Connecting the Dots
Village Gurundia is one of the 51,311 dots on the map of Odisha - one for each village. The dot gets bigger on the district map - showing 30 in all - and changes into an amoeba-like shape on the map of Sundergarh, one of the state's 315 blocks. Zoom in further, and you can see the house of Sunita, 35, who doesn't struggle with cooking for a large family any longer as a new Indian Oil LPG dealership has been commissioned just four km from her house. Getting a refill no longer means a daunting 20 km trek hefting a cylinder.
If three crore women like Sunita in 710 districts are feeling privileged over the past 16 months, it is because, for the first time - with the help of some heavy data analytics - the government is directing opening of LPG distribution centres in areas that need them the most and not where companies see the most demand. In three years, 10,000 LPG centres will be opened so that almost every citizen in each of the 6.4 lakh villages has a distributor within 10-15 km of their homes.
The effort by the government is part of the Pradhan Mantri Ujjwala Yojana under which five crore LPG connections will be provided to below poverty line, or BPL, families with a support of `1,600 per connection. Issuing connections, however, is only half the recipe for ensuring 100 per cent LPG use. Supply will also have to be sustained over time.
This is not an easy task. To plan this, decision-makers needed to find where in the 700-plus districts are LPG distributors located to identify LPG-deprived villages. "We were clear from day one that the five crore women beneficiaries will be identified purely on the digital platform...the scheme will be implemented without any human intervention," says Petroleum and Natural Gas Minister Dharmendra Pradhan. "To sustain a scheme like this, you need information. You need a monitoring mechanism which gives you the ground reality."
In short, you need data science, an optimisation algorithm to identify the ideal locations for the new centres.
Urban Bias
In India, each of the 18,000 LPG distribution centres is operated by one of the three government-owned oil marketing companies or OMCs - Indian Oil Corporation, Hindustan Petroleum Corporation and Bharat Petroleum Corporation - as independent for-profit businesses. They were given the freedom to select and regulate the centres ever since the administrative price mechanism was dismantled in 2002. These companies chose locations based on market potential, leading to high concentration in urban areas. In Lucknow district, for instance, out of 88 LPG centres, 75 are in urban and 13 in rural locations. As against an LPG centre for every 7,500-odd households in urban areas, in rural Lucknow, there is one centre per 21,500 homes. That was the level of skewed distribution the government wanted to address.
So, the ministry set its sights beyond simple market economics and decided to look in the right place - the vast areas deprived of cooking gas. The focus was specifically on the 14 states and union territories where distribution was the most haphazard.
Business Case
"There was an underlying contradiction in the mandate," says Prukalpa Sankar, Co-founder, SocialCops, a data intelligence company that partnered with the government for Ujjwala. On the one hand, the solution had to ensure that the dealerships were viable. On the other hand, they had to be so located as to ensure access for the underprivileged. How would OMCs even begin to solve the conundrum?
There could have been two approaches. The old process, where the OMCs could have identified locations based on population and market potential on a district level. Or, they could have assessed actual need through bottom-up data mapping and identified locations based on an algorithm. For the first time, market potential wasn't going to be the only criterion.
The only data SocialCops had to begin with was on refill sales. This alone would never put an LPG centre near the homes of the poor. So, this was layered with a lot of publicly available data - population, income levels, roads, local mandis, ATMs, post offices, health centres and, of course, LPG penetration - to feed the algorithm for the solution.
Geo-tagging Exact Locations
But there was still a gap. Existing LPG dealership locations had never been correctly recorded. The OMC database merely registered the district of a dealer instead of the complete address. SocialCops decided to record exact geo-coordinates (latitude and longitude) of every LPG dealer in India. An Android app, Collect, was deployed to capture the phone's GPS location even if it was offline. All that the dealers had to do was to download the app on their phones. The work for collecting the co-ordinates - crucial as one wrong detail would have wrecked all other calculations - was done in just a month. The next steps were marking these locations on a map. Maps of districts, sub-districts and blocks were overlaid with distributors' location coordinates to identify areas with and without LPG centres.
