Surprise! The latest release of the Google Search Appliance introduces a much needed – and deeply valuable - capacity for tuning search results.
The background
When Google initially released the Google Search Appliance and its little brother the Mini), they trumpeted its ease of use. No tuning required – simply drop it into your network, identify the content to be indexed and voila, your users have Google-quality access to corporate content.
In official blogs and FAQs, Google proclaimed that the Search Appliance “leverages the work of thousands of engineers to get the correct answer, right out of the box, with no ‘tuning’ required.” This, they say, allows companies to avoid the “tornado of search results chaos” that “tuning search in the enterprise can . . . set off.”
It’s a feature - honest
Analysts were not as enthusiastic as the good folks at Google. Tony Byrne, publisher of the Enterprise Search Report, noted Google’s expertise in redefining a bug as a feature when concluding:
Google’s assertion that their algorithms’ inherent superiority obviates the need to expose their black box is patently wrong…. It’s impossible to know a priori how well one system will work with your collections until you test it — and then tune it.
In a similar vein, Forrester Research named a lack of “control over query and indexing processes” as a key reason for not naming the Google GSA as a market leader in its 2006 Forrester Wave: Enterprise Search Platforms. The Gartner Group chimes in noting in their 2006 Magic Quadrant that the GSA “lacks the sophisticated tuning and customization capabilities” of competing vendors.
In short: The Google Search Appliance could not be tuned for specific enterprise environments – and analysts saw this as a barrier to its wide-spread adoption in the enterprise.
Our experience supports the analyst view. In two recent deployments, we found that the results returned by GSA’s algorithm did not match our clients’ view of the importance of the documents returned.
“Source Biasing”
So we were delighted by Google’s January 4 announcement that the latest GSA release incorporates “source biasing.” As this screen capture shows, source biasing allows administrators to instruct the GSA to increase or decrease the weight (they call it “strength”) of documents of specific types or found in specific locations.
(We’ve made a full size screen shot available as well.)
This allows Google partners to solve a new range of problems. For example:
- The GSA can be “biased” so that all PDF documents are given more weight than Microsoft Word documents (on the assumption that PDFs are final, approved texts and Word docs are working copies).
- You might have two customer data systems, one with legacy information and one with more recent data. The GSA can be instructed to decrease the weight assigned to the data in the legacy system. The legacy data would still be found when very specific searches are used – say “sales to customer x in 1978” – but more current information would be prioritized for simple searches such as “customer x address.”
- The Prime Minister (or President) might appoint a new Minister. This can lead to a sudden (and urgent) request that all press releases and speeches before the date of appointment be “de-prioritized” in favour of documents mentioning the new Minister. Assuming press releases are categorized in chronological directories, the GSA source biasing option now makes this straight-forward. Which is fortunate, as the new, new Minister will be making the same request in eight to sixteen months time. (FYI – yes, we know that Ministers are actually appointed by Her Majesty through her representative the Governor General of Canada on the advice of the Prime Minister but we did not want to completely baffle non-Canadian readers of this blog.)
By any other name…
We really don’t care if Google calls it tuning or biasing or configuring or weighting or a flexible relevance engine - we are just happy they’ve made the change.