Alex Schultz: APIs

Two nations seperated by a common language in search

“Two nations divided by a common language” is a quote often attributed to George Bernard Shaw and I think it’s a pretty good one. In general going global for websites is pretty easy and I have been lucky enough to get lots of traffic for my websites from all the english language speakers around the world. That being said even with english language sites there are super interesting differences. I have spent my time on cocktailmaking.co.uk optimizing for the word cocktails and phrases like cocktail recipes (which I am lucky enough to own in the UK) and “vodka cocktail recipes” etc… Globally this doesn’t look sensible:

Zooming in on UK only data a different result comes out:

The US data completes this picture:

In general I strongly believe that users from different countries are more similar that different. They all want great service, the web is better for all of them when experienced with their friends and a big red button will get anyone to click :). That being said you have to be careful about the details of how you approach things. When british people are looking for a cocktail they search for “cocktail recipes”, “vodka cocktail recipes”, etc… 5x more than “drink recipes” When Americans are doing the same they search for “drink recipes” 3x more than “cocktail recipes”. My british bias made me optimize for the wrong words on total volume.

Crowdsourcing rudeness ratings in cocktailmaking.co.uk

One big issue I had last year is that I was thrown out of adsense for my cocktail site because of rude content. It turns out a lot of cocktails have rude names. As far as adsense is concerned “Sex On The Beach” is ok but some cocktail names are really not ok. That makes sense and I can understand advertisers not wanting to appear next to rude content. The issue for me is I have 10k user submitted cocktails and I am 5k behind in terms of reviewing and rating them.

In order to solve this I defined all new cocktails as “rude” (rudenessid == 3 in my database) and then added a section at the top of each new cocktail asking users to review them. Then once a week I send myself a summary of the “unacceptable” and “rude” cocktails via email for me to go in and clean them up.

The below gives an idea of how much data and how many reports are flooding in to help me fix this :)

Users are super helpful and I have found overall that the crowd comes up with great accuracy each week. Especially when one of my facebook connected users review the content I can really trust user feedback on this front. I am still tuning the system but suddenly users are reviewing more cocktails each week than I have added AND the reviews concentrate in on the cocktails that are getting the most traffic. This really helps me keep the site clean in a crowd sourced manner. My friend Javi has an expression “work to not work” and this is a great example (in my opinion) of that!