The process kept evolving, and so did the work. When the team went in with the first iteration of what it believed was the right solution, minister Dharmendra Pradhan insisted on zooming onto locations he knew well in his home state. The dashboard zoomed into Sundargarh, which showed LPG dealership locations captured via geo-tagging. He was sure there was no distribution centre in Lehripara tehsil. But a pop-up on the dashboard showed a dealer's name and phone number. The minister dialled the number on the screen. It turned out that the centre had opened just three months ago because of a commercial complex that was slated for launch.
Satisfied with the accuracy of the mapping, Pradhan checked out the locations suggested. The northern part of the block was a sparsely populated area, and the algorithm - this was the first iteration that was built solely on the market potential - recommended centres at 20 km from the remotest villages. That's when the minister asked if the solution could propose locations within a reasonable distance.
The team flipped the algorithm on its head, making 'access' the primary parameter. It ran an algorithm over the country to create clusters that needed at least one LPG centre. Next, it assessed the market potential of each of the 5,500 clusters based on population, income levels and LPG penetration.
What followed was a location potential index using parameters like access to roads, proximity to a mandi, ATMs, thus creating a ranked shortlist of 10 locations in every district. The intuitive algorithm came out with customised solutions for different districts depending on how densely populated they are - the market potential going higher with greater density. So, if the algorithm suggested that LPG centres be located at a distance of 15 km in Himachal, Uttarakhand or Chhattisgarh, in UP, it was just half that distance. The solution could vary even within a state; in Gujarat, for example, centres were recommended at a distance of 7.5 km in Ahmedabad and Gandhinagar, in Kachch and Narmada, the distance was 15 km.
Now, 6,000-odd locations are being advertised for tendering. In September, the ministry issued Letters of Intent to over 300 new distributors in UP. In all, about 1,000-odd letters have been given to new dealers in states like Odisha.
"When we started the scheme, we were not sure if we'll be able to roll out uniformly across the country...whether we'll be able to achieve the numbers we are talking about," says Ashutosh Jindal, Joint Secretary in the Petroleum Ministry. Village-level information and solution is not easy, he says. "But technology has been a great enabler".
Switching to clean fuel isn't a one-time process. The consumer pays `649 (the market rate), the money a BPL family subsisting at less than `50 daily will make over 13 days. So, it's not surprising to see beneficiaries drop out at the time of refill when the subsidy is no longer available.
Indian Oil's Manish Grover, industry lead on Ujjwala, however, points out that the refill rate for Ujjwala connections given out until March 2017 - the cut-off chosen to track those due for refill -- is a substantial 75-80 per cent. So, if about a fifth of the beneficiaries have not come back for refill, it is in keeping with the trend in rural households that mostly use LPG as an adjunct to a biomass chulha. The average monthly LPG consumption in rural areas hovers around four cylinders a year; the Ujjwala average is 3.5, he says.
However, affordability and last-mile connectivity are real issues. The ministry is responding with a slew of measures. It started LPG panchayats to address the problem in the last week of September. Smaller 5 kg LPG packs priced at `179 are being provided as specialised offerings for these beneficiaries. There is more. Until now, only BPL women under SECC (Socio Economic and Caste Census) were eligible. An Ujjwala Plus initiative is in the works to include those left out; in fact, anybody earning less than `2 lakh a year can apply without any 'poverty validation'. The government is also exploring business models with microfinance companies to enable daily payments.
A Blueprint
Ujjwala is a unique experiment where decision-makers, sitting in Delhi, used science to come up with a pan-India solution to address the conflicting objectives of both market and access for remote corners of the country. "For the first time, a planning exercise that would typically take two years and suffered from subjectivity was done in a few months through a standardised solution," says Sankar.
The data-based solution for Ujjwala is essentially an exercise in resource allocation. It might well work as the blueprint for newer use cases - to open schools, health centres or agriculture training counters. The variables might differ and the weights might vary but the process remains essentially the same, says the SocialCops co-founder. It can be replicated by multiple government agencies for their schemes. In fact, any industry can use this blueprint for distribution now that the groundwork has been done in creating a database for these markets. Recall how the insurance industry used to keep away from rural markets because distribution was a tough nut to crack in the initial days of deregulation.
The takeaway from the Ujjwala model is not just that there are markets beyond urban centres but that these can be tapped into profitably with data. That's because the cost of data-driven decisions overrides the price of guesstimates.
