Googles Data: Accuracy and Validity

Googles Data: Accuracy and Validity

I’m running a Google AdWords campaign for a client with well over 100,000 keywords in the property market. The campaign is performing well but the client wants to optimise it further – by maximizing their ROI on a keyword by keyword basis. Google’s data is accurate – errors are usually less than 1-2%. But the data can’t be used beyond a certain point for decision making purposes to improve the campaign because it lacks validity. Here’s why:

Invalid Statistics

There have been hundreds of conversions, but of the 100,000+ keywords, only about 10 have converted more than once. This means proving that one keyword is better than another will take years (or decades) with 95% confidence. Well before this time the client’s markets and environment will have changed beyond recognition.

Dirty Google Data

Google’s data is accurate but it’s impossible to compare like for like due to issues with: seasonality, the weather, holiday periods, fads, fashions, media exposure and a whole lot more.

Dirtier ROI

Sales data is even dirtier than Google’s. A keyword might drive a sale from which a profit can be calculated accurately. But in the context of one conversion typically generating a sale of $25,000 or $1m – it lacks meaning. The next sale from the same keyword could be an order of magnitude different in size. So it’s impossible to attribute a profit on a keyword by keyword basis without huge data samples – which aren’t available.

Invalid Statistics x Dirty Data x Dirtier ROI = Fool’s Gold

If this approach was used we’d essence multiply the variability of the keyword conversion ratio by the variability of the ROI. The answer would possibly be accurate – but also, of course, garbage.

Understand the System Limits

Google’s systems are highly sophisticated but only by understanding the limitations of data can we hope to make intelligent decisions. Trying to be accurate without valid data is the road to campaign chaos. Fortunately the client is intelligent and understands it is better to be vaguely right rather than accurately wrong. If you’d like to discuss improving your digital marketing please contact us!