If you ran an online-only store aimed at consumers, taking dozens of orders a day, it would be easy to measure the return on investment of any advertising you tried, and to do so quite accurately. I doubt that many of you live in that world.
Your advertising is more likely to be about generating enquiries rather than orders, so things already start to blur. If these enquiries arrive by all manner of means, even tracking their source is hard. And if you only receive a few each week, the data will never be significant. So there’s an argument for not even trying. Media selection by intuition is the fallback method.
I think we should all try to do better. In the past, when technical advertising was magazine-dominated, many enquiries arrived directly from the magazine. This made things easy. Even so, other enquirers – principally telephone callers – would still routinely be asked what led them there. Nowadays enquirers have probably already spent time on the company website, so may be less likely to recall. It’s still worth asking though.
The minimum we should be doing with each item of advertising is considering, from the start, how its response can be measured, and making the effort. This won’t be the same for every advert, and we won’t get data which might (for example) allow us to measure magazines vs. direct mail vs. online. But we might be able to measure a page advert against a quarter page, one mailing list against another, or a web image advert against PPC search advertising. Just because we can’t apply the same measurements to all media, doesn’t mean we shouldn’t measure at all. It’s all data which will take us one step further than ‘intuition’ in future.