Had an awesome experience zipping down mountains at Neemrana, Rajasthan with Flying Fox
If a picture is worth a thousand words, videos should be worth millions, so I shall spare you any further crap. Just watch and enjoy!
I saw this Complan TVC where they show results of an experiment, where the set of kids having Complan grew by 6 cm as compared to 3 cm for the group that did not have Complan. And then comes the claim:
"Kids who drink Complan grow two times as compared to those who don't"
Picture this. A brand manager would have got this flash of thought one fine morning and decided to conduct an experiment to see how much taller kids having Complan grow. Why would someone want to do that in the first place? I mean ... kids grow, thats it. Some grow an inch or two taller than others. Period. Its ridiculous enough to spend (more time than money) on conducting an experiment like this.
And then, you come up with a finding which reveals a startling 3 cm difference! Inference: Kids having Complan grow twice as tall. According to this conclusion, if two kids, 4 feet tall, take part in the experiment; the one having Complan would grow to 6 feet while the other would grow to 5 feet?It would have been alright if they would have said that kids having Complan grow 3 cm taller than those who don't. But twice taller, based on a difference of just over an inch? Positioning on vital-amines would have been better than on these vital stats I say. Wouldn't parents want their kids to be well nourished as compared to doubly as tall as someone else :)
Complan and Horlicks have been involved in public bickering through their ads of late. I wonder if this "trying to prove one upmanship" takes a brand anywhere in the long run. I think it ends up in brands losing their own platforms, forget about establishing a connect with the consumer, isnt which the ultimate aim of campaigns?
I vaguely remember my Market Research book telling me to establish clear objectives, followed by hypotheses formulation; then collect and analyze data to test the hypotheses and draw inferences. I guess the Complan brand manager either took his MR class long back or just didnt know to read flowcharts. Maybe this would help :)
Long time since I blogged ...
Came across this site http://pricebook.in by eBay.in Motors. The homepage has the following claim:
"Predicting used car and bike pricing made easy Based on actual sales data of over 2,00,000 vehicles"
Its easy to be impressed and get carried away by such a huge sample used for estimation. I am sure the model has been built based on various parameters like brand, age, region, usage to name a few to arrive at the price estimation.
The caveat that I see is not so much in the model as it is in the context it is set in. Most reading this would agree that used car market in India has been highly unorganized and fragmented (as agreed by PriceBook as well here. Another characteristic has been that it has been a seller's market. Due to lack of standardization and information asymmetry, prices at which used vehicles have been sold (and bought) have not been fair to say the least. Now, when you base your model on transactions where probably the most crucial parameter (the price) has been biased with information asymmetry, vested interests (of the dealers, brokers ... remember they usually get a percentage of the deal) among other things, how fair would the estimation of it be? The counter view to that could be the fact that there is such a large sample, it would iron out the irregularities. My view to that is ... if there has been so much irregularity all through, the model is more likely to absorb it in than smoothen its effect.
I got thinking on this concept more because I recently also came across a service at http://vericar.in started by a friend. It automatically triggered a comparison. What veriCAR does is tries to inspect a vehicle's current condition and provide a report along with an estimate of the price (I am sure considering a lot of parameters similar to the ones used by PriceBook). What veriCAR wouldn't do (or I hope so) is take the historical prices as a critical parameter in their estimation. Its a model which takes expertise as input and would learn statistically over time. To me, that is superior to a model that takes biased statistical input (albeit as large as 2,00,000 transactions) and tries to claim expertise. I can hear all those fans of statistics and QM shouting NO NO already!
What the comparison also triggered is an analogy to technical analysis v/s fundamental analysis in predicting stock price behaviour. I am neither too knowledgeable nor very keen about capital markets, but I am sure the analogy makes some sense. What underlies technical analysis, I think, is the efficient market hypothesis, the validity of which I shall not discuss here (frankly coz I don't know too much about it :P).
In case of the used vehicle market, when the inefficiences of pricing in transactions are so well known and accepted, what makes the PriceBook model robust? It might be too immature to comment with lack of information about the model, but what the heck ... its my blog :P
The threat to using estimations from such a model is that they would tend to drive future sales at prices which are seemingly fair. Fair value perception of buyers would rise to the levels set by sellers for so many years. Ain't it a dream for the seller!!
P.S.: If you think I am promoting veriCAR, I am not gonna try to convince you otherwise!
© Blogger templates The Professional Template by Ourblogtemplates.com 2008
Back to TOP
