Bangalore | Fremont, CA
+91-80-67422100 | +1-254-244-4300 |

A superior classification algorithm from my security guard

I am among the millions of worker bees that have made their homes in gargantuan hive like structures stylishly called “gated communities”. Through these arrangements, we hope to keep the big bad world at bay and the hope is that within the four walls will emerge an idyll that will rival the Promised Land. Safety and Security is paramount in the minds of the gated community denizens and therefore a massive investment has gone into the security infrastructure. On a day that my city observed a Bharat Bandh, there were actually barricades, sand bags and flag marches by our private army of security contractors!

Security contractors
Credit: NASA/GSFC/Mark T. Hubbard

The only spoke in the wheel to this very protected existence is the need for another type of army, ie the hired help. An average household in my complex has about 5-6 maid servants doing a variety of jobs. At the lowest end of the spectrum are the clothes and dish washers. Then come the babysitters, next the caretakers of the oldies and on top of the food chain are the cooks. There are also some households with a Chief of Staff maid who oversees other maids and follows the mistress around looking very official and very busy. On the male side, there are numerous drivers, iron-men, dog walkers, cable wallahs and the like. The entry and exit of all hired help is strictly controlled by the security staff. The mornings and evenings at the gate present a scene very similar to the Security Check manned by the Transportation Security Administration at the JFK terminal in New York City. I have seen very intrusive pat downs that would make Dick Cheney and George W Bush very proud indeed.

However, this practice presents a unique problem to the security guards. Unlike the TSA agents who groped SRK and Abdul Kalaam without batting an eyelid, these guys have to fish out the hired help, such that we residents are not subjected to this molestation. You, obviously don’t bite the hand that feeds you, Right? Ordinarily, this shouldn’t pose a problem. Just look for the dishevelled lot, who mouth unfathomable dialects and look like they haven’t bathed for a few months. Easy! But, alas.. not so in our apartment. A full 75% of the hired help arrives impeccably dressed, with the flourish and swagger of a Jack Nicholson. The ladies are quite fashionable and  very pleasant on the eye. On the other hand, a sizeable proportion of the residents do look like they are from a refugee camp. Especially in the mornings, as they would turn up to leave their children onto school buses. I am ashamed to say that my appearance almost always is straight out of Les Miserables.

However, in this impossible situation, I have noticed that the awesome guards have NEVER EVER made an error. Their accuracy is even better than the dabbawallahs of Mumbai in classifying who is a resident and who is the help. I should know because I have never ever been stopped and searched. I have marvelled as to who they are able to do this. The complex has about 550 flats and upwards of 2000 residents and about a similar number of servants. There is also a huge turnover among the guards and fair amount of churn among the residents. So it is impossible to do this classification by hand- that is, identifying by face recognition of the individual. They seem to have developed a superior classification algorithm that so elegantly separates the chaff from the grain. When I paid some more attention, I noticed that they looked for very, very subtle cues like the language, interaction with kids, the level of laid-back demeanour, laughter and the like to identify residents. Since new guards rotate in all the time, their superiors must have taught them the rules of the game. Such precision, efficiency and the ability to train so quickly is awesome.

In so many of the segmentation projects that I have seen, there is always a struggle to classify beyond a certain level of granularity. This is because the information elements that are available as classifiers are not able to discriminate the underlying behaviour very well. Therefore the segments become too amorphous to discern anything of significance. There is a very valuable lesson from these security guards that needs to be integrated into development of data driven classification algorithms in order to be able to construct wafer thin and precise microsegments. The ability to go beyond the obvious demographic, psychographic, usage and value based paradigm to explore more subtle forms of behaviour that will be better able to classify outcomes. This  is the key to unlocking value from segmentation.

The author, Karthikeyan Damodaran, is a Consulting Services Delivery Leader at BRIDGEi2i – A company on a mission to unleash the power of analytics and transform the lives of enterprises and individuals alike. We believe that the solutions to almost all intractable problems lies buried inside the data. BRIDGEi2i has the ability and experience to mine a wealth of unstructured and structured information to help businesses identify prospects, target them through the right channel, maximize cross sell and up-sell opportunities and thereby enhance the life time value of customer relationships.

Connect with us:

facebook  BRIDGEi2i on twitter  BRIDGEi2i on LinkedIn  BRIDGEi2i on Google+  BRIDGEi2i on YouTube

The views and opinions expressed in this article are those of the author and do not necessarily reflect the official position or viewpoint of BRIDGEi2i.