Organizing the Unorganized ­ Employing IT to Empower the Under-privileged
Arun Kumar, Nitendra Rajput, Sheetal Agarwal, Dipanjan Chakrabor ty, Amit Anil Nanavati
IBM India Research Laboratory 4, Block C, Vasant Kunj, Institutional Area, New Delhi - 110070, INDIA.

{kkarun, rnitendra, sheetaga, cdipanjan, namit}@in.ibm.com ABSTRACT
Various sectors in developing countries are typically dominated by the presence of a large numb er of small and microbusinesses that op erate in an informal, unorganized manner. Many of these are single p erson run micro-businesses and cannot afford to buy and maintain their own IT infrastructure. For others, easy availability of cheap lab our provides a convenient alternative even though it results in inefficiency, as little or no records are maintained, and only manual, pap er-based processes are followed. This results in high resp onse times for customers, no formal accountability and higher charges. For the businesses this translates to lower earnings and losses due to inefficiencies. In this pap er, we look at few such micro-business segments and explore their current models of op eration, while identifying existing inefficiencies and pain p oints. We build up on the findings and prop ose an approach for delivering b enefits of IT solutions to such micro-business segments. Finally, we present technology that realizes the prop osed approach in the sp ecific context of two such segments. the pyramid [14], and cannot afford their own IT infrastructure. For instance, in India alone p eople in rural areas account for ab out 70% of the p opulation which amounts to ab out 700 million [5]. Most large cities also have a sizable p opulation consisting of p eople who have migrated from villages to earn a living. Such p eople are typically illiterate or semi-literate and earn around Rs. 60000/- p er annum (approx USD 1500) or less. We conducted a survey of 47 sub jects in our target p opulation across 10 cities and small towns in India to find out ab out their ICT awareness. The professions of the subjects varied from p eople in service industry such as house maids, milkmen, drivers, to very small businessmen such as shop-owners, vegetable and fruit vendors, to skilled microbusinessmen such as plumb ers, carp enters, electricians, handloom weavers, to employed p ersonnel such as security guards, cab drivers, etc. We asked them questions regarding the channels they used for obtaining information, the use of ICT in their professional and p ersonal life, interaction with their clients, use of mobile and their financial status, lifestyle etc. Only 23% of the p eople surveyed try to advertise themselves through some means such as p ersonal contacts, government agencies, phone, or through a contractor. Only 40% p eople owned a phone out of which only 4% had a landline and rest had a mobile. 17% p eople used phone to communicate with their clients, while 4% used pap er based communication mechanism, 2% communicated through government established channels, 2% communicated face to face, and 2% had no direct communication with their clients. Rest 73% of the sub jects did not feel the need to communicate with their clients. For some professions it was understandable such as for auto-rickshaws which pick up passengers when hailed on the street but interestingly many shopkeep ers also did not feel the need to have a channel of communication with their customers. 55% of the sub jects surveyed had TV at home. However, only 32% used it to gather information. Ab out 21% used newspap ers, 2% used radio and 36% used word-of-mouth as their source of information. Rest did not feel the need for any formal channel to gather information. In terms of their financial status, only 21% had a bank account in the family and only 4% accepted cheques (in addition to cash) as a mode of payment. None of them accepted credit cards or any other kind of plastic money. Only a little ab ove 8% make a reservation b efore travelling by train. The authors' own exp eriences, the survey results, and studies conducted by others [3, 6] indicate very little use of ICT by this user segment. This tends to make them very

Categories and Subject Descriptors
H.5.2 [Information Systems]: Information Interfaces and Presentation User Interfaces[Voice I/O, User-centered design]

General Terms
Human Factors, Design

Keywords
Developing regions, Voice applications, Virtual communities

1.

INTRODUCTION

Different service sectors and market segments in developing countries like India are plagued with inefficiencies, low productivity and wastage of resources. A ma jor factor contributing to this is the dominant presence of informal, unorganized businessmen who rely on manual processes and cheap human lab our rather than automation through the use of Information and Communication Technologies (ICT). This is understandable since a significant p ercentage of p opulation in developing countries lies near or at the b ottom of
Copyright is held by the International World Wide Web Conference Committee (IW3C2). Distribution of these papers is limited to classroom use, and personal use by others. WWW 2008, April 21­25, 2008, Beijing, China. ACM 978-1-60558-085-2/08/04.

935


informal in their business or profession and results in lot of disorganization. As a result, there are failed commitments, little or no accountability, and inefficiencies in markets leading to sub optimal b ehaviour [6]. This also enables many of these informal businesses to evade taxes since no formal accounting records are maintained [3]. However, with globalization and proliferation of new technologies these informal businesses are facing difficulties in keeping pace with new comp etition and increasing demands of their customers. Introduction of basic ICT such as mobiles, PC, Internet have help ed improve various sectors such as the fishing industry [6]. There are other efforts such as aAqua [15] that go b eyond the basic ICT plumbing, to offer useful services to this target p opulation. Paper Contribution: There are three key contributions of this pap er. The first is to identify the pain-p oints of the mobile micro-businesses, from two angles. The first viewp oint is to look at micro-business from a service-provider's angle and the second is to seek the pain-p oints of customers of these micro-businesses. For this pap er, we restrict mobile micro-business to household services that are provided by plumb ers, carp enters, electricians, painters kind of skilled lab ourers. We also include auto-rickshaw service in our definition of micro-business. We use user interviews for this purp ose. The second contribution of this pap er is to prop ose organization of these micro-businesses as a solution approach the use of virtual online communities as a sp ecific solution to the pain-p oints for this p opulation segment. Finally, we validate the applicability and usability of the prop osed solution through user-studies. Paper Organization: The rest of the pap er is organized as follows. In Section 2, we study two sp ecific segments and insp ect the op erational models prevalent there. We summarise the pain p oints of these two segments of mobile micro-business. In Section 3 we prop ose a solution approach to overcome drawbacks of current models by prop osing the concept of organizing micro-businesses into a virtual community. We also present detailed results of the user surveys with resp ect to the prop osed community based solution. In Section 4, we present technology solutions for creating a virtual online community for the two micro-business segments. We also p erform a user study to validate the usability of a prototyp e implementation and rep ort the user evaluation results. Finally, we discuss the applicability of our approach in other segments in Section 5 and present some related work in Section 6 b efore concluding the pap er.

