Shashikant Kore interviews me about my previous startup.
(From 2008 to 2011, a doctor friend of mine, Amit Paranjape, and I ran BharatHealth, a healthcare startup, which failed. In conversations about startups, I would often mention learnings from this experience. For a long time, Shashikant kept asking me to write an article about it, and when I kept putting it off, we finally decided to do it as an email interview.)
Please provide your academic and professional background before starting up.
I am currently a co-founder and CTO at ReliScore.com, a startup focused on helping companies filter job candidates based on evaluation of actual job-related skills. I advise a couple of fintech companies and a Govt of India initiative for startups.
In the past I’ve worked for large companies, and small; I’ve worked in India and in the US; I’ve seen a successful exit, and I’ve seen a dotcom failure; I’ve done product development, and I’ve done research; I’ve written consumer software, and I’ve written enterprise software; and I’ve been a developer, I’ve been an architect, and I’ve been a manager (but hated it).
I’ve done my BTech in Computer Sciences from IIT-Bombay and a PhD in Databases from the University of Wisconsin, USA. I am an inventor on 18 US patents, 2 European patents, and 1 Japanese patent. I’m interested in a number of areas of computer science, including: highly scalable systems; distributed and fault-tolerant software systems; text search, information retrieval, and analysis of unstructured information.
You took a sabbatical after leaving Veritas/Symantec. What were your plans for this period?
I had gotten bored of my big-company job, and I was lucky financially to not have to worry for a few years. So my plan was to just work on things that I found interesting.
As a result, I spent a couple of years doing a bunch of things: I worked on my Hindi song lyrics website . I helped Meeta (@meetalks) set up her movie reviews site. I started https://punetech.com because I felt that the tech community in Pune needed a forum like that. I consulted for a few friends’ companies.
How did you come up with the idea for BharatHealth? Who were your co-founders at BharatHealth?
The doctor friend came up with the idea. He felt that a web-based system to capture all the health records and interactions between a patient and a doctor could revolutionize how healthcare is conducted, especially in case of chronic problems where the care requires lots of little tweaking over the long term. (Note – chronic diseases are defined broadly as conditions that last 1 year or more and require ongoing medical attention.)
He approached Amit Paranjape and me and both of us liked the idea. That’s how BharatHealth was born.
Who were the target customers of BharatHealth? What problems did BharatHealth solve for those customers?
It’s easier to explain with an example. Let’s focus on diabetes patients as that was one of the first verticals we targeted.
Our initial target customers were doctors providing long-term diabetes care. The idea was that the doctors would offer BharatHealth to each of their patients. In the beginning, the patient’s basic history, the doctor’s current diagnosis, and medication protocol would be uploaded to the system.
After this, on an ongoing basis, the patient would keep uploading blood sugar readings into the system as recommended by the doctor. The doctor would check this on a regular basis and if any problems were noticed, they would recommend adjustments to the medication.
Similarly, if the patient had any non-emergency questions, they could just ask it in the system itself. The doctor would check the system a few times a day and answer any questions. This was much better than the patient having to come to the clinic and wasting a few hours, or calling up the doctor on the phone. More importantly, the doctor would be able to see the full history of the patient right in front of them along with the question so the answer would be quicker and more accurate.
We felt this was a win-win scenario.
What were your initial hypotheses for BharatHealth?
We felt that this would:
- Improve patient outcomes because of better availability of patient history, symptoms, medication
- Decrease the amount of time spent by the patient waiting in waiting rooms
- Increase the throughput of the doctor because the system would allow answering a question in 30 seconds, whereas the same question on the phone or in person would end up with at least 5 minutes if not more.
- Patients would be happy to pay the doctor for the use of this system
- Doctors would be happy to pay us a cut out of #4
- Once a system like this was up and running, we would be able to monetize it in various other ways
What did the early version of the solution look like? How long did it take to get the first version in the hands of the customers?
This was at a time (2009-10) when touch phones hadn’t yet become common in India. Hence the system was a web-based system. We expected that all doctors and most patients would access it from laptops or desktop computers, while a small fraction of the patients would use smartphones to enter their test results and questions.
We built the system using Python+Django. It took a couple of months to get the first version to the customers.
