The danger of Small Data

There’s a lot of talk at the moment about ‘big data’, but a more pressing concern for many companies’ marketing should be ‘small data’. There are many suppliers in online marketing (including Google, with its AdWords) who are far too keen to sell their capabilities to businesses which just don’t receive enough data to justify it.

We see this with Google all of the time. They encourage almost everyone to use the same sort of ‘conversion’-led management as works for very high volume transactions, when it’s really not appropriate. If you’re selling 1000 hotel rooms a day – directly from an advert – and find that one type of advert produces a lower cost per sale than another, that’s a definite steer as to where you should be putting your money. But if you’re in a business where you’re pleased to get one or two decent enquiries a week, the sources of those successes at the most granular level are almost random. Sure, you might find over the course of a year that most of the enquiries came from one type of advertising rather than another, but once you’re analysing within this, data starts to get fuzzy and perhaps even misleading.

Supposing you’re undertaking a search advertising campaign, and use 100 keywords, which provide 5 ‘conversions’ (enquiries using a form, perhaps), 2 from keyword A, 1 each from keyword B, C and D and 0 from the other 96. You don’t need to be a statistician to realise this tells you nothing of significance. Yet Google will try to persuade you to start investing heavily in keyword A. It might even do it for you, automatically.

This isn’t because Google is daft. It’s because this sort of data use and complexity impresses the users, who end up spending more. In reality, the keyword ‘[Company X] phone number’ registered 0 conversions, but actually caused 100% of the 10 people who clicked on its ad to make telephone enquiries, which weren’t being attributed. Keyword A, which was the company name, got 2 conversions, but people were only searching for that because they wanted to make an enquiry after seeing the company mentioned in the local newspaper, which also got no credit. And so on.

So, should we just throw loads of advertising money out there and hope something sticks? Of course not. But we all need to accept that reports churned out by automated systems need thoughtful assessment beyond the level of just numbers.