2.1.1 Surveys
We surveyed 26 skilled lab ourers in the Delhi region. The aim of this survey was to identify the pain p oints in their current op erational model. We asked them the following questions and the results are summarized in table b elow. · How much time in a week are you free while available for work? · Are you willing to go to far off places due to lack of work in the neighb ourhood? · Are multiple conflicting requests that might result in loss of work, a problem? Table 1: Summary of responses from skilled labourers Questions Yes No 20%-60% Idle time? 75% 25%(< 20% idle) Willing to travel to get work? 43% 57% Conflicting requests a problem? 0% 100% As is evident from the surveys, most often these lab ourers are idle while available for work. For some p eople, this idle time goes as high as 60%, i.e. they get work only on 12 days in a month. This was one of the most common pain p oints of this community. Secondly, since not everyb ody gets work in the nearby area, sometimes they have to accept assignments that are at far off places. Even though almost all travel outside their neighb ourhood at times, 57% p eople do not find it covenient. The travel costs (sometimes paid by customer), problems in commuting through unreliable public transp ort, and fatigue due to cycling to far off distances are a deterrent. Occasionally, a lab ourer may get multiple requests for the same time and day. In such cases, most lab ourers are able to convince the client for an alternate day and/or time. If a reschedule is not p ossible, the lab ourers refer the work to their colleagues. Overall, a conflicting request is not a pain p oint with this segment. Even though there exists such a large workforce, clients often have trouble contacting them or finding information ab out their services and their contact information. We interviewed 25 customers to understand how they currently engage the service of such micro-businesses and asked them the following questions with results summarized b elow. · Do you face any difficulty in finding skilled lab ourers? · Do you feel the need for a service to reach out to skilled lab ourers? · Are you willing to pay a subscription for such a service? Table 2: Summary of responses Questions Difficulty in finding lab ourers? Need a service? Willing to pay subscription? from customers Yes No 44% 56% 91.4% 8.6% 57% 43%

2.

APPLICATION SCENARIOS

In this section, we look at two sp ecific unorganized sectors in India that can significantly b enefit by leveraging IT.

2.1 Mobile Micro-businesses
Many p eople in the lower end of the pyramid esp ecially those who migrate to cities from rural villages, pick up a small trade skill to earn a living. These include plumbing, gardening, driving, welding, carp entry, repairing of bicycles and motored vehicles among others. The size of such unorganized workforce in India is ab out 370 million p eople [16]. This forms 93% of the total workforce in India. A significant fraction of this workforce is semi-literate, i.e. they have less than 5 years of elementary education. Skilled lab ourers such as plumb ers, electricians, carp enters are a part of this workforce.

44% of the p eople said that finding these p eople is their biggest hurdle. 91.4% of the sub jects expressed a desire/interest in a service that could connect them to the local micro businesses. They want it to b e a one stop shop for all their needs. If it provides guarantee of service 57.1% were willing to pay a subscription fee.

936


2.1.2 Current Operational Models
Different op erational models exist today, each with its own drawbacks. Many such micro-businessmen work on their own by visiting houses in a community to provide services. Their business dep ends on the information that the community has ab out them. As the numb er of p eople who know them increases, their clientele increases. The primary mode of advertising for these p eople is word-of-mouth references from their existing customers or their colleagues in other trades. Yellow pages do not work in countries like India for multiple reasons. These p eople cannot afford the cost of such advertisements which are typically not up dated frequently. Even p eople who have a small shop only occasionally advertise themselves through fliers that are cheap er to print and are distributed along with daily newspap er. For customers, it is tedious to search through printed yellow pages and online web based yellow page services such as JustDial 1 are only now b eginning to emerge. However, the most imp ortant pain-p oint for the customers that emerged from these interviews was the lack of accountability and quality of service guarantees that exists in these unorganized sectors. Customers feel more confident to use services of a lab ourer referred through a known p erson rather than using an unknown one from yellow pages. Security concerns also come into play and a referred p erson instills more trust than a completely unknown one. Printed yellow pages do not provide any means to rely on the advertised lab ourer. Customers rely instead on their social network of neighb ours and hired lab ourers such as security guards, maid servants etc. to search for other micro-businessesmen. In most cases, they contact the local grocery store or utility store to contact these p eople. The utility stores charge a small fee for sending them the desired service men. But the choices are limited and customers usually stick to the same p erson even if he does a substandard job simply b ecause the alternatives are not always available. A few organized models also exist. In one such model, these skilled lab ourers get themselves employed at some establishment such as a local electrical goods store (in case of an electrician) or a local builder who constructs residential houses on contract. This gives them credibility and a steady stream or work. However, they get sub jected to exploitation by the establishment and most often get underpaid. In another model, these workers get employed on an ad hoc basis by community housing organizations where their resp onsibility is to attend to repair needs of the housing complex. Here the monthly earnings are fixed, though this is usually lower than what they would earn in their own business. Such workers are contract employees and do not get health and insurance b enefits. Some brick and mortar concierge services also exist today but their op erating costs and overheads allow them to serve only high end customers.

or are rented by him and are op erated autonomously. Such autos ply throughout various cities and towns of India.