How did you reach out to the early customers for BharatHealth?
Quite early we decided that this would be a B2B2C solution. We focused on getting doctors as our customers, and the doctors would ask their patients to join the system. In other words, we would never be selling to the patients directly.
We used our doctor co-founder’s connections to reach out to doctors in Pune, and made the sales the old-fashioned way – by meeting them, showing a demo, and asking them to join.
What were some of the early marketing channels you used? Did you also have a sales team?
We only did direct sales. We did all the sales ourselves. The idea was that once we had figured out how to do the sales ourselves and streamlined the pitch and process, we could then hire a sales team. However, we never reached that stage.
What was the customer feedback of initial versions? Did any feedback surprise you? How did the initial feedback shape your product roadmap?
I’ll answer this in terms of general “customer reactions”. We ran into these major categories of customers/potential customers:
- A small number of doctors quickly understood the pitch and started using the system. They were generally happy how it worked and how it helped them and their patients.
- A number of doctors said they liked the idea but they did not think it would work in their situation because of one of these reasons:
- They felt that a good fraction of their customers were not tech savvy and would not be comfortable using a web-based system
- They would need to teach their receptionist to use the system, and that would be impossible
- The system was in the cloud, and they didn’t think the internet was reliable enough
- Some of the doctors liked the system, sounded enthusiastic about wanting to use it, but when it came time to start using the system, they never got around to it, because always some higher priority thing came up and operationalizing this system kept being pushed lower on the todo list.
What were some of the hardest challenges in terms of product, marketing, sales and the market itself?
An experienced entrepreneur had told us that our product required behavior change on part of the doctors, and that would be very difficult. We never really appreciated how big of a stumbling block this was. In retrospect, I feel that a number of the problems in points #2 and #3 in the previous answer all stemmed from the fact that behavior change was required and our product did not provide a big enough incentive for driving that behavior change.
Speaking of incentives, based on our initial market validation (with about 5-10 friendly doctors) we felt that the significantly increased patient convenience and potential improvements to medical outcomes would be strong incentives for doctors to use this system. However, once we moved past our friendly doctors, the impression we got was that in most cases, “Can this system get me new patients” was the more important question on their minds. And since we did not do a B2C play, our answer to that question was No.
Another problem we ran into was that doctors are used to the pharma-medical-representatives style of sales, where there is lots of hard-sell, lots of money flowing around, and royal treatment of doctors, and without that kind of a sales model, we were not very effective at selling into this market.
Another problem with this market (or the way we were selling the product) is that every doctor wanted customizations to the product to incorporate their own workflow into the system. While this is true of the early stages of any B2B product, I felt that this problem was more pronounced in case of doctors, because most of them have high-throughput practices with workflows that have been optimized over the years, and getting them to change those workflows would require too much behavior change. We tried hard to build a very flexible system that could easily incorporate the requested changes via configuration, but in many cases we had to give up on the customer after a number of rounds of changes.
All these reasons that affected our ability to sell to individual doctors were multiplied by N when we tried selling to hospitals. In the former case, we had doctors who didn’t feel they could convince their patients (and receptionists) to use the system. In the latter case, the problem was compounded because we had hospital administrators who had even less confidence in their abilities to convince doctors (who would then have to convince patients).
BharatHealth came up at a time when mobile phones were ubiquitous, but smartphones were just starting out. How big a role phones played in your solution? Would a smartphone have helped in some way?
In this sense, we were too early to the market. A few years later, when touchphones were truly ubiquitous, the resistance to our solution would have been much lower. Without the guarantee of an always-on smartphone with every patient, the (mental) barrier to entry was too high, and as a result the behavior change required was too high.
What were the signals that indicated the product is not working as per your team’s expectations?
While we made quick progress with the initial set of doctors we had talked to, we soon found that expanding past this inner circle was quite difficult. As we approached more doctors, we found that the meetings and demos went quite well (according to our possibly inexperienced judgement), and they sounded positive about the product, but getting them to take the next step was difficult.
In many cases, in spite of positive feedback, we would not be able to get them to set up their account on the system. There was always a plausible reason for it; some emergency that came up, or some other higher priority activity because of which they wouldn’t have time to learn the system. And in the cases where we got them to create an account and learn the system, we found that they wouldn’t actually sign on patients.
