Our 📝Integrated Financial Model (IFM) forecasts based on transactional metrics rolled up and counted monthly or weekly.
All of the data for our portfolio of 📝Rapidly growing companies lives in 📝BigQuery.
Projects
Background
We use ETL tools like Stitch and Coupler to pull data directly into the IFMs in Google Sheets to do our forecasting.
The process looks like this: once landed in Google Sheets, transactional data from every app is scrubbed and annotated, then pivoted by time for analysis in the IFM.
We use the same process regardless of which app the source data comes from.
It gets less manageable when there are more sources. It gets a lot less manageable when there are more sources and we want to manage the business according to our 📝Weekly Philosophy. By landing the data in BigQuery, we can achieve more reliability, better data for the management team, and less time for our team.
And then we forecast
Today we create our comprehensive business forecast in the IFM with simple math on data abstracted to the highest level. For example, if we have experienced an average of 28% repeat purchases over the last 26 weeks, then we might predict the repeat rate would stay at that average into the future. We might seasonally adjust it. We might allow it to slowly increase or decrease. But we're doing 📝Tops Down forecasting with very basic math.
Time Series Forecasting with GCP Vertex ML
We are looking for help connecting our BigQuery data into Vertex ML and learning how to train models.
(Full instructions to get to this screen here.)