Alexa Web Search Platform

» see websearch.alexa.com as it lived, on the wayback machine

~ the eulogy ~

In December 2005 Amazon put something strange on sale. The Internet Archive's first capture of websearch.alexa.com is dated 13 December that year, which is about as precise a birth certificate as this kind of product gets. Alexa Internet had been crawling the web since 1996, and the archive of that crawl was the seed the Wayback Machine was built from. The Web Search Platform let an outside developer upload code, run it against that stored copy of the web on Alexa's own servers, and walk away with whatever index came out. The code ran on Alexa's machines because it had to, next to the data, since moving a copy of the web anywhere else was not a practical option in 2005.

Nobody else was offering that. The search APIs on the market, Google's and Yahoo's among them, were free but capped, and every one of them answered your question with its own ranking already applied. Alexa's deal was the opposite shape: no cap and no opinion about relevance, with a meter running the whole time.

This is the Alexa that never reached me, and the one that did is a separate grave on this site. In Turkish forums through the 2000s, on donanimhaber and frmtr, Alexa meant exactly one thing: the rank. People put their site's Alexa number in their signature the way they put their processor and their RAM, and arguments about whether the toolbar's sample was representative went on for years. That service, alexa.com and its green bar, limped on until 2022. The thing you could rent by the hour died thirteen years earlier and almost nobody in that argument knew it had ever existed.

~ the dossier ~

~ the price list ~

The pricing was published as four flat numbers, all of them a dollar: one dollar per CPU hour of processing, one dollar per gigabyte of storage, one dollar per gigabyte uploaded, and one dollar for every fifty gigabytes of data processed. It reads like a parody of simplicity and it was the clearest statement of the product's problem. A student with an idea about search could not run a serious experiment without a budget conversation, and a company with a budget already had engineers and a crawler of its own.

~ what the crawl was for ~

Alexa's crawl at the time was described as more than a hundred terabytes, covering 4.5 billion pages across sixteen million sites. In 2005 that number was its own argument. The web was still small enough that one company's copy of it could plausibly be called complete, and large enough that no individual could ever assemble such a thing alone.

What Amazon had actually built was a rental counter for the position Google occupied, and the reason it failed is more interesting than the reason most of these things fail. The platform did not lose to a competitor. It lost because wanting to build a search engine and being able to afford to build one turned out to be almost entirely different populations of people.

~ sixty days ~

The notice went out to developers on 26 November 2008 and gave them until 26 January 2009. New customers had already been turned away before the announcement, so the thing was closed to the outside world before it was declared dead. Amazon's reason, in full, was that use of the service was low and that it was shifting priorities to other areas where it could provide better service to AWS customers. TechCrunch covered the closure under a headline noting that practically no one was affected.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »