Have been chastisized by Tarle for pouring ATF on burning embers in a previous post and not using my praja time on more useful things :).
Quite a few posts on how we should design routes for BMTC buses (regardless of what language the board is in :)). For that we need to know travel patterns on a typical working day (assuming that we want to primarily help people trying to reach work). We should not use ticket sales as an estimate of demand becase demand in that case is biased - biased by whatever route is already there.
I am NOT a market research expert. But this would broadly be the method.
a. Statistics is our friend here. Given the large numbers using buses there will be patterns. So we use statistical methods to do this.
b. We cannot do a census. We will have to use sampling. http://en.wikipedia.org/wiki/Sample_size Based on this we can go for a sample size of 1000 and still get very good of accuracy. If cost is a problem even n=500 gives extremely good levels of accuracy.
c. We need to decide if we sample only users or non-users as well. If we assume that by better routing we will be able to attract non-users we need to bring them into the mix. Maybe 4-5% of the total sample. Or assume that an area is not covered currently - we should definitely include it. A market research person can shed more light.
d. We have to ensure a good representation of all residential areas. Let's assume there are 100 pincodes/200 wards. We have to systematically sample 500 persons from each. Is he/she a user. If yes - where does she most often go to in a bus. How many times a month? Which are the other destinations. Again - market research help is needed
e. Once we have this data, we can find out ratio of users/non-users for that pincode. If it is 1.5/1 we can say that 40% of that pincode/ward are bus users. With BBMP ward data,we can get an estimate of total bus users from that ward.
f. Based on the destination answer, we can get an estimate of proportion of users going to a destination. Have a vague feeling we need to weight by the frequency of usage in a month. Using this we would know what percent of people go to which destinations.
g. Using (e) and (f) we can get an idea of travel patterns. There will be nuances like time of day to be captured, etc. But this in a nutshell would be how I would do it.
Srivathsa
ಈ Blog entry ಅನ್ನು ಸ್ನೇಹಿತರಿಗೆ ಕಳಿಸಿ
ಪ್ರತಿಕ್ರಿಯೆಗಳು
Poll here first?
Can we start the poll here with a few questions as a project and use some viral mktg to get the sample?
thanks
thanks sri. this was in the making for a long time. i like ids' idea and we should keep that as a "deliverable", while we wade through the question.
also khader saar has been raising an important flag about how some regions which are not serviced. so one of the biases to look for might be the bias due to leaving out those unserviced. these are different from those who dont desire the service. khader saar should kindly provide inputs on this.
Start with the end in mind...
Srivathsa,
I have some expertise that might be relevant having run a market research organization for a decade and a half. Will be happy to provide my $0.02 worth of insights into the mechanics of the research process if that would be useful.
One of the first few things we did in any market research project was to mock up the final results and then ask the user(s) of the research to pretend they have all the data they were looking for and then ask how they will use the information. This step ensures that the research objectives are crisp and there is alignment among all stakeholders. While a brief description of this step appears deceptively simple and straightforward, the reality is quite different . This step is the most difficult part of a study.
I have not done any research in the past 5 years (and so my skills are somewhat rusty) and have not done anything at all in India and so am not sure what tools may be available here today for selecting the sample, the instrument for collecting data (paper, phone, internet) etc.
Like I said before, will be happy to assist in the process in any way you think meaningful. We can probably talk some more at the next meeting.
VVR
SY, Thanks for the post!
Yajamanre,
Thanks to you for this important post. Understanding the need is the one of the vital input to any design and implementation. I like your idea of covering all the wards in BBMP jurisdication.
Syed
TS, DULT report confirms the lopsisded connectivity!
Tarle,
The lopsided connectivity is all evident in B'lore city. I am sure this must be true for other cities in India. The reason being, routes and schedules are decided by personal preferences of the people who are in authority. Added to that we as public also doesn't care about the future needs and planning.
My apprehension has been proved by the DULT report itself.
http://www.deccanherald.com/content/18299/bangalores-bus-connectivity-lopsided.html
I am not sure whether this report has details of the areas that are not covered adequately. A fair design would have been that, affluent areas like Koramangala, Jayanagara, Sadashivnagara and the recent ones would have less BMTC services. And areas like J C Nagar, Tannery Road, DJ Halli, Cox Town, Cooks Town, Ganganagar, Sanjay Nagar, Mathikere, areas surrounding BEL, HMT would have more frequency and schedules. but reality is opposite.
In the V2.0 PT system, lets aspire to correct these anomalies.
VVR - end should be more direct connectivity
VVR,
Thanks for the offer for help. What sort of instrument would you use? What questions would you ask?
The end in mind is to ensure that we devise bus routes based on unbiased travel data. So if I know that e.g there are 1000 people taking buses out of say Jayanagar 9th block between 8:00 a.m. and 10;00 a.m. 250 of them go to ITPL and 200 go to E-City. Currently all this load shows up as an increased demand on either routes 18 or 20. So we might end up increasing frequencies on 18 and 20. And end up crowding KBS and SBS even more.