Figure 1: An Auto-rickshaw

2.2.1 Surveys
We surveyed 30 auto drivers in three Tier-1 and Tier-2 cities of India -- New Delhi, Bhopal and Jabalpur. The interview consisted of following questions ab out their work related pain p oints and findings are summarized in the table b elow. · How much time do you sp end on an average waiting for customers, p er day? · Do you own a mobile phone and use it for business? · What is your daily exp enditure? · Do you face any harassment? Table 3: Summary of responses from auto-drivers Questions Yes No High waiting time (7-8hrs)? 58% 42% Use mobile for business? 33% 67% High p enalties? 82.6% 17.4% Customer Harassment? 34% 66% From the interviews we found that most auto drivers operate 12-14 hours a day. A surprising revelation was the fact that out of this, the waiting time accounted for upto 78 hours a day. 58% of the auto-drivers rep orted their daily waiting time to vary from 7-8 hours while 42% rep orted it to vary b etween 4-8 hours. This is the time when they are forced to simply idle around at road sides waiting for passengers to come by. Very few auto drivers are engaged in scheduled pick and drop arrangements. The primary reason for this is the lack of a channel for communication needed to achieve this. Of the drivers surveyed, ab out 58% owned a mobile phone, yet only 33% used it for their clients. The others used their mobiles as a family phone and therefore did not carry them all the time, making it unavailable for business use. The drivers who have a mobile phone are able to accept prior app ointments during morning hours typically for office or school drops. However, after that their location is governed by the passengers that they get and they cannot commit further pickups. The daily exp enditure including rent for the autos is typically b etween Rs 200 (USD 4.5)­Rs 250 ( USD 5.5) p er day, fuel and maintenance is another Rs 100 (USD 2 approx) p er day. Additionally, they have to pay p enalties to

2.2 Public Transportation
The second unorganized micro-business segment that we look at, lies in the transp ortation sector. Auto-rickshaws or simply `autos', such as the one shown in Figure 1, are threewheeled vehicles that op erate like taxis and are a favourite mode of affordable public transp ortation in countries like India. However, they are either owned by the driver himself
1

http://www.justdial.com

937


the traffic p olicemen for road blocking2 and other offences. These p enalties account upto Rs 1500/- p er month (USD 32 approx). Their average daily earnings (after the ab ove exp enditure) vary b etween Rs 100 to Rs 150 p er day amounting to Rs 3000­Rs 5000 (approx USD 70 - USD 120) p er month. There were other characteristics that were revealed. Most of these auto drivers were willing to op erate anywhere within city limits but going out of the limits either required paying extra taxes or was simply not allowed. However, towards the end of day they prefer to only take passengers travelling in the direction of their home location. Some auto drivers also rep orted harassment by customers. Since most autos do not follow meters, many a times customers too go overb oard and demand to b e dropp ed for very low charges. At other times, customers demand to hire the auto even though it is not available for hire (such as parked in a stand waiting for the next shift driver or waiting daily checkup/repair). This is a negative side-effect of efforts by law enforcing authorities to control rude b ehaviour and unauthorised refusal by auto-drivers. We surveyed the customers as well and asked them the following questions: · How much time do you sp end to get an auto? Sp ecify Avg and Max. times. · Would you want to do advance b ookings? · Biggest pain p oints of hiring an auto ? We presented these questions to 33 auto-users and the result are summarized in the table b elow. Table 4: Summary of responses from auto-customers Questions Yes No High waiting time (10-30min)? 57.5% 42.5% Want advance b ookings? 94% 6% Harassment? 82% 18% Service Dissatisfaction? 18% 82% 57.5% of them mentioned that they wait on an average 10 min to a maximum of 30 mins 94% of the p eople surveyed resp onded in p ositive to the second question. Among the pain p oints, 82% p eople complained of having b een forced to pay more than the standard rates or drivers refusing to go the desired destination, and 18% professed their dissatisfaction with the imp oliteness and rash driving of the drivers.

traffic p olice as they are not authorized to park on road. All of these ultimately result in harassment to the drivers and the passengers. Transp ort authorities have tried to create designated prepaid auto stands at different places for the convenience of passengers. Here autos queue up for passengers and pay some nominal fees to the stand op erator and customers get readily available autos at charges fixed by the authorities plus a small premium. While customers are happy with the model as they do not need to negotiate and the autos are registered which adds to the safety factor, the auto-drivers complain of long waiting time and low fare charges for certain locations. For busy places like airp orts and railway station, auto-drivers are allowed to drop a passenger but either have to wait in a long queue to get next passenger or have to travel back to city empty. These inefficiencies result from the fact that no central entity exists that can coordinate the op erations of the autos much in the same way a taxi company op erates by taking in customer requests and dispatching orders from a central location. The difference b etween the two lies in the ownership of the vehicles. In the former, taxi vehicles b elonging to a single company or partners lie under single administrative control whereas in the latter case each auto-rickshaw is autonomous. It b ecomes apparent that organizing these auto op erations while retaining the autonomy of these autorickshaws would provide significant b enefits to b oth commuters are well auto-drivers and owners.

3. ORGANIZATION THROUGH COMMUNITIES
Driven by the inputs from the surveys, we prop ose overlaying or imp osing an organization over the unorganized sectors as a viable solution. Hop efully this can help in alleviating some of the inefficiencies and problems existing in current op erational models. Sp ecifically, we study the service providers in the two segments introduced ab ove and explore whether they are willing to b e organized with the goal of increasing productivity and reducing inefficiencies. Further we study the customer's view to explore the need and willingness from their p ersp ective.

3.1 Mobile Micro-businesses
We surveyed 26 skilled lab ourers in the Delhi region. The aim of this survey was the following: · to discuss with them and find out if organising them can help their business, and, · to verify if their needs can b e served by forming an organization for this section of the society. We provided the survey sub jects with an option of using a hyp othetical organized community service through which sub jects can get information ab out micro-business and can make app ointments.

