Projections and Forecasts

by Neel Desai

If there's one thing we learned in the past few years, is that one of the best tools you have to in order to operate and execute in the present is being able to forecast what the future will look like.

This is why throughout the Trends sections of ProfitWell (e.g. MRRRetention), you can now see end-of-month forecasts for all of your metrics.

They feel pretty magical. How do they work?! Glad you asked. There are three steps in the process.

  1. We find a model that best fits your company. Forecasting metrics for a company with 100 customers is a very different proposition than doing so for a company that has 50,000. Doing it for someone whose company is growing 20% a month is very different than doing it for someone else's growing 20% a year. How do we work around this? We've got six different models (that is, ways of projecting your metrics), and we try each of them on your historical monthly data (in real time!). We select the one that best fits your progression.
  2. We then make an a priori projection. What does that mean? Let's try with an example: It is January 31st at 11:59PM, we have all of your data for January and February is about to start. At this point, we generate a projection for February without really having any data for said month. As the period progresses, we are going to continually (every hour or so) reevaluate the projection we originally made. 
  3. We then automatically readjust the original forecast based on intraperiod data . On ProfitWell, we slice your data by month (see MRR,PlansRetentionCohorts), and by day (see the DashboardGrowth). The original forecast we made with your monthly data will get continuously refined using your daily data (which gets updated in —almost!— real time). So, as the month progresses, the original projection gets revised tens of times a day, getting a little more precise every time.

How Did We Do?