If we knew that people took 18 only to change to 335E then we might as well introduce some new route (say 335J) to take people directly where they need to and bypass KBS.
If we knew this for all Bangalore we might be able to do a better job of providing end to end connectivity for more passengers.
Srivathsa
Drive safe. It is not just the car maker which can recall its product.
Let us start poll
Let us start a Poll asking :
You travel from:
1. South Bangalore - North Bangalore
2. South Bangalore - East Bangalore
3. South Bangalore - West Bangalore
4. South Bangalore - Central Bangalore
1. Central Bangalore - North Bangalore
2. Central Bangalore - East Bangalore
3. Central Bangalore - West Bangalore
4. Central Bangalore - South Bangalore
and so on....
It gives us a rough idea..
vvr
with the data, i would like to be able to create a map that shows trip sources and sinks - for atleast 3 caegories of trips - short haul, medium haul and long haul.
then analyses to identify the profile of passengers: if a person is likely to make more than one trip per day between two points, or atleast one trip between a pair of locations in a couple of days, would he/she take public transport. if yes why? if not why?
these are two important things for me.
POLL
My thoughts on this
1) What is the category of people that the bus routes would cater to ?
This would detemine what type of survey ( online v/s offline, english v/s kannada, kind of questions in the form etc)
If it is a comprehensive survey for all kinds of bus routing - need offline forms in kannada with layman language
2) What data needs to be captured - Travel patterns for all people
For each category of people, you would need a different set of questions which are tuned for them.
for example : let us say categories are working people - (office goers, non office goers), students, retired people, house wives
For example : For working population who go to offices
a) Where do you stay ?
b) Where do you work ?
c) how do you commute to work
d) what are the travel timings to work place
e) do you travel other than the daily trips on work related activities ( in between travel during the day)
f) questions to estimate this travel pattern
Other than these require questions on estimating non work related travel ( this may be common across categories)
a) where do you shop ? how frequently do you shop ? timings of shopping etc
===========================
Another way is to have a house hold survey ( may be more efficient) - collect data about the entire house hold - ( retired, working, housewives, students)
we will have to segment this on socio economic classes - different types of households
===========================
Sampling : to get data about all wards in bangalore, a sample will have to be taken in each ward. We will have to do geographic segmentation - then collect random samples in each segment, collate them and then extrapolate the results.
===========================
( these thoughts are rather disconnected, but main point is that we will need a massive effort in preparing the right type of survey before launching it)
I hope I am not making this
I hope I am not making this too complicated. I just want to make sure we know what the tradeoffs are.
The research question we are contemplating, i.e. quantifying the universe of potential bus travellers and its various demographic and other user need segments to a fair degree of accuracy may not lend itself to a primary research process with a sampled survey being the central vehicle. I am not saying it cannot be done but the cost/effort is huge given the demographics (as alluded to by Thampan). We will need lots of boots on the ground for this effort and the research logistics could be complicated.
Typically, to get quantitative characteristics of an industry (e.g. total market size, market shares, demographic segments etc.) we would go to secondary sources such as trade bodies (e.g. Chamber of commerce, Semicon Associations, etc.) or secondary research companies (e.g. Frost & Sullivan) that track certain select industries and make their money selling their research. So that would lead us on a long and winding road to BMTC's door in this case. Presumably this is the reason why Srivathsa has proposed our own survey. We may still be better off mining what we are specifically looking for from any traffic data that BMTC may be able to provide us. Is this an option at all?
A sampled survey is ideal for qualitative analyses i.e. extrapolating population-wide projections of behaviors (user needs, responses to certain stimulii etc.) from a selected sample.
Alternatively, I would like to suggest that we break down the problem regionally and look at the pieces of the puzzle one at a time. For instance we could study the traffic patterns originating from a heavy population center and suggest remedies. Syed has already mentioned some high-density lower-income (target rich for bus users) areas. We could perhaps start with one of these.
I suspect that the best way for us would be to use a combination of data mining and very specifically targeted survey research.
Start with low hanging fruits!
VVR
Alternatively, I would like to suggest that we break down the problem regionally and look at the pieces of the puzzle one at a time.
Lets get consensus on problem statement and its breakdown.
We may still be better off mining what we are specifically looking for from any traffic data that BMTC may be able to provide us.
I am being optimist to state that it is possible. Here are some low hanging ones.
I do have a copy of the simple one page survey we drafted for the Volvo Survey. Let me know if you would like to start with that.
Syed
VVR - traffic data from BMTC has some basic problems
VVR,
I agree that this is a mammoth task. BMTC will have to hire a proper agency to do this. I don't think for a moment that we can do this via a survey on our site.
Here is my basic problem with BMTC data. Let us say that I see 10000 tickets being sold on route 36 from Srinagar to KBS. Current capacity is only 9000 (say). BMTC might simply be tempted to add 10% more trips on 36. Net result is that more buses than necessary might be flowing into KBS and that place is an absolute mess with logjam being the default state.
