One thing the tech software industry is great at is failing fast. This concept of “failing fast” means, at least in part, rapidly bringing a “minimum viable product” to market, even if it’s a little rocky and feature-light, and then fixing it as it gets adopted by iterating with new releases. We see this all the time in software-driven products and even with some hardware ones, such as mobile phones. Remember the first iPhone? Super cool at the time but, in retrospect, it needed some work.
The fail-fast approach has some significant benefits, as it takes less time and money to get a product to market, something important to entrepreneurs and investors alike. So that’s a plus. It also allows for market feedback into the next version of the product, another advantage. And most of the time, in tech, it just isn’t that big a deal that a feature didn’t work or a product doesn’t have a full feature set. Worlds of Warcraft software won’t let you join a particular guild today? Oh well, hardly a crisis, beyond the existential.
The fail-fast and MVP concepts are now being talked about actively in the healthcare IT/digital health world. With all of the tech people and money being poured into what are now called digital health companies, advisers and business leaders accustomed to the launch-and-iterate model are encouraging entrepreneurs to bring their minimum viable products to market fast and then fix them as they go along, in the same manner as would someone at Facebook or Microsoft or Apple.
Very recently Apple announced HealthKit and Samsung announced S Health and Google announced Google Fit, all within a few weeks of each other. None of these companies are really recognized as healthcare companies, but all are well recognized for software and hardware and integrated tech solutions. I am, admittedly, not that familiar with the history of Samsung product offerings, but I am one of the zillions of consumers who have happily used early Apple and Google offerings only to jilt them later for the next iteration when the newer, better ones came out. Consumers always like the newer, lemon-scented versions, don’t we? Tastes better, less filling.
But in healthcare, can this fly? The whole point of the FDA, for instance, is to ensure that products that get into the market are not dangerous or error-prone or generally worthless. They have specifically declined to regulate most of the new software offerings and particularly those of Google and Samsung and Apple. But is it a good thing if the product they first bring to market isn’t good for consumers the first time out? Minimum viable product is often abbreviated as MVP, just like Most Valuable Player. But I am a little concerned that it might come to stand for Mightily Vexing Problem if we go all-in on this fail-fast methodology in healthcare.
I am not advocating for more regulation, but rather for a far more careful approach when it comes to products that have the potential to affect people’s health. Everyone knows that you can’t intentionally bring a minimum viable product to market when it comes to medical devices, for instance; the damn thing better work. “Oh, sorry, Mrs. Jones, that your automatic defibrillator only works every other heartbeat, but, based on your feedback, we will fix that in version 2.0!”
We do see iteration in the life cycle of medical devices, but the first round is really supposed to do the trick, with the subsequent releases dedicated to incrementally improving already very good outcomes. Pharmaceuticals are supposed to work the same way. You can’t (at least intentionally) bring a partially finished drug to market and promise to make it better in the next formulation. “I am so sorry, the birth control didn’t work, but just think what a lovely mother you will make and we promise to ensure she is an only child….” No, I don’t think so. Yes, it happens, but when it does the product is yanked from the market.
But this is, in many ways, the essence of the software development fail-fast model and it may work just fine in situations that are not critical to life and death. Or can it? Few are yet arguing that the new AppGooSam health tracking platforms are essential to sustain life, but I know the companies are hoping they will become essential parts of consumers’ self-care management. Also, by bringing Epic and Mayo and others to their party, it is pretty clear that Apple expects other key healthcare industry players to think so too. I am sure that a gazillion new health-related apps will be built based on these new platforms. Will it be good enough if they come out limping and take a few years to start swinging?
Regulated or not, people make decisions based on the health inputs they are getting from wearables and sensors and all that newfangled stuff. Or at least they are supposed to. If the products and platforms are mediocre at best (and many of them have been so far), they will not get adopted and they will fail fast in the true sense of the words (aka die).
But if they are misleading because the algorithms aren’t well thought through or the data accuracy is just off, they could actually, even if unintentionally, cause harm. I can think of a few examples, such as wearables that say they measure steps or calories but are so far off from accurate that they are meaningless to consumers. Worse yet are those that purport to measure blood pressure but don’t really work when people are moving around a lot (because blood pressure is really hard to measure accurately when people are moving around a lot). If someone is relying on a device to check their blood pressure and it is often wrong, but they don’t know it, that could lead to a really bad outcome.
I think we have to do some work on this minimum viable product concept in healthcare IT. I don’t think that “works good enough for now” is going to be good enough for patients. On the other hand, it is also really bad (aka, expensive) to be in development mode forever without market testing. The perfect is the enemy of the good, to be sure, but the good enough is also the enemy of the good, in my view. “Good enough” needs to mean good enough for patients to reliably depend on the information and suggestions such that they are not put in harm’s way.
Pilot studies with customers are great for this testing, even though frustrating for vendors (“pilots” that never land have become all too familiar to entrepreneurs), as there is heavy oversight during the process, or at least there should be. We need to see more randomized clinical trails, even if not the regulated type, in the digital health world to ensure that products purporting to emulate medical capabilities are actually functionally and informationally equivalent, if not better. In virtually every medical device/diagnostics pitch I hear there is discussion of sensitivity (e.g., a proxy for the percentage of sick people who are correctly identified as having a tested-for condition) and specificity (e.g., the percentage of healthy people who are correctly identified as not having the condition as a result of a test). I am not sure I have ever heard those words uttered in a digital health pitch, and that is not good. Home diagnostics and sensor-based wearables should be talking about this stuff as part of their MVP. You should not be planning to iterate yourself to accuracy when it comes to important data.
As usual, it is about finding a balance. Talk to customers, get feedback through pilots, launch and re-launch—these are all important parts of getting to market faster and cheaper and, as a VC, Lord knows I appreciate faster and cheaper. I would just like to see a higher standard of care employed in the process when it comes to healthcare than in gaming or inventory management or social networking. I am pretty sure that broad-based failure to adopt a more measured approach to the release of healthcare IT and digital health products could have a negative long-term effect on market size, also near and dear to my heart; This will be particularly true if patients and physicians get too frustrated with badly designed products or, worse, bad things happen.
