With Rotten Tomatoes’ 20th anniversary rapidly approaching, I’ve decided to chronicle some old tales from when we were fresh, young tomatoes full of vigor and energy. My last blog entry covered the early, whirlwind days leading up to our first (and only) round of investment and the decision to switch from our successful and growing web design firm, Design Reactor, over to doing Rotten Tomatoes as a new venture in April 2001.
We were riding high from the rush of closing our investment and finally getting to work full-time on a project that we believed in. The new round of investment meant that we could finally begin building for the future — placing some investements into long-term tech that would allow us to build Rotten Tomatoes from a “hobby-site” in to a full-fledged company.
Building out the Nerdy, Technical Details
First, as I had previously mentioned, we had previously been letting our server bandwidth ride off of the dual T1 broadband lines installed into our office that Disney paid for as part of us developing and staging Disney Channel (and now ABC’s) web sites. Now that we shifted to Rotten Tomatoes’ full-time, we had to find alternate hosting. This wasn’t a small decision — at the time, all of the IDCs (internet data centers) were also finding plentiful business from all of the dotcoms sprouting in the valley. This was before all of the cloud services like Amazon Web Services where you pay on a “pay-per-usage” basis. We ultimately signed a five-year lease with AboveNet, one of the early data center pioneers in San Jose. Even with a “friendly” contract, it was still $1000 per month per Mbps (we were using 5Mbps at the time) AND a server rack leasing fee. It was kind of a shock to sign a contract that would essentially obligate us to pay half a million dollars over the next five years (or little less than half of our first and only round of investment).
It was also the start of my self-learning on the nitty, gritty internals of server installation and maintenance. Again, back in the days pre-cloud, running a web server meant that I had to drive down to Fremont to pick up the actual rackmount servers from the warehouse and bring them back to the office for installation. Another major expense — each server was around $4000 each and our very first rack installation ran off of four servers. Over time, we were adding an additional one or two servers per quarter just to keep up with traffic ending up with about 24 servers by the time I had left. One of the major early technical changes was to shift the entire Rotten Tomatoes web site from running off of Microsoft IIS (Microsoft’s early web server software on Windows NT) over to running off of the now classic LAMP (Linux, Apache, Mysql, PHP – later, switching out Apache for Nginx) stack. The Microsoft stack was sufficient in Rotten Tomatoes’ early days when we were only covering a few movies opening every week with standard HTML templates and server-side includes, but we knew after investment that we’d have to run a professional, database-driven web site that could cover much more. This was all my crash course on how to set up and maintain Linux servers.
The rest of the tech team that I managed worked hard on programming, particularly with George Nguyen handling a lot of the PHP development on our content mangement platform, Eric Yeh doing a lot of database and performance optimization, Pongky Nataatmaja doing front-end PHP and template coding. Also, special thanks to Kai Chang who also did additional PHP coding and a special shout out to Suzanne Wood — she started out as our office manager during Design Reactor, but expressed sincere interest in becoming a developer despite not having a technical background. She interned under Eric and became a full-fledged engineer in our team, leaving the team to pursue a master and PhD in AI and cognitive science (Patrick’s field of study).
Rotten Tomatoes up until this point had been running essentially as a static web site with new web pages added for each movie released. We wanted to be a useful resource for all movies, not just new movies, so we set out to fill the database of movies from the past (catalogue movies, as we called them). Our purpose wasn’t to compete with our major competitor, IMDb, in terms of comprehensiveness. Instead, we aimed to have the basic information available so that film critics reviews for prior movies could be more easily aggregated under the same film title and so that users could find the basic information about the movie to decide whether they wanted to watch it or not. At the time, there were basically only two licensees of movie information with comprehensive data, both set up primarily for retail vendors like Tower Records to offer DVDs: All Movies Guide and Muze. Selecting a data vendor is a pretty momentous decision — all of the information such as cast, crew, genre, etc. pretty much determines the way that you structure your database and how users will be able to navigate your movie information. We ultimate signed a three-year agreement with Muze for another $5,500/month, another nearly $200K contract signed over in our early days. Thanks a ton to George and Kai who immediately set about to integrate the Muze database in to our content management system which we called “Webfarm”. It expanded our coverage of movies from about 1,000 movies pre-Muze to over 80,000 titles (most of them direct-to-DVD releases) as well as a easier way for our editors to use a web interface to edit or add movie information in to our own movies database.
Lastly, we knew that we had to be able to “scale-up” our coverage of movie reviews, particularly as we had just licensed a full 80,000+ title database. Early on, we were primarily bottlenecked on our editorial side where Senh Duong, Binh Ngo, Susan Nakasora, and others were spending sleepless nights selecting quotations and placing ratings on hundreds of movie reviews per movies. We addressed the problem temporarily in the summer of 2000 by hiring four summer interns from nearby Cal to help on the quote selection during the summer movie season, but the workload was too high. At the time, Rotten Tomatoes had a competitor web site the Movie Review Query Engine (MRQE) run by a single engineer in Pennsylvania, Stewart Clamen. Stewart had written a web crawler that crawled hundreds of different web sites and grab links, ratings (when available), and author information for all of the web sites’ movie reviews. MRQE was a night-time hobby for Stewart in addition to his day job as a programmer, and at the time he had no plans to make an independent business out of his hobby site. MRQE had just the data we needed in order to en-masse pair hundreds of thousands of links to movie reviews to all of the titles provided by Muze. In these early days of the internet, there was still a lot of good will between “hobby sites” such as MRQE and Rotten Tomatoes so Stewart gave us the astoundingly good price of $1000/month to license his data, whose comprehensiveness was pretty vital to our early growth. I think he just felt comfortable with Senh, Patrick, and I as “good guy” movie fans trying to build something useful and partially because neither side (MRQE nor Rotten Tomatoes) was 100% comfortable about the legality at the time of aggregating links and data from the open web like we were doing so were more willing to spread the risk around. In any case, our MRQE and Muze licensing relationships lasted through the full six years that I remained with Rotten Tomatoes and were fundamental to being able to turn Rotten Tomatoes from a hobby static page web site in to a useful, professional, and comprehensive resource for movies.
