Predicting customer behavior can be easy if you are analyzing your customers and leads. Using advanced segmentation, you can predict how your customer will respond in a number of scenarios including churn, offers, upsells and more. The challenge is that it requires data. Once you have the data, you can plug in the data to create mathematical models for AI-driven predictions. The key is to assign multiple segments to a customer representing different ‘variables' about the relationship – then these segments can be plugged into AI algorithms to predict behavior.
Beyond Demographics - Segmentation in the 21st Century
As marketers, we need to think about segmentation differently. Real life doesn't follow Marketing 101's 4-Ps – Pricing, Product, Placement, Promotion. People are unique, but you can identify patterns with their behavior.
Here are the types of information you can collect to create advanced segments:
Demographics/Firmographics
Traditionally, this is how most marketers think of segmentation and it's still valid with AI predictions. Most people don't know that a zip code is the highest indicator of socio-economic status - and this information is available for free to download from the government.
Basic demographics like age, income and gender are very relevant for identifying target audiences - however, it makes little sense to market a Maserati to someone who can't afford it. The same is true for B2B marketing - don't market enterprise software and services to a startup and you shouldn't market small business solutions to enterprise companies.
Persona
A traditional marketing persona is represented as a person - "Our customer is Karen who is 34 years old, had 2 kids and drinks wine on Sundays." A generic persona can be used as a tool to help sales understand organizational hierarchies but doesn't really reflect a specific person in the buying journey. People make decisions based on different factors:
- Needs/Urgency: What is the problem you are solving and what is the urgency and budget?
- Communication Styles: People communicate differently and having different types of information available based on the information they need to make a buying decision - did they buy because you are likable and they trust you or because they did a detailed analysis of you and 10 other competitors.
- Positivity/Negativity: People fall into a range of positive attitude vs. negative attitude and it's often very visible on their social media presence.
Decision Drivers
This is a rarely-used technique in order to segment customers. By analyzing decision drivers, you can learn a lot about what 'variables' led to a successful conversion. When someone found you, how did the find you? An ad while searching for something? A referral? An SEO search?
There are multiple variables that you can track to determine the decision drivers:
- Leadsource: This can tell you what methods they used to find you and specifically what keywords they were searching on.
- Call-to-Action: what call-to-action did they respond to? For an ad campaign with multiple permutations or A/B testing set up, determine the call-to-action that was relevant to get them to take action.
- Website Activity: many people do some level of research before contacting you, purchasing, or subscribing. If they were looking at various information on your website - what were they looking at to make a buying decision?
- Competitive Analysis/Questionnaires: If you are in a highly competitive industry and win a deal against one or more competitors, identify what your differentiator was to win the business.
Behavior
Behavior-based segments are completely underutilized and can be used to create deep insights into customer behavior. With every customer they have a level of experience that they had before their relationship with you. Maybe they are a power user, maybe they have been burned before. But, by asking a few questions during the purchase process and observing their behavior after conversion, you can significantly improve the customer experience.
There are multiple variables that you can track to determine the behavior information:
- Error Logs: It shocks me the number of companies that log errors never to look at the data again. Error logging can be used to proactively contact people having problems with your software and can be fully automated to respond to problems your customers have.
- Email Engagement: Most people are buried underneath an avalanche of email and it's only getting worse. Are they reading your emails? Are they unsubscribing? I recently went to a store, went to their website while I was in the store, and was immediately offered a coupon. There were two different issues with this:
They knew my location (it prompted me to allow location settings) so with about 20 lines of code, they would have known I was at their store.
I subscribed to get the coupon, then immediately unsubscribed once I got their first email.
This marketing campaign cost their company $15 of revenue that I otherwise would have spent in their store. By reviewing subscribes and opens, you can get some real insights into how well your email marketing campaigns are working. - Help Searches: Another quick win is in tracking your help searches. If someone is searching your site for information on how to use your product, you might want to make sure that the information that you provided helped. A make-or-break moment for many companies is how they operate when there is a problem. By identifying patterns of problems, you can either proactively serve your customers or you can prioritize your backlog according to the types of problems customers haver having.
- Level of Experience: Are they a novice on your platform or are they an expert? Have they used similar products/services before? By tailoring a small survey during the on-boarding process, you can identify a LOT of information to make your communications much more useful for your customer.