2.2.2 Current Operational Models
Unlike taxis which have their designated stands, autos are hailed from the street and they rarely have designated parking space while waiting for customers3 . As can b e seen from the surveys, there are several inefficiencies in this mode of op eration. The long waiting times for the auto-drivers result in their urge to b e dishonest and to charge more than what is fixed by the transp ort authority. For similar reason, auto-drivers often refuse to go to a place from where they may not get more customers easily. Moreover, often while waiting for customers on roadside or while picking up change for a customer or on other occasions they get fined by the
2 There are very few designated auto-stands and most autos commonly wait for customers in densely p opulated areas such as markets and p opular joints. 3 In Kolkata, however, the opp osite situation applies.

3.1.1 Survey Results ­ Mobile Micro-businessmen
Of the 26 p eople, 12 owned a mobile phone. Most of them were plumb ers, electricians, painters, carp enters. Their current op erating style involves flocking around near a hardware market. None of them advertise themselves through local yellow pages or through newspap ers. They seem to have an impression that they are not a big company who can advertise.

938


Table 5: Mobile workforce survey responses.
Question Need organization? Community-based referrals? Ready to pay commission Use SMS Yes 85% 92% 53% 4% No 15% 8% 47% 96%

Table 6: Mobile workforce customer responses.
Question Need an organization? Ready to pay commission? Need service guarantee? Use SMS? Yes 82% 82% 92% 100% No 18% 18% 8% 0%

85% of the interviewed p eople felt that they need to organize their schedules. An overwhelming 92% said that they would want to register to a virtual community if it can help them in providing more work. 47% of them were not willing to pay a commission to this community. They wanted that the community should charge commission from the customers and not from them. However, the remaining 53% were fine with a commission of ab out 5% to 10%. None of them had problems with the fact that the community may call them back if the work is not good. However, some mentioned that the supp ort will b e provided only if it happ ens within a couple of weeks from completing the task. Almost all preferred to talk to the customer b efore signing up for work assignment on the virtual community. Of the 12 p eople who had mobile phones, only one p erson was comfortable using SMS as a means of communication with the virtual community. Rest preferred voice. Key Insights: This survey clearly identifies that this segment does not get enough work volumes ­ they are capable of hand ling more. It also establishes the need for having a virtual community for these labourers. Moreover, this community should levy a very low commission (if any) from micro-businessmen and should be available over a voice interaction.

this segment of people are literate enough, they are fine with either voice or SMS as a mode of interaction.

3.2 Auto-scheduling
We created a service description, called Dial-an-Auto to explain the scop e and limitation of the environment that the auto-drivers would b e in with an overlaid organization. We presented Dial-an-Auto as an organization that would provide an effective matchmaking of needy customers to autos nearby. This is not p ossible today due to lack of a common communication channel.

3.2.1 Survey Results ­ Auto Drivers
We surveyed 32 auto drivers in New Delhi with the aim of determining whether they would b e willing to sign up for Dial-an-Auto. We asked the following questions: · Would you b e willing to use Dial-an-Auto service? This would mean registering with the service provider, calling in to up date the current location whenever available for picking a customer or to find out if there was customer closer to them looking for an auto. · Would you b e willing to travel with charges fixed by the transp ort department while using this service? · Would you b e willing to accept advance b ookings? · How far would you b e willing to travel to fetch a waiting passenger without charging anything extra? Several drivers outrightly rejected the prop osition -- claiming it to b e imp ossible. Further probing revealed couple of interesting insights. First, they felt that they would get customer requests from far off places and it would b e a waste of time and money to pick them. We assured them that location based matching would solve this problem. Second, local rules enforce these auto-drivers to not refuse a passenger who has hailed them on the street. This was put in practice by the authorities to prevent errant drivers to refuse passengers that wanted to go to less profitable places or who wanted to pay as p er charges fixed by the government. The drivers asked how they could refuse someone on the street if they have already b een b ooked and are on their way to pick the passenger. We took this as an input to next refinement of sour design. However, on presenting clarification most of them could immediately sense the b enefits of such a service and many could relate it to the op eration of radio taxis that are controlled from a central location. We summarise some of the key resp onses. 100% of the surveyed auto-drivers wanted to use Dial-an-Auto. 53% of them were willing to charge as p er the rates fixed by transp ort authority while the remaining 47% expressed their desire to negotiate with the customer. 100% of the p eople surveyed were keen on accepting advance b ookings only in the mornings, since they cannot guarantee their location during day

3.1.2 Survey Results ­ Micro-business Customers
We interviewed 30 p eople who frequently use services from the micro-businesses such as plumb ers and carp enters. Ab out 75% of sub jects mention that they do not get a plumb er / electrician / carp enter when they need them. They are either late or unavailable. Reliability and overcharging is an issue with ab out 30% of the sub jects. Perhaps owing to these pain-p oints, 82% of the sub jects are willing to use an organization for their microbusiness needs. Of this, 68% (i.e. 82% of total surveyed) are willing to pay a nominal fees to use such a service. 70% of the sub jects would prefer to use just Phone/SMS to contact such an organization, while the remaining are comfortable with the Web too. 92% of the sub jects desire a service guarantee from the organization. 70% of the subscrib ers are willing to wait for a couple of hours b efore they get a confirmation for the service. For sp ecial long-term contractual services such as a maid or a driver, 15% sub jects are willing to wait for a couple of days b efore b eing confirmed. A sample of the questions asked to these sub jects are presented b elow: · Do you get Plumb ers/Electricians/Carp enters when you need them ? · Would you rather deal with an organization for your needs ? · Are you willing to pay a fee for this service ? · How would you prefer to contact the organization ? (web, phone, SMS) ? Key Insights: It is clear that the customers who need the services of such mobile micro-businesses are wil ling to use and pay for the services provided by an organization. Since

939


Table 7: Summary of responses from Questions Want Dial-an-Auto? Want to negotiate charges? Advance Bookings? Pick up to 1km without extra charges?

auto drivers. Yes No 100% 0% 47% 53% 100% 0% 25% 75%

system is deployed in the network and accessible through a phone numb er that is assigned to it. For the auto-drivers, we have prototyp ed an auto-scheduling service called Dial-AnAuto. Both the applications are accessible by voice over the telephony channel. Thus these two applications provide a low-cost and low-literacy-requirement solutions to organize the micro-business. Next we describ e the details of the two services.

