Have you ever wondered if all those hours you team spends "engaging" customers on Facebook or Twitter pay off? Or have you ever wondered what the ROI of a loyalty program might be?
There are methods for quantifying the impact of these activities.
Here's one of the easier ones.
Freeze your database as of a certain date, say August 1 of 2010. Freeze customer attributes on that day ... historical spend, recency, membership in a loyalty program, that kind of thing.
Let's use a loyalty program as an example. Create a simple segmentation strategy, then measure average spend per customer from 8/1/2010 to 7/31/2011:
Overall, those in the loyalty program spent 5% more, $2.72 per customer, than other customers.
Now, I realize that I oversimplified an awful lot, I had to in order to make the point. You get to see how high-dollar customers were not impacted as much (on a percentage basis) as low-dollar customers, suggesting that the loyalty program helped low-dollar customers more. In addition, recent customers were impacted more than non-recent customers.
If you're a statistician, use a GLM-style model to estimate if the differences are statistically significant.
If you're in Finance, you can run a quick profit and loss statement. Say you have 100,000 customers in this program. Say that 35% of demand flows-through to profit. Say that your loyalty program costs $125,000 a year to administer, between program costs and perks and discounts.
- Profit = 100,000 * $2.72 * 0.35 - $125,000 = ($29,800).
- You lost $29,800 on the program, or $0.30 per customer.
Again, most analytically minded folks and CFOs will find many ways to poke holes in this methodology ... I'm stating the methodology this way for illustrative purposes. Be creative, enhance the methodology, and come up with your own ideas.
This methodology does a reasonable job of "measuring the unmeasurable" ... you can't execute a test where loyalty members, Facebook Fans, or Twitter Followers are not allowed to engage with you.
Ok, time for your thoughts. What methods have you used to measure issues like the issue outlined in this post?