Yahoo and Prediction
October 5, 2010
Yahoo’s public relations machine is working hard to deal with the flood of news about executive turnover and the questions about Yahoo’s management leadership. I wanted to snag this “What Can Search Predict?” item before it becomes unfindable. Google Instant and Bing are wonder services but pinning down specific documents in the brave new world is getting more difficult in my opinion.
The point of the write up is that user behavior at a point in time provides information about what’s hot and what’s not. I understand this. Analyzing usage data is not a new thing, nor is the math used to clump clicks and plot them, massage them, and extrapolate from them. Most college grads had a chance to try their hand at this type of math in classes from psychology to biology and statistics. (I can hear the groans now.)
Yahoo says:
In many cases, we found that these traditional predictions performed on par with those generated from search. Although search data are indeed predictive of future outcomes, alternative information sources often perform equally well.
The idea is that big data are good but specific, narrow sets of data from specific corpuses may deliver better indicators of user future actions.
Makes sense to me. Big data are big. More precisely constrained data are narrower. When looking for a specific indicator, why not consider the constrained data? Makes sense to me, but I would prefer a method that uses big data, when available, and more constrained data. Two sets of outputs can be examined.
Yahoo adds:
The potential for search-based predictions seems greatest for applications like financial analysis where even a minimal performance edge can be valuable, or for situations in which it is cumbersome or expensive to collect and parse data from traditional sources. Ultimately, search can be useful in predicting real-world events, not because it is better than other traditional data, but because it is fast, convenient, and offers insight into a wide range of topics.
Several questions waddled across my mind:
- What is the current Yahoo use case for its insight? I know that each time I return to my Yahoo Mail, the system does not remember me, nor does it present options to me based on my behavior or a larger group’s in my view. I have to click, click, click to see a list of email. Maybe Yahoo can provide some concrete examples?
- In the midst of the shift to Bing search, where does this predictive stuff fit. I was looking for a “mens black watch” on Yahoo Shopping. Try the query. I am not sure what can be done to improve the results, but search results mixed ranges with specific prices on specific models. Huh? With user data – either big or constrained – predictive methods should reduce confusion, not create a “huh” moment for me.
- Is this a “level” problem? Here’s what I am thinking. The problem in search that Yahoo is addressing seems to be down in the weeds. There are larger findability problems with Yahoo’s system. For example, in the shopping example a user must click on a “more” link in order to access the shopping search feature. Most users don’t know to what that “More” refers. Is this a contributing factor to user frustration which in turn may explain some of the loss of polish on the purple Yahoo Y?
Worth reading and then finding a use case (which I may be missing) before recycling information already in the channel in my opinion.
Stephen E Arnold, October 5, 2010
Freebie
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
