"Build a good product and the world will knock on your door"

This correspondence comes from Cologne, Germany. I'm here with a bunch of other software developers attending my fifth ESWC conference, where an assortment of marketeers try to infuse some business sense in us. Marketing and promoting software is as hard as programming it in the first place, especially for marketing dumpkoffs like most of us :)

Despite the honest endeavours of said marketing gurus, us software types would sooner rush to the nearest pub than attend another presentation on google adwords. These are the context sensitive advertisements that google shows on the right hand side next to the organic search results. If the googler searches for "file manager" and I have an adwords campain that bids on that keyword, then my plain text advertisment will show, and if the user clicks on it, google gets some pennies off me.

Many people swear by google adwords, and perhaps they have some usefulness for some portion of the micro-ISVs, but for me at last, they didn't work. Last time I tried, and assuming that we believe the conversion statistics that google themselves supply, and also assuming that the level of click fraud was minimal, I just broke even. The advertisement campain costs were exactly balanced by the extra sales I made. A zero sum game for me, and a very happy google effectively nicking all my profit!

You hear stories that $1 spent on adwords returns $7 for some, which if true would be heaven. But on closer inspection most people just assume that they are making money, judging by the extra traffic the scheme generates, without being able to quantify the exact number of extra sales. At this point I can hear his Articulateness Dave Collins protest that the secret is analysis of campains. You try a few different ads and settings and see which works best.

I may know little about marketing but I'm very strong on modelling and optimization on account of my day job. Adword campaign analysis sounds good until you realize that the data used to infer the effectiveness are practically rubbish, and that's scientific.

1. The phenomenon we try to assess is very complex so we'd need a huge amount of data (not just a week's worth) to have any statistical significance. Trivializing, if you observe me for three days in a row you may see that I'm breathing well and jump to the conclusion that I will keep on living forever. But alas, we'll all kick the bucket sooner or later, as a longer data collection will show.
    
2. I hear advice on "optimizing" campains e.g. the exact spot your ad will appear on the right hand side (google gives you the option to be first, second, last and so on). Even if these things do matter marginally, you can't unlock the secret, since problem #1 above still holds and worse, your ad is always judged among the "competitor" ads that appear next to it, something that you have no control over. We are getting into game theory territory here. There is no "best" tuning. If there was everybody would copy it and it would lose its edge. Think of satellite navigation systems that direct the traffic away from bottlenecks; if everyone followed the revised route it would become jammed too!

My advise would be to stop wasting time freaking out on google adwords optimization. You don't have any data to judge, just random noise. If you have to use adwords, then set up something reasonable and check on it once every couple of weeks. There are better things to do in life like improve the core program or drink beer from girly glasses. If the software is good and the landing page well made, it will sell.

Later on tonight we'll learn what happened to the epsilon award too.

Post a comment on this topic