Response rates are not a meaningful currency

Marketers sometimes compare response rate metrics. It's often quoted that direct mail campaigns routinely generate response rates of only a percentage point or two, or even lower (typically much lower for on-line advertising). A few lucky souls trot out their stellar double digit response rates in success story case studies. But are these really success stories ? Response rates are a poor indicator of effectiveness, and they are not difficult to manipulate.
The better targeted any marketing campaign is the higher the response rate. This fact is used to justify targeting, and along with it data mining, CRM etc. The logic goes like this... the better targeted the piece is then the more relevant it is to those that receive it, the gain in efficiency comes from not trying to sell things to people who aren't interested (eg women's clothing to men).
DunnHumby report great response rates for coupons printed off for Tesco customers at the shopping checkout. These promotional coupons are targeted to people who buy the brand, and hence are relevant to them.
Now this all sounds highly sensible, very professional and modern. But it shows that response rates can be manipulated by targeting. What's wrong with that you cry ? Surely a better targeted campaign shows higher response rates as a sign of the efficiency of targeting.
Well first there is the problem that response rates are proportional, they don't say anything about total effect size. A 50% response from targeting 100 people is less than a 10% response from mailing to 1000 people. Small highly targeted campaigns typically deliver high ROI but little in the way of absolute profits.
Secondly, what direct marketers call relevance is often another way of saying "they were going to buy anyway".
Here's an example... a bank sends a letter promoting investment products to its customers who are very close to paying off their home loan. They get a high response rate and they credit all the investment product sales to the direct mail campaign. But bank customers have high loyalty, a huge percentage will take out investment products from the bank that provided their mortgage.
Here's another example... a department store combs through its customer database and finds a few women who are regular buyers of bikinis, rightly so they identify that some other women rarely buy bikinis. So they produce a targeted mailing offering bikinis to the more frequent buyers and it gains a higher response than previous mailings that went to all women on the database, and the marketers pat themselves on the back and bask in the glory of a high response rate. Of course it is higher, these women have a higher than average propensity to buy bikinis (and they probably go to the beach a lot too). Even without the mailing they would have bought more than the average.

Plus part of the recorded sales effect is overstated because the campaign brought forward sales that were to occur anyway.

The tighter the targeting, the greater the response rate is contaminated with buying that would have occurred with or without the campaign. Point of sale price promotions are highly targeted at people with a high chance of buying the brand anyway, and typically two-thirds of a brand's sales when on deal would have occurred anyway.

The bottom-line is that campaign response rate is of little concern. What really matters is the total response (after deducting the sales that would have occurred anyway), and the net profit contribution (after the marketing costs).

So while it's a no brainer that targeting people who aren't in the market can reduce marketing efficiency, it doesn't follow that tightening up the targeting delivers higher returns. Tighter targeting delivers higher response rates, but these need have nothing to do with profit effectiveness.
Response rates are only comparable if they compare campaigns sent to identical targets.