time. Ab out 9.3% of auto-drivers surveyed were willing to go upto 0.5 km to fetch a customer without charging extra while 15.6% were willing to go upto 1 km without any extra charges. Remaining 75% were willing to go upto 5 km to pick up a customer as long the metering starts from the location where the call was received. A few auto-drivers expressed concerns if they will have to pay commission for using this service. They were happy to know that no such charge will b e levied to them. They were also concerned if this service would turn out similar to a pre-paid auto-stand describ ed earlier. Long waiting times and trips to areas with low likelihood of returning customers make these stands not very p opular among autodrivers. Another reason is that they lose their autonomy in fixing up the charges. Even in the context of Dial-an-Auto, the drivers wanted to have the option to accept/reject a passenger, and the freedom to negotiate the price directly with the passenger. Key Insights: This survey helped us realise that the solution should be such that it retains the autonomy of individual auto drivers. It should also cater to passengers who do not have a mobile phone. The solution should be able to provide a mechanism for booked autos to refuse passengers. Since many drivers are not wil ling to drive with government approved rates, the solution should provide a mechanism for al lowing negotiations.

4.1 Mobile Micro-business
The VoiAvatar system can b e used by micro-businessmen to create their own virtual avatars by making a phone call. The online avatars of these micro-businessmen can b e accessed to get information ab out them and their work. Such information could include area of op eration of the microbusinessman, service charges and work hours. Once many p eople create their avatars, this can evolve to a virtual online community of micro-businessmen and customers, where b oth can upload and access the information through a phone call. The customers could call in to community p ortal to sp ecify their requirements and the p ortal would then search for relevant micro-businessman registered with it. Alternatively, the customer could browse through avatars of various micro-businessmen to select one for his job. The avatar also acts a proxy for the micro-businessman as the customer can interact with it if the micro-businessman himself is busy with a job or is unable to accept a phone call. This ensures that the user does not lose customers due to unavailability through the phone. The VoiAvatar system has b een develop ed as a sp oken language interaction system. A pre-defined template is provided for a particular typ e of business. The sample template shown in Figure 2 is for skilled lab orers such as a plumb er or carp enter. In this template for example, some of the information asked includes home location, working hours, service charge information and recommendations for this lab orer. The lab orer can also configure his avatar so that it can take app ointments from his customers. The VoiAvatar system also allows multiple avatars to b e linked together through a Hyp ersp eech Transfer Protocol (HSTP) that we have defined and prototyp ed [7]. This hyp erlinking in voice enables creation of a networked ecosystem [10] of such avatars. Traversal of these links through a phone, involves transferring the phone call from one avatar to another along with the context of the application across the two avatars. Such a mechanism can enable customers to make online payments to a micro-business safely through a phone call. The businessman can later modify his avatar by calling the VoiAvatar system.
Calendaring Service Yellow Pages

3.2.2 Survey Results ­ Auto Passengers
We interviewed 30 sub jects who live in different regions in Delhi to find ab out their pain p oints with resp ect to getting an auto-rickshaw. Ab out 80% of the sub jects were willing to make a call to a Dial-an-Auto service to get an auto. Most sub jects are willing to pay an additional 10% to use this service. 68% of the sub jects are willing to do an advance b ooking for an auto. One sub ject had tried advance b ooking over phone with a know driver and he was satisfied with this mode of engaging an auto. Table 8: Responses from auto passengers. Questions Yes No Will you call Dial-An-Auto? 80% 20% Pay extra? 92% 8% Advance Bookings? 68% 32% Overal l, most subjects showed their wil lingness to use the Dial-an-Auto service and they were positive that such a service can be helpful for commuters.

Service Charges

VoiAvatar

Welcome Message

Appointments

4.

THE TECHNOLOGY

References

The insights gathered through the interview process (as describ ed in the previous section) establish a need for organizing the micro-business market through a community based system. As a solution, we prop ose the VoiAvatar system, based on VoiGen [9], that helps these lab ourers to create their avatars on the telephony network. The generated

Figure 2: Creating the Voice Avatar The system is not heavily dep endent on the sp eech recognition accuracy since very few utterances of the end-user

940


Voice Browser

Spoken Dialog System

Location Inferencing Module Matchmaking Engine

Ontology

Get Booking Mode
Passenger

Advance Booking?
N

Y

Record Date and Time

Database

Figure 3: Dial-An-Auto System are converted to text, most is recorded. Since the generated avatar of the micro-businessman is in his own voice, customers can easily relate the avatar to the actual p erson. This service brings some organization into daily op erations of these micro-businessmen. They b ecome emp owered with a means to advertise themselves, describ e their service and corresp onding charges and offer app ointments. Later on, appropriate information logging at the server could help produce usage data and rep orts for a micro-business. Such data collection and analysis is missing for this user segment.

Record Exact Address

Get User Location

Get Destination Details

Get Expected Charges

Register Request

Figure 4: Passenger Call Flow cation inferencing module which deduces the user's location by asking a series of questions. The location is deduced to the granularity level of a small area in the city. The exact location of the passenger is recorded later which is played to the auto-driver. The destination details are obtained in a similar fashion. Finally, the passenger is asked for an estimate of the fare he is willing to pay for the destination sp ecified. This question is optional and can b e removed for places where fixed fares are applicable in practice. In case of an advance b ooking, the user is also asked to record the date and time as to when he will b e needing the service.

4.2 Dial-An-Auto
In order to organize the auto-driver op erations, we prop ose a sp oken dialog matchmaking system that will enable auto-drivers to find the closest passenger and vice versa at any time. The fare prices and decision to accept a passenger can still rest with the auto driver thus retaining the autonomy of the drivers. In this section we will describ e the Dial-An-Auto system. Figure 3 shows the architecture of the Dial-An-Auto. The system is accessed through the telecom network by making a phone call. The call is connected to the sp oken dialog system via a Voice Browser. The sp oken dialog system contains two call flows - one each for passenger and auto driver, and use a location inferencing module and a matchmaking module to serve incoming requests. We next describ e these modules.

