Today's blog post was written by Bryson Engelen, Sales Engineer at Sonoma Partners.
Maintaining and upselling current customers is important to just about any business, and knowing where you stand with a customer right now let’s you develop an ongoing relationship strategy. There are a lot of ways to take a customer’s temperature, and you may already be using them and find them time-consuming, highly subjective, and perhaps inaccurate.
But what if I told you a robot could analyze your relationship with your customer for you with information you already have?
Right now, you probably rely on two basic things to know where you stand with a customer: surveys and your gut. Surveys have come a long way in the last few years, with long, complicated forms that need to be compiled manually morphing into simple five star ratings that can be aggregated and analyzed automatically. And as great, simple, and automated as surveys are getting, there is always that nagging suspicion that the customer is just giving all high marks to get it done, that they are being more generous or critical than they really feel, that you caught them on a bad day, and many other concerns. As for your gut, you know when you email, call, or meet with a client by their tone and expression how they feel about you. Still, you might not know whether these feelings reflect their personal feelings towards you and/or whether that translates to the rest of your company. You only really know your little world. In the end, both of the methods people tend to use right now for checking the health of a relationship are pretty subjective, manual, and incomplete.
Some companies have begun doing more sophisticated relationship tracking and scoring by measuring volume and quality of interactions between their firm and their customers. This usually comes in the form of looking at how many emails and appointments (and maybe phone calls) are happening between both companies and in some cases, there is even a scoring mechanism put in place for the depth of that interaction (emails are a low score, meetings are a high score, 1-1 interactions are high scoring while group emails are low scoring). Using this interaction data, you can see how engaged a customer is with your firm and individual employees, and note trends of upward or downward relationship health. You can even automate notifications if a particular relationship between individuals, or relationships firmwide, are heading south.
While that kind of interaction-based relationship analysis is still pretty rare, it is available today and there are products on the market and components of modern CRM systems that have such engagement tracking prebuilt. The limitation is that these will only track the volume and to a certain extent the quality of those interactions, but don’t really tell you the tone of those interactions. Maybe you have the most emails this year with a firm that really only sends you complaints and doesn’t really like you anymore, but the interaction scores are high. So we know the customer is speaking loudly, but is it a cheer or a primal scream of rage?
Ideally, you should be able to know the health of your relationship with your client based on how they tell you they feel, how your gut says they feel, how much you’re talking to them, and an automated, more objective analysis of every single interaction you have with them. Everything we’ve talked about so far pretty much covers most of those needs, but it’s that last bit that is pretty elusive. And this is where sentiment analysis comes into play.
A lot of players in the cloud world are beginning to roll their investments in AI and machine learning into sentiment analysis for customer interactions and pretty much everything else, meaning a robot can now read every email from every person who works for your customers and determine whether that customer likes you or will buy from you. If you have five employees who interact with five contacts at a customer over several months, sentiment analysis can look at every interaction and score whether it was good, bad, or neutral. Those individual scores can be added up to a total score and broken down by employee, month, or sliced and diced however you want. If the customer uses words like “great” or “helpful,” the interaction gets a positive score and if the customer uses words like “disappointing” or “slow,” the interaction gets a negative score. This means the machine is doing what you naturally do when you read an email, but then making that feeling reportable and aggregating that feeling from all the employees of that customer and from every member of your team. Since the interactions are in the customer’s own words, they are likely a little more genuine and a little less coached than a survey, and they aren’t happening at one specific point in time, but over time. They aren’t happening at particular milestones, they are happening all the time. This is critical.
With a little more work, analysis can be done on particular phrases that can signal interest in particular products or services and do things like let you know if a particular email is more important than others, if a customer is asking a question, if they asked for a quote, or any other specific alert that helps you filter out the noise of your inbox. And some of these more sophisticated analyses are beginning to show up in CRM systems, where emails with the characteristics mentioned above create notifications to you and help you prioritize specific customers over others.
While sentiment analysis in the CRM space is pretty immature, activity scoring has been around for a few years, is gaining in sophistication, and will continue to be a critical component for automated relationship scoring. And when we put all these things together we realize that the same robot working on sentiment analysis can do things like look to see if the amount of traffic between your company and the customer has increased or decreased, if you’ve moved from having a lot of face-to-face meetings to just emails, and if you’re interacting with more or fewer people at the customer than before. Then by adding in things like survey results, you get a really strong picture of not just the overall health of your relationship with a customer, but where the relationship is strong and where it’s weak (a particular person, a particular subject, etc.).
This may even mean that in time surveys themselves are dead, because if every interaction you have with a customer is scored in real-time, you’re essentially taking their pulse all the time and can know via machine learning whether they like you or not. And by analyzing sentiment over time and across different customers, different interactions, different sales or service cycles and more, you will begin to know when is the right time to talk about specific topics and begin to have predictive models on what next steps should be.
The whole idea of predictive selling, of a robot telling you what to talk about next with a particular customer, isn’t new; it’s been a point of discussion in the CRM space for years. But what is new is the ability to leverage your customer’s feeling about you in real-time to determine when those predicted actions should happen. And predictive selling products are now getting integrated more and more into CRM and email traffic to enable their accuracy to skyrocket.
If you gauge the health of your relationship with customers using a relationships score, health score, or some mechanism around whether the customer will remain a customer and buy more, you will soon be able to do this far more accurately and without the manual work. Soon you can rely on a robot to tell you which of your customers like you and will buy more from you and which won’t. Because your customers are probably already telling you what you need them to tell you, but you don’t have the tools to pick up on it. And those tools are getting better and better all the time.
To get see a more concrete example of this sentiment analysis at work as it relates to scoring emails, check out this post.