Software-as-a-Service (SaaS) continues to be a hot area for startups. The Responsys IPO filing shed more light on the numbers behind a larger scale SaaS business, including ratios of license to service revenue as well as growth over many years. As a SaaS company, whereby clients essentially rent the software, and thus are financed compared to paying a large up-front license fee, it is critical to understand if you’ll be making money over a long-term horizon as there’s a great chance you’ll lose money in the short-term due to the nature of the business.
Here are some SaaS growth metrics we track:
- Churn rate in terms of number of clients as well as in dollars
- Monthly, quarterly, and annual recurring revenue growth
- Client acquisition costs as well as how many months/years it takes for a client to be profitable
- Omniture Magic Number – ratio of sales and marketing costs two quarters ago to new annual recurring revenue from last quarter
- Average revenue per customer/user
- Lifetime value of the customer as well as the lifetime value discounted against the cost of capital
- Cost of goods sold (typically hosting and customer service fees) per client
Managing and tracking these SaaS metrics help us better understand our company as well as benchmark us against data from publicly traded SaaS companies. My recommendation is to prepare a monthly analysis of this type of information.
What else? What other SaaS startup growth metrics do you track?
David Cummings on Startups 
David,
Great metrics! Many SaaS companies offer a free trial. Frequently the highest rate of cancelations occurs during that free trial – once customers get over the hump, they’ll stay a while.
We track the % of customers that stay with us past the trial each month. A small bump in this metric can have a big impact on future revenue.
David, love the blog. A question, when starting out, what is the best way to project the average lifetime revenue per customer for SaaS? We (kinda) know the price per month, but how do define the lifetime? Assume an average duration (eg. 2,3,4 years) or base it off of an average or expected churn rate in the beginning, until you have more accurate information?
Determining churn without some operating history is difficult, as you’ve already diagnosed. I’d model out different time frames from 2 – 5 years, as you mentioned as well as look to comparable products or industries that would be good proxies (e.g. publicly traded companies).