A curmudgeon might ask "what is the difference between product market fit and having a good product?" Which I sometimes struggle with - is this just old wine in new bottles? But there are some subtle distinctions worth exploring, and in particular it is what if you have a good product but aim for the wrong market?
Riding into Product Market Fit
An example from Clayton Christensen's The Innovator's Dilemma: Honda and motorcycles. Honda had been very successful with selling motorcycles in Japan and sent a high level employee to Los Angeles to open up the US market. And this employee quickly became despairing, as the small, light motorcycles that sold so well in Japan were completely uninteresting to US motorcycle fans who loved their powerful hogs. So he thought he would be returning home a failure, completely losing face.
So the employee tried to burn off his frustration by taking long rides on his inventory which would not sell. Rides into the chapparal of Southern California off-road. Get away from it all and try to escape from his unfolding career disaster.
Except, people saw him riding on trails and stopped to ask: where can I get a motorcycle like yours? This was a previously non-existent market, and one ill suited for traditional American monsters. And so Honda built a completely new market
Drop-letting Into Product Market Fit
Sid brought up a great example from genomics: 10X.
Originally 10X Genomics applied their droplet barcoding technology to generate genomic linked read libraries. It's an elegant concept: reads from the same input molecule could be grouped together to capture more structural and haplotype information. You could apply hybrid capture to these libraries and have linked read exomes. Maybe assemble very large metagenomes. The informatics to support this was a bit slow to build but still a darling of many bioinformatics labs. Except the product no longer exists.
Why not? Because it didn't turn out to be a big market. Not enough demand for linked reads, and the rise of true long read technologies offered a better way to get structural information and haplotypes. So when 10X had to reformulate their barcoding scheme after an IP lawsuit, they dropped the linked read product.
10X Genomics didn't disappear - they pivoted to single cell RNA-Seq. And that market proved to be much larger. It didn't hurt that they didn't have much competition early on to create single cell libraries from many thousands of cells. Plus they kept innovating on library subtypes - antibody and T-cell, 5' libraries, etc.
It also didn't hurt that sequencing costs kept dropping. After all, if you are capturing tens of thousands of reads per cell, and doing so for tens of thousands of cells, that product is a lot of sequencing reads. And the informatics community became very interested in the topic and started pumping out tools to use all this data. And biopharma companies saw single cell as an important component of discovery programs, and wanted many samples run through - far, far more than they ever would have ordered for linked genomic reads.
Which Way to Product Market Fit?
So if you are designing products, how do you find product market fit? How do you avoid costly flops like Juicero, the fresh juice appliance that turned out to be replaceable with a pair of hands? Or avoiding Revolocity, the genome sequencing workcell which belly flopped?
Clearly one important criterion is to have a high reliability product: both Mondrian and NeoPrep gained reputations for failing. If you don't have reliability early on - Oxford Nanopore struggled early in the MinION access program with shipping flowcells worldwide that still worked on arrival - then get serious about solving the problem. In ONT's case, it was tiny bubbles destroying the pores during shipping, so they identified the problem and then completely changed the design, construction, and fabrication of the flowcells to nearly eliminate the problem
Sid's Arima Genomics is one of three providers specializing in HiC kits, the others being Dovetail Genomics and Phase Genomics (Oxford Nanopore also has their Pore-C kits and various molecular biology shops sell kits) HiC was originally developed to resolve the 3D architecture of chromosomes, but that remains scientifically fascinating but not yet widely applied - a niche market.
HiC can also be useful for establishing genome contiguity over highly repetitive regions, but use there is being pressured by improved ultra-long nanopore protocols, and generating telomere-to-telomere genomes seems in any case to still be a low-level activity mostly used to generate an initial reference.
But, you can't ultra long read DNA unless you have ultra high molecular weight DNA to start with. The dominant means of storing clinical samples is Formalin Fixed, Paraffin Embedded (FFPE) - amazing for preserving cellular ultrastructures for decades of ambient storage, but horrible for DNA. DNA recovered from FFPE is typically hideously shredded, though it is possible to recover a tiny number of long DNAs from it. But HiC adapts well to FFPE - the first step in HiC is formalin fixation - and so can extract long range structural information from preserved tissue blocks. So the HiC companies see the clinical market as an opportunity to gain product market fit in markets much larger than that for chromatin structure or T2T genomes.
Of course, it won't be simple. There's still a wide range of possible stumbling blocks - threading regulatory requirements in diagnostics, bioinformatics which delivers to clinicians the information they find useful, getting reimbursement codes and convincing payors this is valuable.
What's Your Favorite Product Market Fit Story?
There are numerous other stories I know - and must be many, many fold more I don't. There's many ways to fail, so hard to succeed. For example, I can think of a number of products that were basically too late to the party. What Genapsys released with late in the 2010s would have been amazing in the early 2010s. There's also competitive issues - Ion Torrent would have made far more impact if Illumina hadn't launched MiSeq at the same time.
Please feel free to leave your favorite example in the comments or send me a private message. I don't have any illusions that there will ever be a simple recipe guaranteeing product market fit, but studying successes and failures might make product development a bit less prone to failure. And after all, we all like it when new products fit the markets we care about.
Great write up - discussed product market fit in my multi-omics class this year. Challenged students when attending meeting to ask their big wigs about it. It seems like most genomics companies make a lot of these classic mistakes. I think Ont is probably a great example in a few areas: correct me if I’m wrong but my initial recollection was that they were all about speed and native nucleic acids but with long reads as a secondary benefit - obv now length is the primary with a focus on genomes. I also think the folks adapting nanopore sequencing for small molecule (non nucleic acids) could end up being a great example if they somehow overturned the entire mass spec market.
ReplyDeleteAlso thinking about the history of column based purifications