Get ID
Auto-Driver

New Member?
N Update Status (Location, Availability)

Y

Register

Matchmaking

4.2.1 Location Inferencing Module
Well structured network of streets and highways in develop ed countries make it very easy to sp ecify a precise location using location p ositioning systems. Applications like Google maps and Yahoo maps can b e used in location based services to give directions or to guide someone to a particular location. The same is not true for countries like India where the roadways are highly unstructured, milestones or street identifications are missing at cross roads and there are multiple names for the same location. Locations are sp ecified with the help of landmarks such as school, hospitals, market places as opp osed to just the names of streets. For the Dial-An-Auto system, we develop ed the prototyp e of a voice based location inferencing system which makes use of an ontological model to narrow down to the exact location of the user. The details of the ontological model used is out of the scop e of this pap er.
Play location
Y

Match Exists ?

N

Thank You and Wait

Figure 5: Auto Driver Call Flow Auto-Driver Cal l Flow: When an auto-driver calls the system for the first time, he is asked to register his p ersonal information: Working hours, home location, phone numb er etc. Some of this information is later used by the matchmaking engine to determine the b est match for the auto-driver. For example, it takes into account whether the auto driver prefers passengers heading towards his home at the end of the day or how far he is willing to go to pick a passenger free of charge. Once the registration is complete, a 4-digit PIN is given to the auto-driver which serves as his identification. The system then asks the auto driver for his current location. The matchmaking algorithm determines the b est match among all the registered passengers. In case a match exists, the passenger details along with the exact recorded location is played out to the driver. Otherwise the driver is asked to call back again after sometime. If the auto-driver has a cell phone, the system calls the auto driver as soon as a passenger b ecomes available. We have implemented a matchmaking engine [8] that not only does traditional parametric matchmaking but also takes

4.2.2 Spoken Dialog System
This module presents a different call flow for passengers and auto-drivers. It guides the user by registering his callinformation and request and transfers control to other modules to process the request. Figures 4 and 5 show separate call-flows that are designed for the Auto-driver and the passenger which can b e accessed on different telephone lines. Passenger Cal l Flow: Passengers can call the system to make two typ es of b ookings: current or in advance. In case of current b ooking, the control is the transferred to the lo-

941


into account real time up dates such as location information of the auto-driver. The parametric matching is done on the preferences of the auto-driver sp ecified at the time of registration.

4.3 User Feedback
Since the VoiAvatar system is intended to b e used by micro-businessmen, we invited some plumb ers, electricians, carp enters etc (subscribers) to use VoiAvatar. They created their own avatars. Secondly, we invited p eople who are typical users of such services and asked them to call the avatars that were generated by the subscribers. Thus we evaluated the entire process of generation of avatars (by subscribers) and then the use of these avatars (by users). In this section, we elab orate on the profile of the sub jects who took part in the evaluations. Later we describ e the evaluation process and conclude by providing the evaluation results and insights gained from user study.

For the second process, we explained the concept of providing services through telephony infrastructure to the prosp ective users of avatars. Then we asked them to make a phone call to the avatars that were generated as part of the first process. We asked the following questions to them after the phone call: · Was this voice based system (i.e. avatar) informative? · Do you normally have problems in getting touch with the plumb er, electricians or such service providers?

700 Time (in seconds) 600 500 400 300 200 100 1 2 3

"voigen_completion" "subject_introduction"

4.3.1 Profile of Survey Subjects
The system was evaluated on two processes. The first process focused on the usability asp ects of the VoiAvatar system. The target p opulation chosen for this task are the p eople who work as freelancers and have a sp ecific region of operation (a few kilometers). They are typically skilled lab orers (such as electricians, plumb ers, carp enters) and charge on the basis of the amount of work that is required of them. We surveyed 12 sub jects of which 3 were carp enters, 5 were plumb ers, 3 were electricians and one was a drilling p erson. The second process involves use of the avatars that are generated in the first process. These set of p eople, who we foresee would use these avatars, are from well-to-do families who need services to fix water taps, electricity problems, among other work items. This section of the p opulation is relatively more exp osed to IT services. Some of them have used interactive voice resp onse systems and know ab out or use the Internet.

4 5 6 7 8 9 10 11 12 Subscriber subjects

Figure 6: Time taken by each subscribers to understand the avatar concept and to build their avatars.

4.3.3 Survey Results
Out of the 12 sub jects that were used to generate their avatars, 10 were able to successfully create the avatars. Figure 6 shows the time that was sp ent to build the avatars for the 12 sub jects. Sub ject 1 and 7 were not able to create their avatars. As seen, most sub jects were able to generate their own avatars within 4 minutes. A 4 minute phone call in India costs less than 5 rup ees (10 cents). This is despite the fact that none of these sub jects knew or had used a voice based conversational system b efore. Surprisingly most of them were comfortable with using the VoiAvatar system. More imp ortantly, all of them were able to identify the p otential that having a avatar can have in increasing their business. However each such phone call had to b e preceded by an introduction to the concept which took ab out 5 minutes for each sub ject, as is seen in Figure 6. All showed tremendous interest in the concept of avatars and the fact that their work can b e advertised without them actually requiring to purchase any additional equipment. Ab out the usability of VoiAvatar, 2 sub jects felt that the system was not designed prop erly. They highlighted that it asks difficult questions rather than just asking for simple things such as name,occupation or area-of-work. We also p erformed usability surveys on 12 sub jects who were asked to make a call to the avatar that were generated by one of the carp enter. We asked them if the information provided on the avatar was useful to them. While most p eople found the information to b e useful, 3 sub jects mentioned that such a avatar will only b e useful if they are able to make an app ointment with the system. We also asked the sub jects if it was an advantage to have the entire avatar in the voice of the actual carp enter, rather than it b eing a system voice. To our surprise, 66% of the sub jects felt that they would prefer to hear to a standard voice that is clear and does not have a heavy accent. They mentioned that they were not

