Thanks for pointing out that my figures were misleading (or, more to the point … that I failed to qualify them properly). Looking back at the post now, I don’t think I could obfuscated the matter any worse :-(. What I said:

Their future sales and profit were 46% and 57% lower, respectively, than for ads run previously at the regular price (i.e. the LTV of this new group of customers was dramatically lower than existing ones);

Two words were conspicuously (and fatally missing), and I would have to come to precisely the same conclusion as you in this case. What I should have said was:

Their future sales and profit were on average 46% and 57% lower, respectively, than for ads run previously at the regular price (i.e. the LTV of this new group of customers was dramatically lower than existing ones).

In short, the future value of the ‘via discount acquired’ customer segment proved to be many-fold less than their normally priced counterparts. Not only did they order less frequently, when they did purchase their average order size (and profit) was significantly lower, too. It was also pointed out that repeat customers who responded to the discount promotion subsequently reordered at an 18% lower rate than those who did not.

// To avoid any further confusion I’ve edited the original post. Thanks again Dave!

Your point is well taken (that you have to look at the back-end), but in the example you give, the discounted offer is still better. Assume your control would have yielded 100 customers worth $100; then your discounted promotion would have yielded 400 customers worth $43 (i.e., 57% less). That’s a total value of $17,200 (400 x $43) for the discounted promotion vs. $10,000 (100 x $100) for the control. Which amount would you prefer in your bank account?

Of course, you could change your assumptions to yield a different result. Either way, it comes down to the same thing: you have to run the numbers.