Facebookâs trillion-edge, Hadoop-based and open source graph-processing engine â Tech News and Analysis
Geege Schuman stashed this in FB
And yet, despite all of this technology, the ads are still quite mediocre.
A little less technology - A little more common sense
We've gone a little technology crazy - the fact that we can now process a trillion node table in n dimensions in a reasonable time without breaking the bank gives data science departments the mandate to do just that. I recently heard tell of a data scientist that spent 2 weeks analyzing smart-grid data to conclude that certain customers turn their lights on and off throughout the day and others turn the lights on in the morning and off in the evening. Ten days to figure out that there are business customers and domestic customers. Hmmmm.
I've been followed by certain retargeting programs as I surf the web, and it's visible and annoying. Especially when the recommendation engine has not noticed that I have already bought what they are advertising. This is when data scientists become too smart for their own good.
What about a low tech solution? Whenever I want, I tell Facebook about somewhere I am going or something I am looking for. It then weights the ad's in favor of those things. If I click on them, the weighting goes up. More machine learning and less predictive analytics.
Or even a medium tech solution? Anytime I book travel with my credit card, the card knows where and when I am going. If the credit card company is able to provide that data to the ad' networks with my permission, then in the weeks leading up to my trip, I see ads relevant to the trip. When I return, they stop. Some good can happen as we join up the data, as long it I get a say in it.
I think we need a low-tech approach to data science before we throw technology at the problem.