If you’re more than 20 years old you will remember the annual delivery of your local telephone directory; the thud as it lands heavily in your hallway, often with a crumpled cover, a testament to the efforts of the delivery boy to fit the tightly published pages through your letter box. For me, this directory was my first exposure to search engine rankings, with its wonderfully named 001 Aardvark Taxis vying angrily with 001 Ace taxis for first place in the listings.
Pigot’s directories, published in 1814, were the first widely available English reference source for everything from tradesman to public houses. After Bell’s invention of the telephone in 1876, directories became common, first to the secretaries of the nobility and then to the middle classes.
Coming First
The 001 Taxi problem is as alive today as it was in phone directories of my youth. Today it manifests itself in the ‘war’ of search engine rankings – a continuous cut and thrust between businesses seeking to steal a march on their competitors whilst the search engine developers try to outwit the tricks of web site developers. For some businesses, failing to appear on the first page of Google’s search results means anonymity and business failure; paying for adwords is expensive.
I became acutely aware of this problem in our recent development of Social Care information and advice portals. These portals, legal cornerstone of the Government’s 2014 Care Act, require local authorities to establish and maintain an information and advice service in their area. At OCC, we have sought to develop a smart ranking system that works on keyword relevance (i.e. the prominence of the word in the service profile). However, service providers may list keywords several times in their description to optimise their profile which we don’t want. We wish to rank services by relevance so if the keyword (or thesaurus equivalent) is found in the description once it should get the same relevance points as finding it 10 times. Google’s famous Page Rank algorithm is not available to us in the Social Care space, so we need to find another way to help our users find the highest quality information for their search.
To display our results we have introduced a secondary search ranking facet based on service profile quality score. The idea is that we establish a scoring mechanism for service profiles similar to the relevance score; points are awarded for logo, accreditation, quality of description, images, contact details, etc.
However, this still leaves a very serious problem. Serious for service users, providers and Local Authorities, namely that if you type in the same keywords you will get exactly the same search results in the same order every time. This is a reality – most people will use the same words to search for services: words like ‘Home Care’ or ‘Meals on Wheels’. So the same providers will always come first in your list; the same providers last. This is bad for creating a diverse and competitive market and bad for citizen choice. The reality is that well over 20 appropriate residential care providers may all have scored somewhere between 98% and 100% in our ranking algorithms. The small percentage differences are vital for the order of results but we surmise that they are trivial in terms of the actual suitability of the service for the citizen.
Randomization
One solution – randomization. We could take results that have scored almost the same (and we can define what ‘almost’ means in percentage terms) and then randomize the results. All those providers that have scored ‘almost the same’ are presented in a different order each time you search – the actual order being determined by a randomization algorithm.
This simple technique helps to create a diverse market; shares business across a range of suitable providers and helps ensure a vibrant and competitive market. There is no incentive for Providers to give themselves silly names; no more 001 Aardvark Care Homes. However, we’re worried about the impact on usability – how many search engines do you use where the results come out different each time for the same search? If you send your search link to a friend and tell them to look at result #2, they will see a different service in that position.
If we’re going to use randomization to help create a vibrant Social Care market we’re going to have to make sure everybody understands what we’re doing. Thoughts on a spare page of an old Yellow Pages please.