With the licensing of data from Muze and MRQE along with the construction of our “Webfarm” content management system (CMS), we were able to create a full flow for our editorial that covered the full history and catalogue of movies. Muze covered basically any title that was available on DVD (almost the full history of movies). MRQE was indexed movies by their IMDb ID so as we got movie reviews in from MRQE, we would have to match up the movie’s IMDb ID within the MRQE feed over to the appropriate Muze ID in addition to matching up the film critic and publication to our internal ID system. For movies in the MRQE datafeed that didn’t match up to a Muze ID, we would have to either manually search Muze and IMDb for the appropriate film titles and create an alias or we would have to create a whole new custom movie entry in our database (the case for many independent or film festival titles that didn’t have Muze entries yet) and eventually consolidate them when Muze eventually did create an entry. This was basically how we were able to rapidly populate up to 20,000+ films with Tomatometer reviews over the course of a few months using the small handful of editors that we had started out with.
As time went on, we used the increase in editorial productivity to tackle not just films already released, but to cover films while they were just gettting announced or in production. As aforementioned, our Webfarm system enabled us to scan Variety and The Hollywood Reporter on a regular basis and add newly announced productions in to our system. This was an important element to our traffic growth — by being amongst the first web sites to actually have a web page created for these newly announced titles, the other movie web sites such as IMDb or search engines would begin linking to us eventually leading to our film page showing up amongst the top search results when searching for the movie.
Now with the coverage of not just newly released movies, but also catalogue movies and recently announced movies, we achieved our goal of becoming a fully comprehensive resource for all movies.With our comprehensive movie data, more and more sites began reaching out to us as a movie resource within their own services. Initially, we were happy when other, bigger companies began freely linking to us: Google Desktop in 2002, an early desktop search widget, began providing Rotten Tomatoes data whenever you searched for movies. Our neighbor in Emeryville, Ask Jeeves also began freely displaying the Tomatometer in their movie search results. The added demand for our movies and Tomatometer data led me to establish our data licensing business where we exported basic movie data, Tomatometer ratings, and review ratings and links. For a period, we had Netflix, Microsoft (via their WebTV product), and Adobe/Macromedia, and others all paying a monthly licensing fee to use and display Rotten Tomatoes on their respective services. We also had both Variety and The Hollywood Reporter licensing our Tomatometer data to display in their respective publications and web sites, a way to get the Tomatometer better accepted in the film industry as a standard. Of course, later other notable services such as Apple iTunes also began using the Tomatometer within their services.
Here’s a fun story: During the tough times around 2002 when I just got started on data licensing, we were approached to build a “white label” Rotten Tomatoes web site for the world’s largest network of adult web sites at the time, controlling a network of over 300 popular porn web sites. I never actually met my contact face-to-face and it appeared that the small number of web site operators were all spread over the country working remotely. Nonetheless, their request was simple and clean — they were willing to pay us to create a version of Rotten Tomatoes with no branding (neither Rotten Tomaotes’ nor their own) but presented the Tomatometer and links for new movies just like Rotten Tomatoes that they could host on their own servers. Their rationale was that they felt that their porn web sites had relatively short session times and they wanted to explore more forms of content that would interest users in staying on their sites for longer even if they were devoid of porn or advertising. Seeing as how there wouldn’t be any objectionable content or branding on these white label pages, we eventually agreed to their terms. Despite being in a questionable industry, they were a near-perfect data licensing customer and comprised about 30% of our data licensing revenues over several years and paying their fees like clockwork every month. They even would all us and renew their licensing contract for the same terms a full 60 days BEFORE the expiration of their contracts. What a great customer!
The early days of Rotten Tomatoes immediately after our investment were spent building up the site from a hobby in to a full fledged business. We applied to tech to not just keep our team relatively small, but also to make Rotten Tomatoes a comprehensive resource for all movies and, eventually, leveraging that tech and data into its own revenue stream via data licensing. The data licensing had a side benefit of getting the Tomatometer more broadly distributed across other resources such as industry publications, search engines, and movie destinations like Netflix and iTunes and played an important role in making Rotten Tomatoes the industry-accepted standard it is today.
Footnote: Yes, there was a brief time that Netflix presented the Tomatometer and critic review quotations when you ordered movies. I was an early Netflix fan before the time of streaming downloads and back when they delivered DVDs in the mail via the “red envelopes”. In 2002, just as we were getting our data licensing program ramped up, Netflix contacted us out-of-the-blue and asked whether the Tomatometer to Netflix’s movie pages. From 2002 until 2005 (or 2006?), Netflix presented a fresh or rotten tomato, Tomatometer rating, and few select Cream of the Crop critic quotations and links within their movie pages via our data licensing. It was great to have Netflix as an early paying customer of our data licensing program. I found it to be really helpful when selecting movies, but I believe around 2005-06, Netflix wanted to tilt more heavily into their content recommendation system (even running the “Netflix Prize” to improve their recommendation algorithms) and subsequently removed the Tomatometer from their pages.
For more of my Rotten Tomatoes 20th anniversary articles, check out my other memories.