Creating Microsegments
The culmination of these data points are what I like to call micro-segments. Using the data that is pretty readily available, you can identify:
- How much they make
- Who they are
- What they need
- How are they engaging with your company
Types of Predictions
Predicting the future starts with understanding the past. By collecting information about your customers and their behaviors and by creating micro-segments, you can begin to use this information in order to predict behaviors of current and new customers.
Here are a few types of customer behavior predictions you can make:
- Retention/Churn
- Satisfaction
- Engagement
- Training Needs
Retention/Churn
Retention/Churn
The culmination of these data points are what I like to call micro-segments. Using the data that is pretty readily available, you can identify:
- How much they make
- Who they are
- What they need
- How are they engaging with your company
Retention/Churn is probably one of the easiest metrics to calculate. Retention predicts the probability to remain a customer, while churn predicts the probability of cancellation. They are opposite metrics – just some people like the positive spin of ‘retention' over the negative sentiment of ‘churn'. Demographics can be used to predict this, but by leveraging all four types of segment information to create AI algorithms, you can drill down to create accurate predictions of cancellations and can proactively address problems.
Some of this is just using the data and information at your disposal – for example if someone gets an error and you proactively contact them to provide steps to resolve it – there's no AI involved in that – however, this is the next level of retention.
If someone can't afford your product/service – you were marketing to the wrong person in the beginning of the process. If someone experiences a problem that can be solved, then focused effort needs to go into saving the sale.
By going beyond basic demographics, you can see clear patterns that can help you identify marketing problems and can help improve long-term retention over time.
Satisfaction
Satisfaction is sometimes measured within a company with NPS scores - "How likely are you to refer us to a friend", but the problem is that because MOST companies don't actually do anything with the data - people have become jaded in offering feedback.
Just monitoring social media channels and tracking sentiment over time can do a lot to improve overall satisfaction levels and can be used to drive new innovations and improvements in your product or service.
But - the key here is communicating and being transparent. There's nothing worse than being told "we've logged an issue" and it going into a black hole - never to be heard from again. I managed support for an enterprise platform with thousands of users. We created a process where EVERY ticket was linked to a problem. I had a tremendous advantage when a squeaky wheel would complain and I could point out that issue was #35 on the list of problems and our top three problems were prioritized because we received dozens of complaints. The side effect was that once our customers knew that we were monitoring the data, they started submitting requests when the problem happened and we were able to prioritize requests accordingly.
Imagine using a community Q&A forum where people can search, ask questions, and get responses, then monitoring sentiment using Natural Language Processing to categorize sentiment, problems, training and usability issues.
You can also leverage engagement information and behavior information to can predict satisfaction and work proactively to improve the customer experience.It's pretty easy to predict that if a customer experiences a hight number of error messages or does a lot of searching on your website for help, that they would have low satisfaction. The data exists to track this information and respond accordingly, we just aren't using it.
Engagement
Engagement is the next area of personalization and the ways that you engage with your customer can be predicted and optimized. If you are finding a customer isn't opening your spam - I meant to say 'drip email campaign', then identifying the most effective ways to engage with you and for you to engage with your customer or lead is pretty important.
You can predict the ways that similar people with similar attributes engaged with you and tailor your communication strategy accordingly.
Training Needs
Going back to the behavior-based segmentation - it's pretty easy to identify triggers that can identify when someone needs assistance and the best way to communicate with them. In addition to automating help with training information based on error log history and help search history, by identifying their level of expertise (are they a power user or novice?) you can customize your engagement with a customer based on their level of experience with your product or other similar products.
Creating a Data-Driven Culture
For many companies, this is a tremendous transformation in the way that they operate. If your team isn't used to being measured on their performance or isn't used to responding to issues prioritized by data, then they will have huge problems in the near future.
Here are some resources to help you transform the way your organization thinks about data, knowledge sharing, your organization:
Measure What Matters: How Google, Bono, and the Gates Foundation Rock the World with OKRs
This book is about the types of transformations that occurred at Google after implementing OKRs and how they continue to use OKRs to measure business success.
The Fifth Discipline
This book tells the story of how organizational silos can destroy a company and how to improve communication within your company and across departments.
The Ultimate Sales Machine
This book not only covers ways to improve the sales in your organization, but it's also a book on transformation and how to change the culture within an organization.
