I would have expected Emerald Cloud Lab or similar competitors to go a lot and be successful over the last five years. As far as I know, like Emerald Cloud Lab only had modest growth and there aren't competitors who grew strongly. Outsourcing to cloud labs seems like it allows the laberatory to have benefits of scale and virtualization that drives down costs and is easier to use then working in a wet lab. Is there something holding back this trend that I'm not seeing? Alternatively, what's going on?
Ask Related Question
Disclaimer: I do not work in a lab, and never have beyond a short stint as a research assistant in undergrad.
That being said, I can think of several reasons. In no particular order:
- Habit: researchers are used to what they have been doing, and do not want to change.
- Control: they are a hands-on scientist who likes to roll up their sleeves.
- Secrecy: if a cloud lab does the experiment, then a cloud lab has the data.
- Not Free: why spend money on a cloud lab when they have already-paid-for equipment and interns in their lab?
- Training: most of the actual routine work is done by grad students; this is a critical part of their training as scientists. If experiments are outsourced, how will they learn to use the equipment and to design experiments of their own?
- Cognitive burden: another tool chain to remember? It's not even a python script!
- Publication bias: I have read several accounts of papers being rejected because they used the wrong code in their analysis; the reviewers preferred R or Python. Do any journals accept a description of a proprietary software workflow from a single company in the methods section?
- Experimental design: things have improved from a replication standpoint since the replication crisis, but I don't see much movement on the bias towards novel results. The scalability argument Emerald Cloud Labs is making doesn't appeal as much if the goal of an experimenter is to design the most novel possible experiment.
- Inadequate discovery equilibrium: this is essentially another facet of the previous point, but researchers may assume that because the experiments are so easy to replicate and scale that their efforts will not be sufficiently rewarded, even if they can think of good experiments to run.
- Too few non-academic researchers: business investment in R&D has plummeted from its previous levels, as most corporations moved to shift investment into shorter-payoff projects. They are likely not even evaluating this kind of product anymore.
- Competition from computers: a significant chunk of the big data/machine learning revolution is going into producing better models and simulations; this directly competes with the repeatability and scalability pitch that cloud labs are making. Come to think of it, the best use might be validating or building a model or simulation.