As a result, growth just slowed to a crawl.
At what point did your team decide to pause efforts on this product and company?
During a long stretch of low/no growth, Amit and I were involved in some unrelated initiative (part of our PuneTech work), and we hit upon the idea for a different product in a space that we understood much better (hiring for software industry) and we felt it would be easier to sell because we understood the potential customers and their motivations much better (having worked in that industry for 20+ years each). As a result we decided to stop BharatHealth and ReliScore (our current startup) was born.
Last decade saw significant improvements in technologies like smartphones, wearables and machine learning. You mentioned smartphones would have lowered barriers for adoption. What healthcare problems, that you are familiar with, can now be solved in a better way with new technologies?
Since I moved to a non-healthcare startup, I haven’t really kept close tabs on what all is going on in the healthcare tech space.
The potential for use of smartphones, wearables, and machine learning in healthcare is huge. You pick any aspect of healthcare and it has the potential to be revolutionized by these technologies.
However, the challenge primarily lies in the “behavioral change” that I mentioned earlier. This falls into two categories: preventive healthcare vs acute healthcare.
Preventive healthcare could be applied to everyone, and it could have a huge impact on the world. However, getting people to start using an app and/or providing it with the data it needs is not easy. Most people are quite lazy about taking steps now to prevent future problems, especially if it involves a payment. And if it doesn’t involve payment, then there is always the concern about how the company will use the data and whether privacy will be preserved. As they say it is easier to sell aspirin not vitamins. Because of this the companies most likely to make good progress in this area would be those like Apple, who have: 1. a good brand image, 2. an app or device that is already in people’s pockets, and 3. the data entry is minimal-effort, ideally zero-effort.
For acute healthcare (i.e. solving a problem you have right now, i.e. aspirin instead of vitamins), doctors and other providers have to be in the loop (given the current state of technology and regulations), so this involves handling behavior change and incentives for those players. As is clear from our failure, this is not something I understand well, and I haven’t yet seen a compelling model from someone else.
Since you faced challenges with B2B2C model, which seems to be the primary route to market, what are your thoughts on using the B2C approach?
I think the right kind of B2C approach could succeed, but the focus has to be more on psychology, incentives, distribution, rather than the technology. And would need deep pockets also. We considered pivoting to a B2C model, but we decided that we neither had the expertise nor the budget to pull it off.
You mentioned, you thought about B2C pivot, but didn’t work on it for a variety of reasons. Do you remember what some of those B2C ideas were?
We thought of a lot of ideas, only some of which I remember. For example:
- Allow end-users to capture their medical history and reports, and enable easy sharing of this data across patient, doctor, pathologists/labs, family (e.g. NRI children), and for second opinions.
- Same as #1, and get medical insurance companies to offer discounts to patients who show good control over time.
- Allow end-users to capture their medical history, and allow them to ask questions to our team of doctors
- Same as #3 but have software triggers and/or doctors looking at the data to warn patients of potential problems that they’re ignoring
- Same as #3 but allow the patients to suggest which doctors they would like on the system and use that to create pressure on doctors to join.
How long did you work on BharatHealth, from starting work on it, till deciding to stop working on it? I could add it to give some idea of commitment for new entrepreneurs.
Approximately 3 years.
If a new entrepreneur wants to explore this space again, what advice would you give?
I would say this:
- Focus on psychology, incentives, distribution, rather than technology. Technology is the least of your challenges. Ask yourself if you can do the first 3 to 6 months of your startup using just WhatsApp and an Excel spreadsheet and you personally handle all the customers. You’ll be surprised at how often this is possible.
- To be able to do #1, steep yourself in the culture of this domain. Talk to doctors who are not your friends, hospital administrators who are not your friends, and so on.
