Noticed a bug in the Yahoo iOS Stocks App. When viewing the detailed metrics for individual stocks, the field for average volume is showing wildly incorrect values. Based on the examples included below, there doesn’t appear to be a pattern or obvious root cause to the problem (i.e. simply missing the M for million) as the numbers are all over the place.
In an earlier post, I wrote about the +1 component of a stock in Google Finance. Recently, I noticed that stocks in Yahoo Finance had a similar social component: a module that captured the count of Facebook Likes. So I was curious as to how the counts of +1 in Google Finance compared against the counts of Facebook Likes in Yahoo Finance for a given stock.
In this post, I compare the number of +1s of a stock in Google Finance versus the number of Facebook Likes for a stock in Yahoo Finance for the top twenty US companies as ranked by market capitalization.
Here’s what the +1 module in Google Finance and the Facebook Like module in Yahoo Finance look for an example company, such as Apple (AAPL). (Note: red arrows added by me)
Here is the data for the top 20 companies:
The first initial reaction is that, in general, the number of Facebook Likes is higher than the number of Google +1s. One big outlier is Google, and as explained in the earlier post, this is because users who use Google Finance are more likely to be employees of Google or even fans of Google products and thus in both cases more likely to be fans of the stock.
So what was the point of all this? I think that these numbers are a function of two parameters: (1) How effective the respective finance site is at drawing user traffic, and (2) How popular the Google +1 and Facebook Like features are for users who use such finance sites.
I’ve noticed a lot of low quality results when looking at the news article feed for stocks in the Yahoo Finance iPhone Stocks App. It seems that for a given stock, for example eBay, the app is suggesting any article that even has a mention of eBay.
Recently, I’ve been noticing a lot of articles that are relevant for Yahoo Inc., but not eBay. The reason is fairly obvious: the current CEO of Yahoo Inc., used to be an employee of eBay Inc. As such, many online articles about Yahoo point out the fact that Scott Thompson used to work for eBay.
That being said, the App can be a little smarter in deciding which articles to show. For example, it can analyze the click through rate of a given article that shows up both for eBay and Yahoo — and most likely for these types of articles, users viewing the Yahoo ticker symbol are going to be more likely to click through on the article as it’s more relevant to them. Once that occurs, the same articles should move down the search rankings when they are in the eBay stock ticker news view.
One subtle, yet important, aspect of search is how a search engine handles null search results. This is the use case when the search engine cannot find anything relevant to the user’s search query. The ability to handle this use case is important for traditional search engines (Google, Bing, Yahoo), vertical search (i.e. Yahoo News or Google News), or even an e-commerce site such as eBay or Amazon.
I ran a search query in Yahoo News that led to zero results being returned. So I tried the exact same query in Google News and saw plenty of results.
So what’s going on here? I know people prefer Google search to Yahoo search and algorithmically, one could make the case that it has a smarter search “brain”. But for one site to have plenty of results and for another to not have any results is probably not a deficiency in the search “brain” but rather the product choices for the Yahoo news search engine.
Perhaps the Yahoo news search is looking for an exact match of all three search terms consecutively in an article i.e. “mega millions tragedy”. Perhaps Google news search is more loose and simply searches for articles that contain at least one instance of all three words, or even more loosely at least one instance of one of the search terms.
The point is, in the case of null search results, the product is better served trying to find a partial match for the user instead of admitting defeat and returning nothing.