The promise of data and analytics for product companies is that they can help you understand usage, and improve your ability to build, deploy, and service products to customers much more accurately and efficiently. By building a product analytics capability, you will be able to leverage data based on how your customers are actually using a product—rather than how they or the business think they want to use it—to ensure that you are making the best decisions in product development and targeting.
There are three key areas of investment that can lay the foundation for this functionality:
Understanding the customer life cycle. Leading companies invest in integrating and analyzing data across the life cycle—from sales, to product use, to support—to identify patterns, accelerate beneficial outcomes, and delay or avoid adverse ones.
Driving product engagement with behavioral data. Identifying areas of struggle and opportunities for enhanced user experience through behavioral data is also called data-driven product development.
Building an integrated analytical environment. You can deeply integrate its analytical infrastructure with product and sales infrastructure to enable differentiated user experiences across product use, sales, and support.
In this post, and a future one titled “Driving Product Engagement with User Behavior Analytics,” we’ll expand on these opportunities by describing their business rationale and expected value, as well as potential approaches to realize that value. This post will look at understanding the customer life cycle; the next will look at driving product engagement with behavioral data and building an integrated analytical environment.
Understanding the customer life cycle
Your customers develop a relationship with your products and services over time. They begin using an initial product, then discover other uses and opportunities within your suite of offerings, and hopefully acquire additional features and entitlements. Developing a deep, data-driven understanding of that life cycle across different customer and user types can help you accelerate events such as the purchase of new entitlements, and delay or avoid negative ones such as feature or entitlement abandonment and cancellation. Leading software companies now build their products from the ground up with the instrumentation to support this kind of analysis.
Creating value via customer segmentation
It’s important to understand how your users get value from your products, and how their relationship with you grows over time. Customer representatives may understand each customer as a whole, but may not have insight into the customer’s individual users, and they rarely have visibility into user behavior across all of your customers. Profiling and segmenting users both within each company and across companies can provide valuable insight into your customers’ needs—insight you can use to shift revenue forward and drive incremental revenue.
SVDS has developed many such segmentations with our clients who seek this kind of industry-leading capability. The general approach is to identify and collect behavioral characteristics of users, engineer features (derived data fields) that express these characteristics, then employ one or more clustering algorithms. The result is a set of behaviorally distinct groups, whose activities can be interpreted to drive beneficial life cycle outcomes.
The value of segmenting each customer’s users comes from insight into how the customer uses your resources, what their patterns of behavior are, how many are involved, and other finer-grained usage details. Such insights could:
Drive cross-sell and upsell opportunities by identifying users with similar behaviors who may use other products (more on this below).
Identify new user segments by analyzing behavior across your customer base. These segments can be an input to development of new products and features, or refinement of existing ones.
Allow reporting of key KPIs by user segment to understand the health of different user types over time as well as differentiate the effects of new features and product changes.
Inform new product packages or offerings by identifying the differentiated needs of users.
Customer segmentation can take advantage of data on how each user interacts with the product, sales, and support. You might already collect usage and support data in varying degrees for each of your products. Clickstream-level data is not required, but at least a log of requests or commands executed per session, and possibly the duration of use, is ideal for best differentiating behavioral patterns in product usage. Demographic data (job title, etc.) on each individual can be useful, if available. The ability to relate support interactions to product use and contextual CRM data can enable a very powerful understanding of customer life cycle capable of driving tremendous value.
There are two focus areas where customer segmentation is often used to drive value: actionable customer intelligence that supports cross-sell and upsell opportunities, and churn prediction and avoidance.
Cross-sell/upsell
Often, each customer relationship is unique; the past behavior, needs, and preferences of each customer are carefully considered in the sales process. These preferences may not always be stated explicitly, but can be inferred from the mix of products or services with which a customer has already interacted. The focus of this capability is to deliver insights to your sales organization in order to better identify cross-sell or upsell opportunities. Specific relevant analytical capabilities include:
Identifying cross-sell opportunities for existing customers via frequent itemset, market basket analysis, and/or collaborative filtering. Such analysis would involve identifying the sets of products or services which are most commonly used together in order to answer questions like, “How much more likely are you to be a user of product/service A, given that you’re a user of product/service B?”.
Anticipating customer needs by understanding common patterns in the sequence of “touch points” that customers engage with throughout the business and within a given product. By identifying common paths that customers take through your product offerings, you can make informed product recommendations to them based on quantitative evidence.
Identifying common customer pain points by analyzing customer support data in conjunction with usage data.
Such sales-enabling analyses could rely on a number of data sets, including: product usage data across your entire suite of products, historical entitlements, sales interactions, purchase history data, support tickets filed by users, and other explicit feedback/requests from users.
This work creates value by allowing you to :
Enable your sales team to close more new contracts with existing contacts by leveraging internal data sources as rich sources of cross-sell and upsell opportunities
Anticipate customer needs by understanding common trajectories across your services
For this approach to yield the most success, the results must be fully integrated into the sales process and adopted by your sales teams. Such adoption can be tuned to increase performance over time. By integrating feedback loops that help to identify the right set of offers to the right set of customers at the right times, you can continually improve your sales process to to execute the maximum number of upsell and cross-sell opportunities.
Preventing customer loss through detection of early indicators
In many situations, customers do not abandon a product or feature suddenly, but exhibit signs of disengagement prior to leaving (churning). The purpose of churn prediction is to learn patterns of user behavior preceding churn that enable you to intervene before a customer leaves. Once you learn these patterns, there are two actions enabled. First, any behaviors that indicate a deficiency in the product experience can be addressed by product developers (more on this in our future post). Second, you can monitor behavior patterns indicative of disengagement, and your customer service team can be alerted of the potential customer issue. When such behavior is detected, you can intervene in some way to prevent the loss. The particular intervention actions are up to the customer service team, but they usually involve contacting the customer to assess dissatisfaction and possibly offering incentives to stay. These interventions should be tested, as some retention efforts have been known to backfire.
The primary value of churn prediction is loss prevention, including loss of customers to competitors. A secondary value of churn prediction is the insight that churn models may bring. By capturing the behavior patterns of customers just prior to churning, the model may reveal information about customer (dis)engagement that is useful to the sales and product teams.
SVDS has developed churn prediction capabilities with several clients in varied contexts. The general approach involves first modeling customer life cycles and turnover rates to determine a baseline—what a typical “life cycle” is, how often customers churn, how successful our client is at retaining them, etc. We then apply machine learning techniques to historical data to learn the typical behavior patterns that occur prior to churn. The result is a model incorporating these patterns. We take measures to ensure that the model works as well on additional historical data to ensure it will perform acceptably in the field. Finally, we deploy the model and monitor its predictions, building an alerting mechanism for the account management or sales team.
In our future post on this topic, we’ll take a closer look at driving product engagement with behavioral data and building an integrated analytical environment. In the meantime, if you’d like to talk with one of us about better understanding your customer life cycle, please get in touch.
The post Data-Driven User Engagement appeared first on Silicon Valley Data Science.