Documentation Debt: The Interest You're Paying and the Pay-Down Plan

Watch what happens when someone on your team hits a question they can't answer. They ask in chat. Someone who knows — usually the same someone — types out the answer in ninety seconds, the asker says thanks, and everyone moves on. It feels like efficiency. It's actually a loan. The answer existed for exactly one person, for exactly one moment, and the next person with the same question will take out the same loan from the same lender.

That's documentation debt: the accumulating gap between what your team collectively knows and what your team has actually written down. Like its financial cousin, it isn't inherently bad — early on, borrowing is rational; you're moving fast and half your processes haven't stabilized enough to be worth writing down. But debt carries interest, and the interest on undocumented knowledge is uniquely sneaky because it never arrives as a bill. It arrives as texture: the fourth repeat of the same Slack answer, the new hire's sixth week of asking, the deploy nobody dares run because the person who understood it is at a wedding.

What the interest actually costs

Three payments, all recurring. Repeat questions are the visible one — your most knowledgeable people taxed a few minutes at a time, all day, precisely because they're the most knowledgeable. The tax reads as "being helpful," which is why nobody books it as cost. Slow ramps are the second: every undocumented process turns onboarding into archaeology, and a new hire who should be productive in four weeks takes ten, learning by interrupting. And concentration risk is the one that keeps owners up at night once they see it: knowledge that lives in one head is an outage waiting for a vacation, a resignation, or a bus. The debt metaphor is honest here too — a little spread across many heads is normal; a lot concentrated in one is a balloon payment with a due date you don't control.

Size the debt before you pay a dollar

The instinct, once the problem clicks, is a documentation sprint: "everyone writes everything down this month." Resist it. Sprints produce a wall of hastily written articles nobody maintains — you've met this wall; it's called the knowledge-base graveyard. Debt gets paid down by triage, not by enthusiasm.

Spend one week building the inventory. Two sources tell you everything. First, the question log: for five working days, every recurring question that gets answered live — in chat, in meetings, over shoulders — goes on a list. One line each, no ceremony. Second, the dependency list: each person names the three processes only they can run. That's it. At the end of the week you have the debt ledger, and it's usually twenty to forty items — not the four hundred a "document everything" sprint would have invented.

Now score each item on the two axes that matter: traffic — how often is this needed — and risk — how bad is it if the knower is unavailable. High-traffic items are costing you daily interest. High-risk items are the balloon payments. High-both is your top of the queue, and there are rarely more than eight of them. That short list is the 20% of the debt generating 80% of the interest.

The pay-down cadence

Two articles a week. That's the whole program — not because two is magic, but because two is sustainable, and a pay-down plan you abandon in week three is worth less than a modest one you run for a year. Each week, take the top two items off the ranked ledger and write them properly: the SOP format your team will actually follow — trigger, steps, judgment notes, owner — not a transcript dump. Assign each article an owner and a review date at birth; an article without both is future debt wearing a present-tense costume. In a quarter you'll have cleared the high-both queue, and the texture changes noticeably — fewer interruptions, faster ramps, calmer vacations. The ledger keeps working after the backlog clears: new recurring questions go on it, get scored, and enter the same queue.

Stop the new borrowing

Paying down old debt while taking out new loans is running on a treadmill, so the second half of the program is a single habit change at the moment of answering. When the recurring question arrives in chat, the knower's move is no longer "type the answer." It's "answer once, capture once": write the answer as a draft article — even three rough sentences — and reply with the link plus any personal note. Same ninety seconds, permanently different economics: the next asker finds it instead of asking it. The habit feels slightly slower per incident and is enormously faster per quarter, which is exactly the trade a debtor climbing out has to internalize. One caveat from the trenches: only recurring questions earn capture. Documenting one-off curiosities is how graveyards start.

Where the tool fits

The failure mode of every pay-down plan is friction at the capture moment — the knower is busy, the blank page is unappealing, and "I'll write it up later" is the lie that restarts the borrowing. That's the moment KnowledgeByDesign is built for: it turns a rough chat answer or a two-minute voice note into a structured draft article, keeps the owner and review date attached so nothing enters the base as future debt, and — because findability is the other half of the bargain — makes the article surface when the next person searches the way they actually phrase things. The judgment of what's worth documenting stays with your team. What gets removed is the blank page standing between a given answer and a kept one.

The bottom line

You never chose to take on documentation debt — nobody does. It accrues by default, one helpful chat answer at a time, and its interest compounds in the three currencies a small team can least afford: attention, ramp time, and resilience. The way out isn't a heroic sprint. It's a ledger built from one week of real questions, a ranked queue scored by traffic and risk, two properly written articles a week, and a capture habit that stops the new borrowing at the source. Six months from now the difference isn't a bigger knowledge base. It's a quieter one — fewer interruptions, faster new hires, and vacations that don't require a briefing.

— Tom