"We could build this in-house." Usually correct. Usually incomplete.
"We could build this in-house." Usually correct. Usually incomplete.
I hear it in many calls with interested pharma and chemical companies. And they're right - they usually can. These are organizations with serious data science teams and people who understand their own chemistry better than any vendor ever will.
If you build it yourself, you get something shaped exactly around your workflows. Your data structures. Your naming conventions. Your particular way of running a campaign. No compromises, no feature requests sitting in someone else's backlog.
That's real, and I'm not going to pretend otherwise. I'm also obviously not a neutral party in this conversation - but I've seen enough internal tools work well to know this isn't a one-sided argument.
Here's the part that tends to get underestimated:
- Your best people build infrastructure instead of doing science. The data scientists writing your optimization backend aren't working on the problems only your organization can solve. That's a real trade-off, and it rarely appears in the business case.
- It takes longer than the estimate. It always does. A team that starts in January isn't running real campaigns that year. Much of what a vendor already knows about user workflows has to be discovered from scratch.
- It survives exactly as long as the person who built it stays. Internal tools are built by one or two brilliant people who document maybe a third of what's in their heads. When they leave, the tool doesn't break - it degrades. People quietly work around the parts that stopped working, and eventually it's replaced by a spreadsheet.
Most internal scientific software doesn't fail technically. It just stops being understood.
So the question isn't "can we build this?" You almost certainly can.
It's: is this our competitive advantage, or is this infrastructure?
A simple test - if a competitor had the identical system tomorrow, would you have lost anything? If not, it's infrastructure.
Almost nobody builds their own ELN anymore. Nobody runs their own email server. Not because those things are unbuildable, but because the organization decided its scarce engineering talent belonged closer to what makes it distinctive.
That's the actual decision. And it deserves a proper conversation, not a default answer.
