I just ran a query for one of your peers ... well, not a query, but a model. I'll simplify the model for you, so that we can all understand the message behind the model.
I analyzed first-time buyers in the past twelve months, modeling the amount of future demand these customers will generate. There were only two dependent variables.
- Dependent Variable = Demand Spent Next Year (2012).
- Independent Variable = Demand Spent Last Year (2011).
- Independent Variable = 1/0 Indicator ... Did Last Year's Purchase Include A Discount/Promotion Code ... 1 = Yes, 0 = No.
Here's the outcome of the model (again, this is over-simplified to make a point):
- Future Demand = $28 + $0.220 * (Demand Spent Last Year) + $8.00 * (Did Customer Purchase Include A Discount / Promotion Code).
This company mails about 12 catalogs a year to this audience. This company converts 45% of demand to profit. This company produces catalogs that cost $0.50 each. This company spends an average of $2 per customer on online marketing.
This company utilizes 20% off promotions.
- 10% of future demand among customers not historically using a discount/promo code are discount focused.
- 20% of future demand among customers who historically used a discount/promo code are discount focused.
We now have enough information to determine the future impact of discounts/promotions on this audience. Let's assume that each customer spent $150 last year.
- Full Price Customer Future Demand = $28 + 0.220*150 = $61.00.
- Discount Customer Future Demand = $28 + 0.220*150 + $8 = $69.00.
So far, the discount customer is "worth more", in terms of demand ... 13% more.
- Full Price Customer Profit = $61.00*0.45 - $61.00*0.10*0.20 - 12*$0.50 - $2.50 = $17.73.
- Discount Customer Profit = $69.00*0.45 - $69.00*0.25*0.20 - 12*$0.50 - $2.50 = $19.10.
Oh oh.
On the surface, with all of your KPI-infused dashboards, you're increasing customer value by 13%. Google Analytics tells you how successful you appear. Great! Except you're not successful. You cut future profitability by 7%.
By the way .... this is a common outcome. In fact, the outcome often looks worse than this.
Of course, you have to do the math to know this.
And almost nobody does the math.
Maybe this is a good time to do the math? Or to hire me to do it for you (click here).
UPDATE 2013.01.07 - 12:29pm PST: One of our readers noticed that the discount customer is more profitable in this example. The reader is correct. In the example I cited, discount customers spend $8 more in the future ... when I ran the models, discount customer spent less in the future. I didn't want to use exact results, so I tried to protect the discount shopper, and by doing so, I altered the results, and then completely invalidated my own story. This is my mistake.
UPDATE 2013.01.07 - 12:29pm PST: One of our readers noticed that the discount customer is more profitable in this example. The reader is correct. In the example I cited, discount customers spend $8 more in the future ... when I ran the models, discount customer spent less in the future. I didn't want to use exact results, so I tried to protect the discount shopper, and by doing so, I altered the results, and then completely invalidated my own story. This is my mistake.
Seem to me the material factor in this formulation is:
ReplyDelete"20% of future demand among customers who historically used a discount/promo code are discount focused."
That % seems like a hard thing to know, and even if it's knowable, I would bet that 20% is low.
This data is from an actual customer project.
ReplyDeleteIt's knowable. It's an easy query to run. Freeze your file as of 1/1/2012, identify customers at that time, then measure the % of volume from 1/1/2012 to 12/31/2012 that is discount oriented.
Now, if you're Lands' End, well, then you have problems, because 89% of future volume might be discount oriented.
This is realistic, actual customer data. No need to speculate if it is hard to know or guess if it is low. This is what happened.
As mentioned in the update, Doug, I made a mistake in the math, which almost nobody caught. The actual result (% of future demand) is what happened ... the lift for discounting was, in the actual project, a loss, making the math even worse for discounting.
ReplyDeleteTo your point - for many companies, the future discount percentage is higher than what I found in the example I cited.