4.3.2 Survey Process
For the first process, we briefed the subscribers for ab out 10 minutes to motivate them of the use and advantages of avatars in their daily lives. Then we briefed them with the usage of the VoiAvatar system for ab out 5 minutes. Then we asked sub jects to make a telephone call to VoiAvatar. Sub jects were asked to interact with VoiAvatar and resp ond to the commands and provide the relevant answers. Finally, we asked the sub jects a set of questions to get an understanding of the usability of VoiAvatar and of the p otential of avatars for the masses. The questions asked were the following: · · · · Have you ever used an automated voice driven system? Was this method of generating the avatar easy to use? Are you interested in having your own avatar? Do you think your business will improve with the use of your avatar?

In addition, the following observations were made during their use of VoiAvatar: · Time required to describ e the concept of avatar. · Time sp end by the subscriber in creating the avatar using VoiAvatar.

942


able to understand the voice of the carp enter in some of the prompts. However the remaining 33% were impressed with the fact that they could hear the carp enter's voice and mentioned that the voice would help them judge the carp enter. All the sub jects used for this usability test were educated IT professionals. They were from different regions of India and not all of them had Hindi as their first language. Summary: Overall, the concept of providing services over telephony infrastructure was found to b e very imp ortant to the subscrib ers and they realized its p otential. The concept of avatar was also found to b e very easy to learn and use. Based on this survey study, we realize that since the access to avatars is through a telephone, it has a wider reach. Secondly, since voice interface is used to access the services, the learning required to use such services deployed on the telephony infrastructure is minimal. This is critical since our target p opulation is not exp ected to b e educated enough to learn the complex and non-natural interfaces such as keyb oards. Finally, the model of avatars is such that it has minimal cost implications to the subscribers and users. The system evaluation through these user studies act as a proof p oint to the p otential of VoiAvatar in developing countries.

5.

DISCUSSION

The approach prop osed in this pap er has b een demonstrated to b e applicable in two sectors. However, the approach is a generic one and is equally applicable in other sectors as well. Retail industry is one such sector where the problems faced by small and micro-retailers due to lack of organization has b een recognized. The Indian middle class represents a huge burgeoning market. This realization has led various retail giants such as WalMart [4] to enter Indian market. Several new players such as Subhiksha [2], FoodMart as well as local companies such as [1] have also ventured out into this space resulting into mushrooming of various shopping malls and retail stores across the country. This trend presents a challenge to survival of a huge numb er of small retailers that form an essential part of the country's economy. It has b een argued that the small retailers offer p ersonalized service to their customers including free home delivery, attending to them at odd hours, rememb ering their preferences etc. and therefore they cater to a different segment than shopping malls [17]. Nonetheless, surveys indicate that such mom and p op grocery stores and retail shops are b eing hit hard since they are unable to match the reduced prices that big stores manage by ordering in bulk, sourcing directly from the producers and optimizing their supply chain. Information Technology can again play a key role here in enabling these small retailers to organize themselves into virtual communities or coop eratives. This would give them bargaining p ower with suppliers, help them source products directly from producers and amortize transp ortation, warehouse and other cost across all memb ers. We b elieve that organizing the unorganized as an approach enabled through virtual online communities has the p otential to impact several sectors.

city p olice and the transp ort authority, called Easy Auto 4 . The service has a call-centre based model. Customers must register with Easy Auto to receive door-to-door service. Non-registered customers can avail the service only at Easy Auto stands. The autos can b e b ooked through the web, an Interactive Voice Resp onse (IVR) system or by sending an SMS. Locations are sp ecified using location codes that are provided in the passenger registration kit. Auto drivers need to register. However, they do not have the flexibility to choose their passengers. The fares are fixed by the transp ort authority. The control is more centralized and this model works like the regular cab service model. Comparatively, Dial-an-Auto strives to preserve the autonomy of the auto drivers while at the same time provides similar b enefits to the passengers. The automatic matchmaking will also eliminate the call-center costs which account for a large part of the op erational costs in Easy Auto. The Easy Auto is still in its initial phase of launching and not many auto drivers have registered for this scheme yet. One of the incentives for auto drivers to enroll is that they get insurance protection for their families. In the micro business area there have b een efforts to enable the organization of the micro businessmen to reach a larger clientele using web based solutions. The idea is to create virtual market places where micro businesses can directly sell their products. PEOPLink5 allows artisans in developing countries to sell their products online by using a simple catalog generator software. Local catalogs are aggregated into online market places. The customers can order products through the web and the payments are made directly to the artisan. B2BPrice.com allows farmers in Philippines to trade their goods online. Dilli Haat6 is a central market place in Delhi, where artisans from all over the country are invited to exhibit and sell their products. It is a brick and mortar effort to organize the p eople in this segment. It is a government driven initiative to enable artisans from various parts of the country to reach out to a larger market. The market place simply provides a common venue for the artists and customers to interact. Another common model followed in India is where fruit and vegetable vendors gather at a common market place to sell their products. This is an example of a loose, informal organization of micro businesses where they still function like autonomous entities. A recent effort is babajob.com, a social networking website for finding jobs with a difference. It enables p eople who have access to web (called mentors) to create and maintain online profiles of under-privileged p eople essentially incorp orating them into an online social network. Since this is a new site, at the time of this writing, it is not known whether this service has b een accepted by a lot of p eople or not. We b elieve that finding mentors will b e a key challenge, as will b e the up dating of the social network links as time progresses. The use of voice-driven interfaces for our target p opulation has also b een suggested [18] and exp erimented with [11, 12]. Authors in [13] p erformed a user study to evaluate the acceptance of voice-driven application in rural India based on a sample sp eech-driven agriculture query system. The
4 5

6.