Point is that tickets sold on 36 cannot tell us the ultimate destination of that passenger. How do we correlate ticket sales across routes? If a person changes from 36 to 252 or to 78 or to 315 how does one capture that?
Many of these routes were fixed by BMTC more than 25 years back. Bangalore has changed significantly since then. We know that NM is going to make its appearance in 2012 or 2013. Should the routing not be reworked in the nex context?
Srivathsa
Drive safe. It is not just the car maker which can recall its product.
SY, this is good information
A survey is a good tool for us to project some characteristics about the entire population. For instance, let us say that the entire population we are studying is community of BMTC users starting their trip at Srinagar. Knowing that 9000 (or however many) people per day took an outbound trip is a huge step forward. This is the number that I thought will be difficult to obtain from our studies.
Now given this information, we can extrapolate the following info (I am just making up or rather mocking up some numbers here) through a survey:
5400 people continued their trip beyond KBS (60%)
3240 people are very or somewhat dissatisfied with their onward connection at KBS (36% of total sample)
68% of those that are dissatissfied with their onward connection at KBS are headed for the North-East quadrant of Bangalore
etc. etc.
Of course this assumes a lot of things. We are able to sample the population well (i.e. we get a good cross-section across age groups, time of day, professions etc. etc.) and we are able to ask unambiguous, unbiased questions in the proper order (i.e. follow all the rules of a good survey design).
Looking into the future, we can bring in "potential users" (e.g. teo-wheeler and four-wheeler users who might consider a switch to public transportation with NM) into the sample. We can also design in questions that consider the NM impact.
Data that Ashwin Mahesh has
I liked the way Ashwin Mahesh's project collects data. Mentioning it here since many may not know.
He has cellphone density data from towers of cellphone service providers (airtel if I recall right). I am guessing this is how it would be working
So now, by comparing cellphone density data at various points of time, you can get a sense of population movement patterns.
I am not sure whether he gathers data on movement pattern of each cellphone, or if it was just density data from base stations. If each cellphone in the city can be tracked for a couple of days (which I guess Ashwin might have done already), you would have pretty good data points on population movement patterns in the city.
But yeah, you would have to make a guess on how many of these were moving on private transport, and how many otherwise. Wonder if movement data is detailed enough to spot trends like
Either way - this approach would be a goldmine of market research data for BMTC. Decent sample set - everyone who works and commutes carries a cellphone.
SB, can you ask for details!
SB,
Could you pls ask AM for more details on data being collected using cell phones?
If my understanding is correct this is what we are looking for:
If AM has such data then we all set for extrapolation and other analysis
Easier done at big interchanges
Spoke to my wife who is a market research person.
She suggested that we survey people at KBS, SBS and other important interchanges (Richmond Circle?, Corporation?, Silkboard?) to get a quicker survey done.
People at these places are there most likely because they want to change routes.
Non users of course will not be covered. For them some of the big office hubs (E-City, ITPL) basement car parks is a good place to start.
Srivathsa
Drive safe. It is not just the car maker which can recall its product.
Diverse market makes things difficult
Hi All,
Whilst most recent, current data & future forecast/s are an absolute must for any sort of planning, the realities in India make accurate data collection a formidable task with downright minor telephone & internet penetration & poor levels of literacy. Consequently, planning has been abysmal & out of sync with latest needs & trends.
An example: Metro phase-1 routes were planned based on echos of aged surveys with minor modifications based on a more recent small-scale study. Click here for a previous post about old study reports & explains how the routes had been decided. Nearly the same route proposals had been put forward each time by various study committees without much updating with current or prevailing patterns. If the data had been updated, the N-S route would have linked Tumkur road with Hosur road. The Jayanagar routing was clearly based on age old surveys.
It is obvious that all along, data have been weak, except for CTTP-2007, which is the only study of recent vintage.
I strongly doubt if modern market research methods (by telephone, internet or physically with a small sample) will actually be efficient in determining all the diverse needs since transport is very basic unlike consumer goods.
It is for this reason that I had suggested a much broader & more inclusive research to determine bus routes on this post.
Captain, Agree this is the way to go in future!
Captain,
It is for this reason that I had suggested a much broader & more inclusive research to determine bus routes on this post.
I agree with you that moving forward, the urban transport authorities in cities should start collecting this kind of data for the planning and augmentation. This method is being used in many places. While I was in Los Angeles, my company on yearly basis had to submit this data to the city of El Segundo, CA. Every employee was required to submit the information.
But this would take long time. It needs the framework to layout the logistics of collecting, authority in place to collect it and most importantly making use of it. Also in places where, companies doesn't pay govt dues for years I am not sure how would they comply with this requirement. But I feel this is the way to get quality data on daily commuters.
Meanwhile, it is no harm if data can be collected using conventional methods by surveying population at different strategic locations.
Sample One Page survey questions.
BMTC Team at Praja had designed this simple one page survey for use at places like ITPL etc. Even though done for promoting Volvo service, but questions are formed to collect basic data. Please feel free to comment, critique and suggest additional/better questions