Browsing a tale of 5 cities

You can learn a lot about a city by the kind of activ­ity it dis­plays. In our case, we were in­ter­ested in when a city wakes up – vir­tu­ally.

Using data from a lead­ing in­ter­net ser­vice pro­vider, we looked at the time at which people log on to the Internet.


That’s the av­er­age brows­ing be­ha­vi­our in 2011 across India. The dark­er blue in­dic­ate more people brows­ing on­line at that time. Most people seem to be brows­ing at 10pm (re­mem­ber: this is primar­ily a do­mest­ic ISP), and there doesn’t seem to be a huge dif­fer­ence between the days, ex­cept that people wake up slightly later and sleep slightly later on Sundays.

However, between cit­ies, there is a con­sid­er­able vari­ation in this pat­tern. Let’s com­pare Bangalore and Mumbai, for ex­ample.


The red areas in­dic­ate times when Bangalore-ans browse more, and blue areas in­dic­ate times when Mumbai-ites browse more. Folks from Bangalore are re­l­at­ively early risers, start­ing as early as 4am on most days, and re­tain the lead un­til around 10am, when the Mumbai us­age catches up. Mumbai leads in the af­ter­noon, while Bangalore re­cap­tures the lead a bit af­ter 6pm. Most Bangalore-ans are asleep by 11pm, though, and Mumbai-ites zoom past, cap­tur­ing the lion’s share of brows­ing between 11pm and 4am at night.

Also, it ap­pears that Mumbai-ites work a lot less than Bangalore-ans on Sundays – opt­ing to go on­line only af­ter 9pm at night.

On the whole, “work-hard, play-hard” might cap­ture the spir­it of Mumbai-ites, while “early-to-rise” seems to define Bangalore. (From per­son­al ex­per­i­ence, we find that a bit hard to di­gest, but the data is ir­re­fut­able.)


A com­par­is­on with New Delhi shows a some­what sim­il­ar pro­file through the hours of the day, but the week­day be­ha­vi­our is sim­il­ar. So Bangalore does rise earli­er and sleep earli­er than New Delhi, but works about the same on week­days and week­ends.


In con­trast, Bangalore and Chennai seem to have quite a sim­il­ar brows­ing pro­file through the day. However, on week­ends, Chennai seems to browse a lot less than Bangalore, thus qual­i­fy­ing as “early-to-rise”, and also for “relax-on-weekends”. (From per­son­al ex­per­i­ence, we find that quite be­liev­able.)

If we take Chennai and com­pare that with Mumbai or Delhi…


… the pro­files are quite sim­il­ar. As be­fore, both cit­ies wake up and sleep late. Mumbai is a bit more act­ive on week­ends than Delhi.

Just based on this data, if we sum­mar­ise our ex­per­i­ence, this is what it ap­pears to be:

City Behaviour Rises early Rises late
Works on week­ends Bangalore Mumbai
Takes week­ends off Chennai New Delhi

The visu­al­isa­tion we used above is a heat-grid. It’s ana­log­ous to scat­ter plots, but for dis­crete data in­stead of con­tinu­ous data. Some of the situ­ations where we’ve used heat-grids in­clude:

  • Plotting call volume in a call centre by date and hour
  • Plotting num­ber of com­plaints by cus­tom­er group and reas­on
  • Plotting train­ing course ex­pense by type of course and in­struct­or
  • Plotting prof­it­ab­il­ity by geo­graphy and branch type

… and so on.

If you’re in­ter­ested in cre­at­ing heat-grids your­self, you will find it fairly easy to do on Excel 2007 and bey­ond. Just fill in the val­ues you want in, se­lect them, and choose Conditional Formatting – Color Scales.


Update: We got re­quests to add a few cit­ies – Hyderabad in par­tic­u­lar. I’m afraid that this ISP does not have a strong pres­ence in Hyderabad, so data is sparse. But here’s Pune:


As you can see, Pune seems to be a late riser. Probably not too dif­fer­ent from Mumbai…


… and oth­er than check­ing mails a bit earli­er than Mumbai-ites on Sunday night, it’s pretty much the same pat­tern!

Leave a Reply