Reply to thread

So I pulled some reports and evaluated the correlation between Qweb and Sales.  I pulled data on 6 stores as this has to be done one at a time and is kind of a PITA, the results were getting clear so I figured it would be enough to tell the story.  Basically Qweb Used goes up so does used car sales (58% is a strong association)

Qweb New = Qweb criteria + they looked at a new VDP

Qweb Used = Qweb criteria + they looked at a used VDP

I would also use this an efficiency metric from marketing engagement to sales.  The higher the correlation the smoother / more consistent the process.  More random = high variability.  High variability means the store is not managing to leading indicators.  Each month is bringing a major course correction from one extreme to the next.  This can show up for a number of reasons, here are some common examples we see.

1. Used car supply goes up sales go up, then used car supply falls sales falls, process repeats
2. GM compensation based on net profit influenced by previous month overall bonus payouts.  Dealership does well / bonus payout goes up.  Next month GM doesn't try as hard (as they are hit with large increase in expense) results go down, process repeats.  SPIFS cause this as well.
3. Every other month direct mail/TV/SEM fill in the blank here.
4. Dealership does well, dealership hires more, dealership converts worse and sells less, people leave, process repeats

If you think about your store just remember variation is bad.  If you want to get nuts on reducing variation check out Six Sigma (the DMAIC model is your friend)

Here is the detailed data on the that orange Ford Anomaly.  This essentially means the results are completely random, this is a completely possible result, but definitely an outlier.