Every successful SaaS company focuses intently on their customer’s success. If your customers are not successful, your company will not be successful. A successful customer achieves both tangible and intangible value from your product. They execute their tasks quickly and effectively. They discover new functionality that fixes more of their problems. They develop strong loyalty to your product and your brand. Successful customers become your power users.
Customer success is hard. Teams are often human driven and difficult to scale. This can be solved with product-driven customer success. Through personalization, your product can enhance and scale your customer success efforts. Product-driven personalization facilitates successful on-boarding, reduces churn, and creates power users. It serves the users who can be successful with personalization alone. It can flag which customers are most likely to up-sell or churn. It is a natural extension of your customer success team.
Tangible Customer Success Benefits
Some companies treat customer success as customer support. It becomes an afterthought. Customer success is not customer support. It engages users and preempts problems. It does not react to problems. It provides the customer with solutions to extract maximum value. It generates revenues and drops churn. Jason Lemkin, a VC and the founder of SaaStr, says, “Customer success is where 90% of the revenue is.”
Driving Successful Upsells
Customer success drives revenue through upsells. A successful upsell is not forced. The user grasps their problem, and perceives how your extended offering will help. The customer success team should focus on making the user aware of their problem, and presenting them with your extended solution. If done correctly, it will sell itself.
Customer success reps should not only drive extra revenue directly from the customer. Not every customer can be upsold. As a customer I may only need your base solution. If I perceive more value than cost, I will not churn. If a representative attempts to force an upsell on me, it will backfire. No one wants to be sold something they don’t want or need. Even if you manage to force the upsell on me, it will eventually fail. I will eventually perceive more cost than value and tone down. I may chose to cut your product completely.
Increasing Word of Mouth Referrals
Customer success also drives revenue through word of mouth referrals. This can be far more common than direct upsells. For a customer to refer you, they must be extremely successful with your product. Just solving a tangible problem is not enough. They should have strong positive emotions associated with your brand. The solution cannot only be good, it must be great. Customer success metrics like Net Promoter Score (NPS) understand this. They only account for “great” ratings, dropping the “good” ones.
Customer success drops churn. If a user perceives more value than cost, they will not churn. Customer success should drive the customer to value quickly. The customer should have their problems solved. The problem you solve should also be meaningful and difficult to solve. They should not consider leaving after using your product.
The value of a successful customer success effort cannot be overstated. Executed at the highest level, it can result in negative revenue churn, the holy grail of SaaS. Your revenue will grow regardless of the quantity of new customers. Your new customers lifetime value will surge. The value they provide to your business continues to compound.
The graphs below describe a business by the revenue from the view of monthly customer cohorts. So each bar is the amount of revenue per cohort. The top has cohorts whos revenue shrinks 5% month over month. The bottom one grows by 5% month over month. The amount of new customers and initial revenue per new customer is the same. The difference in end result is striking. The compounded effects led to 2x the monthly recurring revenue (MRR).
This model ignores the increase of word of mouth referrals that the company described by the second graph would acquire. They would be making even more money. Customer success works.
Customer success may reward you generously, but executing it to this degree demands hard work. In a perfect world, one of your customer success agents could be sitting with a customer at all times. They could answer their every question and handle their every need. This is impossible. Even if you had the people power, your customers will often prefer to figure things out themselves.
Personalization-Driven Customer Success
Your product must become a virtual customer success representative. Your user experience must be personalized
A customer success representative works by understanding the intent of the customer. They identify the problem the customer faces and advises them on how to proceed. Concretely, this can mean showing the customer what to do on your application to solve their problem. It can also mean presenting them with the correct content, or pointing at the next tab.
This interaction with a customer success representative can be mimicked by the product. The virtual assistant can do the same. Based on the behaviour of that user and similar users, your system should be able to predict intent. The UI will be able to take this intent and react.
Personalization = Intent Prediction + Reactive UI
After a user uses your application to sign a contract, they may intend to send the contract to them next. The application should react and transfer the user to that page.
If this user tends to use email first thing when logging in, the application should react by displaying the email tab first.
If a user writes an email that they intend to send to email@example.com, your application should display that email as a suggestion.
If a user was setting up an email integration, then clicks on documentation, they intend to read about your email integrations. The application should react and display that at the top.
If a user tends to look at billing first thing Monday when they log in, the application should react by uncollapsing the tab. It could even react by emailing it to them weekly.
If a user begins to upload many rows one at a time. The user likely intends to upload more. The application can react by highlighting your mass upload feature.
Why it works
The goal of customer success is to help the user get initial value with the base solution, help them discover incremental value through your extended feature set. Personalization does both here.
It begins by leading the user to their intended location. This can be done by segmenting your user base. Different segments likely have different intents. By quickly classifying the user by their behaviour and their attributes, you can optimize for this.
Once the user starts using the application more, you can streamline their process by reacting to minuet intents. This will help them discover new functionality in your application and transform them to a power user.