3 thoughts on “Learnings from Building a Healthcare Startup in India”
Dear Navin – must say that this was quite an exhaustive account of your experiences and will no doubt be invaluable to anyone that looks to get into this space. That said wanted to highlight some key areas where my views diverge. I write this from being at the customer end of such a tool for my better half (bang in the middle of your target space) and honestly am still in the market looking for a solution. Things that I would look for in a solution such as this:
1. Privacy, Privacy, Privacy of data.
I saw that you had taken a pot shot at the medical community in general (as being freeloaders) and I am sure that your have reached out to a large cross section of them, to have arrived at your conclusions. I still am a believer of the noble profession and anyone who spends 9 years studying, at the very least, would graduate with a modicum of “vinayata” (for the lack of an exact english translation). The primary issue, the way I see it is how safe is the patient information with the company that claims to store it in the cloud. What kind of indemnities would you provide for a breach. What stops the company from making a fast buck by selling the data to other startups that cross/upsell a lot more.
With the pedigree that you and your partners command, I would think that the latter would be a less of a worry, but in the age of ransomeware, it would take a lot more meat to pitch the solution.
2. Data Input
On both sides – patient and doctor.
As much as I admire doctors, their tech skill leave a lot to be desired. An intuitive and slick user interface for a casual user (Gen X, Y & Z included) may not be used as the benchmark for doctors. It is almost like they are taken aside in Med school and threatened such that they are scarred for life. So, typing as the input method is woefully inadequate. Voice as an input method has come a long way now and should be explored as an alternative – but deployed ONLY after due testing to ensure that at least 95% is coherent and only <5% need manual keyboard overrides, else this is DOA. For the user side, its a different game altogether: Similar to Garbage in = Garbage out dictum in CS, if the inputs by the users were recorded incorrectly (on account of typo / incorrect understanding), the doctor's decision / response may prove quite an issue. No matter whose burden of proof, this will have a very negative user experience all around and may be life threatening. Also a lot of the patient community in the geriatric age-group may find the tools hard to navigate and even more so on a smartphone. Here to integrating voice may be the ultimate panacea.
3. Build for the user
I went through the blog, and may be I did not see the specs, but there are a few bells and whistles that would ring a bell with docs closer than you thought.
eg. tagging key drugs prescribed – ability to extract (anonymized) patient report that can be used for studies that a lot of doctors conduct. This can be a much bigger draw than the paltry incentives that you as a tech provider may be able to give. Working with Pharma companies to offer such studies may be an added incentive.
4. Always on network
You would be amazed how difficult it is to get reliable coverage and network access in clinics and elsewhere. Until we cross that hump, I would always recommend an offline entry with synch capability for the tool-set
5. Get scale in distribution
Pharma companies are always looking to up their game. Making them your sales channel by offering your licensing for a royalty, may help them get stickiness with the docs (something they would give an arm & a leg for) and also differentiate their offering. They would happily pay the subscriptions for doctors. The game is not to fall for quick buck and selling patient data. But there are better ways to monetize – just imagine the possibilities where let's say Merck's drugs where available for a particular diagnosis would default on the prescription, and the doctor has to manually override the default. I am not even starting to talk about the immense potential for co-branding and such – the possibilities are endless. Honestly, your sales channel and distribution simply lacked the ability to scale.
Lastly, your learnings and the sage advice that you have at the bottom of the blog (last question) are worth their weight in GOLD
I hope you take my critique in the right spirit and happy to connect if you want to have a follow up conversation. I have shared my email, and you can reach out to me through that.
Especially, the space that you are exploring as part of your new startup also happens to be another domain that I am quite passionate about – happy to share notes and help out in any way.
Background: M&A / Tech / Education
For point #2: keep in mind that this was 12 years ago. Audio-to-text was nowhere close to being reliable, especially for a domain like medicine.
For point #3: we spent a lot of time with doctors, and we did talk about the advantages of having all the data available for research. Most of them weren’t interested.
For point #5: we did talk to a few pharma companies to see if they could be a distribution channel. But they weren’t interested primarily because we were too early to the market (and hence the same reasons as I have listed in the answer to the “hardest challenges” question.)
For points #1 and #4 I think I’ve said the same thing, so I’m not sure why you’re calling those a critique of my post.
Also, I don’t know why you claim that I characterized doctors as freeloaders. I haven’t implied that anywhere.
Wonderful lessons, expressed so simply. Thanks Navin, this should be a must read for all budding entrepreneurs.