How We Increased Gross Revenue Retention by 20%? Part-1

7 min readDec 30, 2021

In SaaS (Software-as-a-Service) business model, gross revenue retention measures how much of your monthly recurring revenue (MRR) you retain each month after you’ve subtracted the effects of churn or downgrades to lower-priced products, but not the effects of upgrades. Gross revenue retention can be calculated monthly, quarterly, or annually depending on your selling model and typical subscription term.

Gross retention gives you a view of how stable your revenue is when you’re considering customers who either renewed with you, decreased their monthly spending with you or stopped buying from you. Having the increasing new customer acquisition cost, retaining revenue stream can greatly help any SaaS business.

What is considered good gross revenue retention?

For SaaS companies selling into small and medium businesses (SMBs), a good Gross Revenue Retention Rate is 85%. For Enterprise SaaS, 95% is considered a good Gross Revenue Retention Rate.

My journey to increase gross retention

Back in 2017, when I took the CEO role at a MarTech SaaS company, our business struggled. One of the challenges we had to solve was the low gross revenue retention rate. Back then, we had over 2,000 customers where about 10% of them were small businesses, 70% from the mid-market, and the rest from the Enterprise segment. The gross revenue retention was at the low 60%. Way below the benchmark.

When you have such low gross revenue retention, you know that something is wrong in the way the business operates or something in the product is wrong, or any combination of that. Where do you start?

First discovery round

The first thing we did, to learn about the challenge, was to interview the existing customers, churned customers, and our own sales and support teams. We asked questions about the reasons leading our customers to churn, or not to renew their subscriptions. The results of this discovery project were mixed. While some said that the product is too expensive for them to purchase, or renew, others blamed the product for missing features. On the positive side, almost everyone valued the user experience of the product as well as the responsive and valuable customer support that we had.

Having these findings, we decided to accelerate our investment in the product to close the main gaps. We started to add features all across the products. Dozens of new features every quarter. We also invested in re-training our sales team about the main product features so they can better present them. The account managers were also re-trained to help customers understand the product features better.

First results

Ten months into the journey, we started to see results. Our gross revenue retention is now hovering around 70%. We felt that we managed to find the root cause and the trend will continue. We expected to hit the 80% target soon. We continued to invest, work hard, and even harder, however, the gross revenue retention didn’t increase further. Despite all the efforts and time, the gross retention bar remained around 70%. At that time we knew that the root cause of the problem wasn’t identified, nor solved. More work is ahead of us.

Second discovery round

Two years into the journey, we looked for a different approach to solve the problem. If the first round helped us to increase the gross retention by 10%, there must be another discovery that will help us to add an additional 10% at a minimum.

Customer health index

One of the ideas we had was to create a health index for each customer. The health index will be calculated weekly to give us a leading indicator of healthy or sick customers. The health index was calculated based on the volume of activities each customer had with our product every week. We used different machine learning and prediction models to fine-tune the model. The outcome of the health index was simple to use. We bucketed customers into 5 health index buckets, where 5 = healthy and 1=sick. We could also calculate the probability of churn for clients in each bucket. The customer health index had a churn prediction accuracy of between 75–85%. Now we had a good model to predict churn.

Although the model helped our sales operation team to refine the quarterly business results prediction, it didn’t help to improve the gross revenue retention. It didn’t help us identify the root cause of the problem — why do we have unhealthy customers that eventually churn? What do we need to do to move these customers into healthy buckets (buckets 4 & 5).

Despite this initiative, the gross retention bar remained around the 70%. Where do we go from here?

Third discovery round

We didn’t give up. There must be a solution that we need to find. We knew it is there, we just need to be more creative in finding it.

One day I asked the team, “What do we know about how customers are using our product?”. I received many answers. From the features they use to the questions they ask as well as the wishlist they have. I could not get a clear answer to my question. We didn’t have any analysis to answer the question. So here is where our third round of discovery started — how customers were using our software.

How customers were using our software?

In November 2019, we celebrated 10 years for the business since it was founded. Ten years of hard work in making our product great. Ten years of features over features that were added to the product by dozens of developers at a high velocity. The product was feature-rich. Although we did have in-product analytics (we started with the free Google Analytics and migrated to the paid MixPanel in 2018) that collected every click the customer had on our product, the collected data was mainly used to measure usage per feature by our product management team, as well as identify errors for our engineering team. We never tried to cluster features into groups and measure the usage of each group nor the correlation between them.

The goal of our third discovery round was all about measuring product usage and correlation between clusters of features, and product modules, to learn what are the most used clusters and how customers are using them. The result of this project opened our eyes to something really important about our customers and how they use our product.

In our study, we clustered features into eight buckets. Each bucket was called a module. We then measured the usage (or activity) of each module by each of our customers. We normalized the data and used a stacked bar chart for visualization. We created beautiful art that very nicely visualized how our customers were using our software. It was a groundbreaking moment when we first saw the results.

Visualizing product usage by clusters of features

The following chart presents the usage of module-A (brown) across our customer base (volume of activities). Some customers are using it almost exclusively, while others are using it together with other modules. We identified a very interesting correlation between module-A (brown) and module-B (blue), while module-C & D (orange and yellow) are separate from them. What does it tell us? We learned that we have a group of customers who focused on modules A & B while almost ignoring the other modules. In other words, they will not care much about a new feature we will add to module-C & D, but will highly appreciate improvements in A & B. That also applies to how our sales team will present the product to such customers or where our account managers will spend their time with the customer.

As we looked further into the usage of each module, we discovered more and more patterns in the data. For example, we identified a specific cluster, module E (blue), that was completely isolated from the rest. We learned that we have a group of customers who are using our product just for that module and almost do not care about the rest.

Churning customers and their usage of the product

Another view of the data was about churned customers. How did they use each of the modules in the product that eventually caused churn?

For example, we learned that module-C (orange) was very dominant across all customers that used it and churned.

At this stage, we felt that we have a much better understanding of what is causing the churn, what modules are driving retention, what are we missing in the way we operate, and what our way forward should look like.