The importance of data accuracy
Google thought I was interested in espionage and security the other day. If you’ve never looked at its ads preference pages, that from your web history, Gmail and searches broadly, tries to work out who you are and what you’re interested in I’d take a look. I’ll be surprised if it gets everything about you right.
Because, I’m not particularly interested in espionage by the way. I may have looked over the Snowden revelations when they emerged, but no more than the next man, and that perhaps is the point. If Google can get it wrong, given they have more data (and smart developers able to derive insight from the rest of us), what hope do the rest of us have? In fact, in the case of a recent and insightful talk given (and widely shared) by developer, Maciej Ceglowksi, Google thought he was an Arabic speaking woman, age 25-34, whose interests include "Babies & Toddlers", "Meat & Seafood", and "Greater Cleveland" – which he isn’t.
The secret is that data collection is messy. Even social media data, which because we’re creating (not having it mined) has an error rating of 3%. This may not sound a lot but consider this, for Facebook’s 1 billion users, 30 million – the whole of Belgium and the Netherlands for comparison will be getting served useless adverts. For many other forms of data, the error rate is much higher – particularly as people don’t tell you when their circumstances change if you’re mining the data – there’s no way for people to tell you you’re wrong. Then combine data types together in the novel way and the propensity for error is even higher.
But this shouldn’t be a problem - In fact, it may well be an opportunity. Because social data – which is explicitly created and Google’s opt out form provide oversight. Provide it for end users and actually accuracy rates will be being higher – reducing error, increasing effectiveness and may even open up new opportunities.
Consider for instance a fashion brand said it wanted to check your movements and when you were wearing its clothes, so that it could work out when you needed a new pair of trousers. You’d consider that intrusive. But packaged as a data service for customers, this is exactly what Nike+ has done. It’s collected tonnes of insight, and because it provides a service, it’s actual able to provide consumers with a perfectly targeted message, so effective, it almost doesn’t feel like advertising.
Why can’t other parts of the data mix be like this? Providing services and insight for their customers. Every misplaced targeted ad destroys brand equity even if the short term sales rise. Providing services driven by data both builds brands and also helps trigger sales. It’s also a more sustainable future. Because surveillance marketing as it has been recently been called is facing a rising tide of awareness and opposition and inviting consumers in will improve everyone’s life.
Consider the change over the cookie rule as a small example in the right direction. Everyone thought it would be disastrous but actually people are happy to accept from brands they love as they see a more positive user experience. It’s time to stop hiding in the shadows, spook-like, and behind TLDR legalese and be open and flexible in what we hold about the public. More data, does not provide better targeting, unless it’s more accurate, so we need the feedback and involvement of the very people we’re trying to serve – not the companies, but their customers – people like you and me.