Know-How Exchange : Research/Metrics

One of the best ways to demonstrate differences in impact due to a marketing program is to run a test on one audience, saving a portion of that audience as a "control" - a group which is not exposed to the program. You need to make sure that the test and control groups are as similar as you can possibly make them and that they are made up of similar proportions of different customers (e.g., in your case, subscription length and other demographic attributes that are fundamental to your business.)

I personally like the using analytics to measure results. The idea already floated above of using a control vs the new program allows for continuous improvement and more importantly PROOF>

The test/control approach is ideal. If that's not possible, though, you can forecast what you would expect sales to be with an extrapolation of your current trend line. Then measure actual sales against that baseline forecast.

If you want to be a real purist, you may want to "clean up" your trend line before you project it into the future ... take out any seasonal effects, unusual events in the base period that might skew the trend, promotions, competitive activity, etc.

We've used both approaches -- occasionally at the same time -- and they seem to work fine.

If the forecast trend line is really a critical measure, you might consider having an analytics expert consult with you for the set-up. There are a lot of nuances in the process that can have a material effect on your measurement.