RELATED WORK

An attempt has b een made to make the process of hiring an auto simpler by an initiative launched in Bangalore by the

http://www.easyauto.in http://www.p eoplink.org 6 http://www.delhitourism.com/dttdc/dilli-haat.html

943


study reveals that even illiterate users were able to navigate through the dialog system, though the numb er of errors for such users was higher than those with literate p eople. VoiceGuide7 is a system that allows fast and easy creation of IVRs. It offers an easy to use application development visual editing tool for IVR creation. While this may b e convenient for application programmers, our target segment will b e unable to use it. Our VoiGen system [9] allows anyone to simply make a call to create their VoiAvatar, making the process all the more easier.

[6]

[7]

7.

CONCLUSION

[8]

In this pap er, we rep orted the insights gained from a survey of mobile micro-business service providers (and their consumers). These represent some of the 700 million unorganised workforce in India. Based on the collective painp oints of this community, we prop osed the use of virtual online communities as a mechanism for organizing this workforce and hence improve their work efficiency and productivity. The surveys with the service providers and customers present the need for such a virtual community for the two verticals that we used for this purp ose. Thus we identified the problems, proposed a solution and validated the applicability of solution while constantly engaging the end-users. A key consideration as also observed from the surveys is that this approach has to b e applied while retaining the autonomy of individual participants. Such efforts involving government results in participants losing their autonomy such as in Pre-paid taxi/auto stands and more recently in the Easy Auto service b eing piloted. Without the use of technology, similar practices have b een successful in different forms in developing countries where p eople try to organize themselves to gain some business advantage. Examples include group organized auto-stands, semi-formal fruit and vegetable markets, etc. However, our b elief is that technology can help scale such self/group-organization efforts and let private entities participate in enabling multiple such solutions under a sustainable business model. In the future, we intend to pilot some of these services and explore the applicability of this approach in other vertical sectors as well.

[9]

[10]

[11]

[12]

[13]

[14] [15]

8.

ACKNOWLEDGMENTS

The authors would like to thank Sameer Agarwal, I IT Guwahati and Neera j Chaudhary, I IT Delhi for their help in implementation of Dial-an-Auto VoiceSite.

9.

REFERENCES

[16]

[1] Reliance Industries Limited. http://www.ril.com/html/business/business retail.html. [2] Success Story of Subhiksha, India's largest retail chain. http://www.rediff.com/money/2007/feb/05bsp ec.htm. [3] The Use of ICTs by Small and Informal Businesses. http://research.microsoft.com/ jdonner/ Pap ers/donner micros slides.p df. [4] Walmart.com - Always Low Prices. http://www.walmart.com. [5] United Nations Department of Economic and Social Affairs, Population Division. Urban and Rural Areas
7

[17]

[18]

2001. http://www.un.org/esa/p opulation/ publications/wup2001/2001urban rural.p df, 2003. R. Abraham. Mobile Phones and Economic Development: Evidence from the Fishing Industry in India. In IEEE/ACM International Conference on Information and Communication Technologies and Development (ICTD), Berkeley, USA, May 2006. S. Agarwal, D. Chakrab orty, A. Kumar, A. A. Nanavati, and N. Ra jput. HSTP: Hyp ersp eech Transfer Protocol. In ACM Hypertext 2007, UK, Septemb er 2007. D. Chakrab orty, K. Dasgupta, S. Mittal, A. Misra, C. Ob erle, A. Gupta, and E. Newmark. Businessfinder: Harnessing presence to enable live yellow pages for small, medium and micro mobile businesses. In IEEE Communications, Issue on Networking Technologies in Emerging Economies. January 2007. A. Kumar, N. Ra jput, D. Chakrab orty, S. Agarwal, and A. A. Nanavati. Voiserv: Creation and delivery of converged services through voice for emerging economies. In WoWMoM'07 Proceedings of the 2007 International Symposium on a World of Wireless, Mobile and Multimedia Networks, Finland, June 2007. A. Kumar, N. Ra jput, D. Chakrab orty, S. Agarwal, and A. A. Nanavati. WWTW: A World Wide Telecom Web for Developing Regions. In ACM SIGCOMM Workshop on Networked Systems For Developing Regions, Aug 2007. I. Medhi, A. Sagar, and K. Toyama. Text-Free User Interfaces for Illiterate and Semi-Literate Users. In IEEE/ACM International Conference on Information and Communication Technologies and Development (ICTD), Berkeley, USA, May 2006. T. S. Parikh and E. D. Lazowska. Designing an Architecture for Delivering Mobile Information Services to the Rural Developing World. In Proc. Intl. Conf. on World Wide Web (WWW), May 2006. M. Plauche and M. Prabaker. Tamil Market: A Sp oken Dialog System for Rural India. In Working Papers in Computer-Human Interfaces (CHI), 2006. C. K. Prahlad. The Fortune at the Bottom of the Pyramid: Eradicating Poverty Through Profits . 2004. K. Ramamritham, A. Bahuman, C. B. S. Duttagupta, and S. Balasundaram. Innovative ICT Tools for Information Provision in Agricultural Extension. In IEEE/ACM International Conference on Information and Communication Technologies and Development (ICTD), Dec 2006. S. Sakthivel and P. Joddar. Unorganised Sector Workforce in India - Trends, Patterns and Social Security Coverage. Economic and Political Weekly, 2006. M. Seth. Kirana vs sup er-market. Business World, http://www.businessworldindia.com/july0504/ casestudy01.asp, Jul 2004. J. Sherwani. Are Sp oken Dialog Systems Viable for Under-served Semi-literate Populations? PhD Thesis Proposal, Carnegie Mel lon University, http:// www.cs.cmu.edu/ jsherwan/JS-proposal.pdf, 2005.

http://www.voiceguide.com

944