolation.?Getting the real insights to make critical decisions on spend, capacity, training and products won’t be found in this approach. In order to truly make great strategic decisions it means understanding exactly what is and is not working when the key drivers of your business are considered as a whole, and then, identifying the anomalies.?

An anomaly is where high impact potentially lies. While there isn’t a Webster’s dictionary definition of a “customer anomaly”, I’ll take a shot at it.?A customer anomaly is a statistically meaningful subsegment of customer attributes that is significantly over or underperforming against a key business metric.??(Note, you could replace the term “customer” here and use the term “business” if you’re in operations, supply chain, risk or other functions).?

Let's look at a real world scenario. In late 2020, the SVP of operations from a large company was asked to reduce costs by 3M$ to support corporate cost cutting measures. In her role, she ran a number of large call centres. ?Her primary cost drivers were headcount, training and technology.?For discussion purposes, we’ll say her cost per inbound phone call is $8/call.?So, to achieve her objective, she needed to reduce the number of customer calls by 400,000 odd calls.?That’s not an insignificant amount of calls - how was she going to do that??

Her dashboards confirmed the issue but didn’t tell her what her best next actions were. I'm sure you are in a similar situation constantly - there is a critical number you need to hit and many levers you can pull, but are you quite certain which are the best? ?

Our protagonist understood this wasn’t going to come from a single broad action.? She realized the only approach was to deeply understand what was driving customers to call with enough specificity that she could implement corrective actions within the business, while being assured they were the best actions based on available resources.?She knew the insights she needed were invariably to be found in the staggering amount of customer data that was collected on a daily basis. However, that data was spread out across systems and formats - customer tickets in ServiceNow, voice recordings, and the Salesforce CRM system.?Just getting this data together in a data warehouse to begin categorizing it would take months and that’s before the data team could start running manual queries to look for answers.

Let's dig into this problem for a minute. Have you ever considered how many potential insights are in your data? The table below is a bit shocking. The left side illustrates what it's like to build a dashboard, when you're looking at a single variable (i.e. region) with an X and Y axis. The right side talks about the amount of combinations that need to be tested if you want to have insights based on combinations of attributes in your data like region, income, ad campaign, gender, AND product.

Now, I won’t get into a long technical description of how, but the way to solve this type of problem is by what’s called “multi-variate analysis”. ?It means looking at all the factors in your business and customers as a whole.?And if you’re in sales, marketing or operations, you can’t just assume your data team can do this.? Because it’s absolutely, insanely hard.?Multi-variate analysis is essentially looking at all the attributes of your often siloed data, in every single combination possible, then surfacing those combinations that are significantly under or over performing (the anomalies) against your key metric (say, revenue) for discussion and prioritization.?For our Operations SVP, a key anomalous multi-variate insight was “you are spending 55K per week on calls coming in from customers in Spanish speaking households in the Southwest that have been with the company for 17 days or less”.? The anomaly here is that this "pattern" was growing at a rate 10x faster than any other combination in her data. Her "aha" moment was realizing that creating Spanish literature and having Spanish speaking installers in that region for the digital products getting installed in the homes of new customers would save the company 2M$ a year. Try finding that on a dashboard.

The power of the anomaly

The point I’m trying to make is it’s the anomalies that tell you exactly where to focus and dive in.?Like understanding exactly what role, in what industry, and what message is landing you the most new logos. It’s also the anomalies that will tell you that your managers who work in more than one retail outlet a week are 5 times more likely to quit after two years with the company.?Or that shipping costs are going through the roof due to a staffing shortage on Thursdays in your central warehouse.?It’s anomalies that tell you what to change to move the needle.?So, you could, I suppose rely on what you’re used to.?Or you could try to get your data team to test a billion plus hypotheses in your data themselves.? Or you could take a little bit of time to find out how Unsupervised AI solves the problem above and how its quickly becoming a must have for companies that are embracing innovation, data-driven decision making and crushing their numbers.?