The four stages of enterprise software
How we get from an idea to a standardized tool
My next blog post is going to look at software for collecting lab data, so as a warm-up I wanted to discuss a pattern I’ve seen in the development of enterprise software.
This pattern is triggered whenever a shift in the context/environment creates either a new opportunity or a new need within organizations. ELNs and LIMS are near the end of this process. Data Catalogs and Data Lakes are entering the finals stages. MLOps software is closer to the beginning. For software connecting lab data to data scientists, I think we’re at the very start.
Here’s what it looks like:
When a new idea becomes suddenly relevant in a particular context, the teams working in those contexts notice it first. Since there aren’t off-the-shelf options, they start building their own internal version of the tool/software.
Eventually, someone notices that all these different organizations are building essentially the same tool, so they create an off-the-shelf version - either starting from scratch or by repackaging an internal version.
Because the off-the-shelf versions aren’t fully baked yet, and don’t exactly fit most organizations’ needs, the home-brew versions continue to be developed while the commercial versions iterate and generalize. Most of the teams that adopt the off-the-shelf software are late-comers or large enterprises that are more biased towards buy over build.
Eventually, the off-the-shelf options begin to beat the functionality of the home-brews, and address the general problem well enough to cover what most of the market is looking for. Building your own version becomes a clearly bad idea, so most organizations migrate their home-brew systems to off-the-shelf options and almost never start building their own from scratch.
Stage three is the longest and most painful part of the process, both for the vendors who are trying to stay afloat and for the internal teams agonizing over the build vs buy decision.
In my blog post in two weeks, I’ll describe a need that has come up in the past few years, as ML and data science have become a larger part of biotech research, and for which I think we’re mostly still in stage 1. Stay tuned!