Inc42 UpNext: UrbanClap On Designing AI Systems To Expand And Scale Up Sustainably

To celebrate India’s rising startups, Inc42 is profiling a new soonicorn every Friday in the Inc42 UpNext: Unicorns Of Tomorrow series. For the next few months, we will be speaking to founders and cofounders at these potential unicorns and shining light on their journeys and growth stories. We begin the series with a look at local home services marketplace UrbanClap.

Gone are the days when Indians — even in the large metros — had to wait days on end to get a leaky tap fixed or fix the fan ahead of the summers. Home makeovers and improvements are never more than a tap away these days, just like food, groceries or shopping. Just like small restaurants have seen an uptick in reach thanks to food delivery apps, a city’s disparate local service providers such as masons, electricians, repairmen and more have also found a digital home.

Riding on the high adoption for consumer services, UrbanClap has become THE marketplace for home repairs, maintenance and beauty services across India’s major cities. Founded in 2014 by Abhiraj Singh Bhal, Raghav Chandra and Varun Khaitan, the company has adopted a marketplace model similar to Uber, Zomato or Fiverr, and has constantly tailored its services to meet the demand in the market. UrbanClap offers beauty treatments and spa services at home, appliance repairs, plumbing, maintenance, carpentry, cleaning, and painting services.

With volumes touching 6 Lakh requests per month, running operations can be hard but not impossible. But it’s tougher to build a platform which treats both users and service providers with empathy with standardisation. This why UrbanClap has taken to integrating artificial intelligence engines into all its systems, but before that the marketplace dealt with a slew of problems along the way. 

UrbanClap cofounder and technology head Chandra told Inc42 that the company does a lot more than simply listing professionals and customers. It solves for standardised pricing and catalog, skilled training, product warehousing, and financial-success of the professionals.

Building A Bulletproof UrbanClap

Chandra went deep into how UrbanClap scaled up its software, decided on the monolith vs microservices debate and how the company has used team culture to build what he called “bulletproof platform”. From beauticians to plumbers to electricians to massage therapists, UrbanClap’s platform handles many different services across India, itself a rather fragmented market. So what goes into designing a platform with so many variables?

Chandra said that the long list of categories like beauty, home, painting, international etc are not separated on the engineering side, which reduces the complexity. “These platforms serve as the infrastructure for every category. Think of training as a platform, finances, product procurement, standardisation and pricing engine, market access as a platform. So now any category can plug into this and utilise this platform.” 

This has allowed the five-year-old startup to scale up rapidly and adapt to the changing trends in the market.

“For example, if you take the training. Training is a huge problem. Because we have very skilled tasks such as plumbing or ac technician, as opposed to say delivery. We can ‘techify’ training and even make evaluation tests.”

Unlike a ride-hailing startup, where the driver comes pre-trained more or less, for UrbanClap, one of the major investments in training and skilling people to fill the services gap. And top-quality talent becomes harder and harder to find so training has to get nuanced. UrbanClap aims to train a million professionals in the next 5 years to meet the demand in Tier 2 and 3 cities. “We have 20K professionals whose 100% revenue comes from UrbanClap,” Chandra added.

In addition to its training programme, the company is now working with government entities such as the national skill development corporation (NSDC) for the mobilisation, training and certification of service professionals across India. It currently has around 20K service partners, and daily, it fulfils 25K – 30K services in a day.

As cofounder Bhal told Inc42 in an earlier interaction, “The one thing that has helped us thrive is a very sharp focus on the quality of experiences, both for our service professionals as well as our users.”

The Logistics Of UrbanClap

Even if 1% of our users register unsatisfactory service, thats 6K unhappy customers in a month.

Chandra reiterated how it’s not just about having the service partners on board. There are some logistics challenges as well such as making sure the assigned service person is reaching the user’s house at the assigned time. Chandra claims UrbanClap has solved it to a certain extent.

“If you see any of the new marketplaces (food delivery, ecommerce or ride-hailing) it seems like a solved problem but in the beginning, it was a huge issue. Now it looks like a piece of cake but it takes the ecosystem a lot of time to adapt and it takes a lot of structuring and innovation to make it seem like a sustainable model. And it only gets worse when it scales.”

That scale is something UrbanClap is dealing with now. It inched closer to the unicorn club with the $75 Mn Series E funding round led by Tiger Global with participation from existing investors Steadview Capital and Vy Capital. With the round, the company is within touching distance of the unicorn status that many startups yearn for.

UrbanClap reported operating revenue of INR 116 Cr in FY2019, which is a 150% rise in comparison to last year. It reported fulfilling 3.3 Mn service orders in FY19, which is a 3x growth. And this growth in usage has come with technical and technological challenges.

Chandra circled back to the problem of service partner arriving on time. “For a service partner to show up on time, it takes complex math to understand what their behaviour has been in the last few months, and what kind of jobs they like; do they travel to the location of the service request? What is the acceptance ratio? We analyse this through prediction engines.”

With its tech and prediction engines guiding the scale-up strategy, UrbanClap is on track to hit 7.5 Mn service orders for FY19 with an annual revenue run rate was pegged at about INR 200 Cr.

The Orthogonal Approach To AI Systems

Some problems are harder to solve than others. “Matchmaking became one of the most complex systems for us. This is the heart of the problem we were solving. It’s tough because it needs to take into account every single data point, when and where to take demand, time slots, people’s behaviour and more”

The art in artificial intelligence comes from how your different systems interact.

Chandra explained that UrbanClap uses an algorithm for availability, another for predicting if the service professional is likely to take leave on that day, and yet another for the rating and quality assessment of professionals. “So they are all orthogonal systems, each has a different metric it is optimising for. Then there will be one papa (father) system, which takes these predictions and turns out a super prediction of what is right for the top-level business metrics’.”

These systems have helped UrbanClap streamline its operations and now the company is shutting down its other non-core businesses, including wedding services and photography. But it will scale up its operations for home services and beauty, cofounder Abhiraj Bhal said recently. UrbanClap is also planning to introduce luxury services to compete with high-end luxury spas.

One big debate in engineering is how should teams be structured, which actually percolates down to culture. Chandra described an assembly line-like system of passing down values and vision. “This is a factory setup, that means every team makes one part of the solution and passes it on. I think this is a non-ownership model, there is autonomy.” One big debate in engineering is how should teams be structured, which actually percolates down to culture. Chandra described a pod based structure at UrbanClap where designers, product managers and engineers work on problem statements. “Missions drive org structure, which means every team is independent and fully equipped. We solve for cross team dependencies by having a high bar on engineering design. I think this drives most ownership, there is autonomy,” he added.

Artificial intelligence is not just allowing startups to automate parts of their operations, but also scale up sustainably. Chandra said UrbanClap has had to account for this as well, but it’s not a problem fully solved. “We have to keep going back to the drawing board all the time. Different scales have different designs. Every time we reach a significantly higher scale we need to rejig because each of the underlying systems become too complex and has to be broken down.”

Note: We at Inc42 take our ethics very seriously. More information about it can be found here.