Welcome!

The Hidden Cost of Broken Knowledge

TL;DR; Knowledge that lives in one person's head — or in a Slack message, or in a Google Doc only three people know about — is broken knowledge. It costs you far more than you think. Most of the cost is invisible because no one attributes it to training.

This is one of the core ideas in Xavier's masterclass. Once you see it, you can't unsee it.

What broken knowledge looks like

Broken knowledge is the stuff that your team knows but hasn't written down in a way that anyone else can find.

A few common shapes:

  • The tribal elder. There's one person on the team who knows how the payment processor integration works. Every time something breaks, everyone pings them. When they're on vacation, something breaks, and no one can fix it.
  • The Slack archive. Someone explained the correct way to do X in a Slack thread a year ago. The explanation is perfect. It is also completely unfindable, because it's deep in a channel no one searches.
  • The half-written wiki page. There's a Confluence or Notion page for "how we do code reviews." It was written in 2023. It's half-right and half-wrong now, and no one knows which half is which.
  • The "ask me in person" policy. The onboarding process for new hires is "sit with this person for a day and absorb it." It technically works, but it costs one senior person a full day every time, and the quality depends entirely on whoever happens to be available.

Every one of these is broken knowledge. The organization has the information, but it isn't stored anywhere durable, findable, or maintainable.

The cost

The cost of broken knowledge is hard to see because it shows up in ways people don't label as "training problems."

Here are the most common symptoms:

  • Senior people are permanent help desks. They're constantly interrupted with questions they've already answered before, which means they produce less actual work.
  • New hires take forever to ramp up. Onboarding is measured in months instead of weeks, and the quality of the ramp-up varies wildly depending on who's mentoring them.
  • The same mistake keeps happening. Someone makes a mistake, figures out the fix, tells two people — and six months later someone new joins the team and makes the exact same mistake.
  • Handoffs are terrible. When someone changes roles or leaves, the incoming person has to reconstruct everything from git history and Slack archaeology.
  • Decisions get re-litigated. A team decides to do X for reason Y. Two years later, no one remembers reason Y, and the team starts debating whether to switch to Z. They might even decide to switch to Z, and then run into reason Y again, and switch back.

None of these are labeled "training problem" in a tracker somewhere. But they are, at root, all the same problem: the knowledge existed, but it wasn't captured in a way that lasts.

Why organizations tolerate it

Broken knowledge is tolerated because the cost is diffuse and the alternative is expensive.

The cost is diffuse: a senior person losing 20% of their time to answering questions feels like "just how the job is." An incoming hire taking four months to ramp up feels normal. Repeated mistakes feel like individual failures. No one sums up the costs and attributes them to the real cause.

The alternative is expensive: writing good documentation is tedious. Producing a video is a multi-day project. Keeping content current is a chore no one wants to own. So most organizations decide, often implicitly, that it's cheaper to tolerate the diffuse cost than to pay the concentrated cost of fixing it.

This calculation only makes sense if producing training is as expensive as it used to be. When the cost of producing a 5-minute video drops to an afternoon — which is what AI tools like Lupo make possible — the math flips. The diffuse cost of broken knowledge starts looking very expensive compared to the concentrated cost of fixing it.

The fix is behavioral, not technical

Lupo helps, but a tool alone doesn't fix broken knowledge. The fix is behavioral: you have to build the habit of capturing what your team figures out, the moment they figure it out.

A few practices that work:

  • When you answer the same question twice, write it down. Don't wait for the third time. Open a library in Lupo, create a content item, paste in your answer, and send the learner to the link instead of retyping it.
  • When a senior person figures out something non-obvious, record it. Five-minute Lupo video. Ten-minute script. Done before anyone forgets the context.
  • When you onboard a new hire, turn their questions into content. Their confusions are tomorrow's FAQ. Capture them while they're fresh.
  • Retire content that's out of date. Broken knowledge is worse than no knowledge. If a doc is wrong, fix it or delete it — don't let it rot in place.

The masterclass has a lot more on how to actually make these behaviors stick in a team. This page is just the sketch.

Where to go next

  • Foundations of Adult Learning — once you decide to capture knowledge, how do you make the content actually teach someone something?
  • Go Deeper — the full masterclass goes into much more detail on diagnosing and fixing broken knowledge.