I used to be the PM for one of HubSpot’s freemium email tools. It was a great experience launching something brand new in front of thousands of people at our annual conference, and then growing it to hundreds of thousands of active users. We watched our metrics extremely closely, and we knew how fast we had to grow to hit our goal for the next year. I would frequently get questions like “are we having a good week?” and I felt the pressure to give a reliable and accurate projection. If things were going poorly, I needed to be able to understand why and sound the alarm so the team could take action. That said, you can’t have many false alarms. You don’t want to come across as the boy who cried wolf.

I ended up building a simple model in Excel (Google Docs link) that helped me visualize the current week against previous weeks, so I’d be able to make a reasonable guess as to whether we would grow, hold steady, or lose ground.

For example, lets say that 100,000 people used an app on a weekly basis (WAUs). Not every person used it every day, so our DAUs (daily active users) was less than 100,000. Lets say it was 65,000 DAUs. I recorded the cumulative number of people that had used our app so far that week. It looked like this:


I then added columns going back in time, and kept track of it going forward. This gave me a nice set of data to play around with:


The pattern of usage each week was incredibly consistent. I graphed each daily total against the final count of users active for that week:


This chart shows that our pattern of usage each week is incredibly consistent. Between 60 and 70 percent of our weekly total uses in Monday, and between 75 and 80 percent of our users have used in by the end of the day Tuesday.

The way I read this chart is that we are better able to predict what our total user count will be by the end of the week. As our user base grows larger, it’s harder for our user acquisition to have as much of an impact. New users can definitely increase, but that’s a good problem to have.

Using these numbers, I updated the model to project what the high and low bounds of what I’d expect for our total usage for the week:


Interested in doing this kind of analysis or learning how to build world class products through quantitative analysis? We’re hiring for data analysts at HubSpot, read about the position here.

Disclaimer: all of the data in this sample and the screenshots are fake. They were made up and not reflective of any HubSpot product.