You can also help them discover new features that are only available in the next tier to drive up-sells.
The Human Touch
Your personalization tool generates invaluable analytics. It will segment your user base for you according to behaviour. The intent prediction may also be used internally to qualify a user for upselling or to flag a user who may churn.
Personalization does not replace human customer success representatives. It makes them more effective by acting as a tool to delivering the ideal experience. It serves the users who can be successful with personalization alone. It flags users that require a human touch. Personalization augments and empowers customer success representatives.
Segmentation can help define different user types and their behaviour. Trend analysis can explain how each segment uses your application. A human customer success representative can apply this information while talking to customers. By asking the right questions, or seeing that customer’s behaviour, they can understand how that customer will want to use your application.
Personalization requires two components, intent prediction and a reactive user interface. Both components require data analysis to design.
You must collect user event data. Actions performed on your application should be captured on your UI and stored in an event collector service.
You begin this process by tagging each UI button and action that has semantic value. Name each action and track each event by sending it to your event collector service.
We opted to build our event collector from scratch for seamless integration into our personalization platform. Our event collector batches and sends events through its client library. The service runs on AWS and pushes events to Kinesis where it is simultaneously stored in a few different services, like S3 and Redshift.
Building from scratch is difficult to maintain and scale. You’ll end up collecting huge amounts of data from a variety of sources and having to store them for efficient retrieval.
Data must then be analyzed. You should begin with high level metrics, such as number of clicks and average clicks per user. Companies often build a dashboard for their global metrics. It makes usage and business state transparent.
You can continue to enhance your analytics capabilities. You can analyze user paths through your application to understand typical user flows. This technique is commonly known as flow analytics. Your UI can be optimized by adding shortcuts or simplifying pages. Your UI should match your user’s journey through your application.
Flow analytics allow you to identify key intents. It facilitates breaking down your UI, and clustering it into hierarchies. It classify each intent a user might have on any given page.
With your foundation in place, you can begin implementing personalization. It’s simplest to begin with static personalization. This means showing users differents pages depending on attributes you know about them. For example, a university should show a different home page for a faculty member and a student.
Machine learning holds power in this space. Unlike other concentrations of machine learning, intent prediction is nondeterministic. Two users who’ve done the exact same thing may have different intents in the same context. The lack of context and data further confounds the problem. You only have their events on your website to work with. You are given a chain of sparse events unevenly dispersed along a time axis.
One approach provides a model a user’s attributes (their location, filled out forms, etc.) and the current page they are on to predict their intent. This can suffice for simple applications; however, intent prediction often requires the past series of events from that user.
An intricate approach feeds a recursive neural network the series of events, and predicts what a user will do next. This model can generalize user’s by their actions and can perform well in production. This model suffers from the cold start problem. Users who have not done much will be unpredictable. This model cannot take attributes into account such as location and user type which may resolve this in the simpler model.
Triton’s system uses a hybrid approach that can outperform both of the above methods by over 40%. It can take attributes and all past events into account. It also can cluster users to enhance your analytic capabilities.
Your application should be smart enough to react to intents. If your intent prediction system infers that the user will want to sign in as faculty, the application should be able to change to facilitate this.
Reactive UI must be designed. It cannot be automated. It requires you to understand your user flows and figure out where the forks are. These are locations where significant populations of users act differently. The prediction mechanism should be able to predict which fork the user will take or communicate that it is unsure. Depending on the prediction the UI should change to facilitate the flow. This can be minor, like reshuffling the order of the top bar. Google often takes this approach.
If you search dogs, Google will prioritize images and videos.
If you search dow, Google will prioritize finance and news.
If you search buy shoes, Google will prioritize shopping and maps.
Some landing pages personalize aggressively. Most companies have different customer segments that are interested in different things. If you know the customer on your page is enterprise, through their email or their behaviour, you should show them a landing page that highlights your strength for them. If your user is in the farming space, you should focus on case studies and quotes from that space.
Personalization ends up working like a customer success representative in simple cases, and augments them in more difficult cases. It facilitates the user providing value at every turn. Product-driven personalization scales and fine tunes changes in ways that even a customer success representative could not. Product-driven personalization makes your users successful.
Customer success drives revenue through upsells and referrals and reduces churn. It provides your SaaS company with large tangible and intangible results. Customer success is hard to scale, especially if you sell to many SMBs.
Product-driven customer success can scale the process and provide value at every turn. It requires a reactive UI and intent prediction. These components require a data-driven approach to design and build.
This article glosses over many intricacies and minutia required to build a personalization system. It is hard to do in practice, and even harder at scale. We’ve built Triton to help you personalize your application.
Triton personalization turns your application into an experience. We spearhead your personalization efforts. We provide the data collection and analysis required to effectively design your reactive UI, and provide APIs for recommendations and intent prediction. Our mission is to make your user experience feel human. If you’re interested to learn more, send me an email at simba at triton.cloud or sign up on our website.