Go-Digital Growth Insights
- SQL
- Power BI
- →Three cohorts: attended and bought, attended only, bought only
- →+88% vs +44% vs +15% submissions, six months before to after
- →Custom pre and post eligibility window built into the model
Context
Go-Digital lets a prospective dentist take an iTero scanner for a three-month trial, submit real Invisalign cases on it, and buy it afterward at a discount. It is aimed mostly at doctors already doing Invisalign with traditional braces. My manager wanted a dashboard that showed how the programme was actually performing, not just how many people signed up.
Framing it as a before and after
The question was whether the programme changed behaviour, so I set it up as a cohort comparison across three groups: doctors who joined and then bought a scanner, doctors who joined and did not buy, and doctors who bought without joining. Each doctor is measured on Invisalign submissions in the six months before versus the six months after.
The eligibility window
For the comparison to be fair, every doctor needs a full six months of history on both sides. A doctor who starts a trial on 1 January finishes it on 31 March and only becomes comparable at the end of August. I built that into the model: while a doctor is still inside their post-trial waiting period they are held out, and they enter the analysis only once six months of post-trial data exists.
What it showed
Doctors who attended and bought increased Invisalign submissions by 88 percent. Doctors who attended but did not buy still increased by 44 percent, which suggests they took the training back to their practice regardless. Doctors who bought a scanner without attending increased by only 15 percent. The programme, not just the hardware, is what moves submission volume.
What I'd improve
These are raw before-and-after deltas, not a controlled result. Keen doctors self-select into the programme, so some of the lift is selection rather than effect. A matched control group or a difference-in-differences design would firm up the causal claim.