Knowledge Base Governance That Cuts Support Volume Across Email and Chat

We treat our help content like a product that needs pruning, not a pile that only grows. What to create comes straight from support volume. If the same client question lands in email or chat more than a few times in a month, that's the next article or Loom we make. For updating and retiring, we run a quarterly review where anything that hasn't been viewed, or has gone stale against a recent platform change, gets rewritten or removed.

The rhythm that lifted quality most was a simple handoff. Whoever answers a tricky question well writes or updates the doc that same day while the explanation is fresh, then a second person reviews it before it goes live. Repeat questions fell away once the answers were genuinely findable.

At Santa Cruz Properties, our "knowledge base" is really the library of answers our buyers ask about owner-financing, lot availability, and loan servicing. Here's how I decide what stays, what gets rewritten, and what gets retired.
First, I let the inbox and chat logs tell me what to write. Every week I pull the top questions coming into our Edinburg office across email, chat, and phone, things like "Do you run credit?", "What's the down payment on a lot in Falfurrias?", "How do I make a payment on my land loan?" If a question shows up three or more times and we don't have a clean article for it, that's the next thing I write. If we already have one and people are still asking, the article isn't doing its job and needs a rewrite, usually shorter, with the answer in the first sentence.
Retirement is just as important. Lot inventory changes, pricing changes, and county-specific details in Hidalgo, Cameron, or Starr shift over time. Anything tied to a sold-out tract or outdated closing step gets archived so buyers aren't reading stale info and calling us to "confirm."
The review rhythm that moved the needle for us is a monthly sit-down between marketing and our loan servicing and sales teams. Sales tells me what prospects keep asking on the lot tour; loan servicing tells me what existing landowners call about after closing. I bring the analytics, which articles get opened, which ones lead to a ticket anyway. We pick three to update, one to retire, and one new article to publish before the next meeting.
The handoff that raised quality most was simple: every customer-facing teammate can flag an article as "confusing" in one click, with a sentence on why. That feedback loop turned our FAQ from a marketing page into a real self-serve tool, and it cut down repeat questions on financing terms noticeably.

At Davila's Clinic in Weslaco, our "knowledge base" is really our patient education library plus the FAQ content we publish for the Rio Grande Valley families we serve. Here's how we decide what lives, gets refreshed, or gets retired.
We start with the front desk and phone log. Every week our team tallies the questions patients ask most, "Do you take walk-ins after 5?", "What's a telemedicine visit like?", "How do I prep for my physical?", "What does chronic disease management actually include?" If a question shows up more than a handful of times across email, chat, or the phone, that's our signal to create or rewrite an article. If an article hasn't been clicked or referenced by staff in a quarter, we either retire it or fold it into a stronger piece. Clinical content gets reviewed any time guidance changes, outdated info is worse than no info in healthcare.
The review rhythm that moved the needle for us is a simple Friday huddle. Justin Davila, our FNP-BC, sits down with the front-office lead and me for 20 minutes. The front desk brings the top five repeat questions from the week. Justin signs off on the clinical accuracy, I rewrite it in plain, bilingual-friendly language, and it's live by Monday. That single handoff, patient-facing staff flagging confusion, clinician validating the answer, marketing translating it, cut down a real chunk of repeat calls about hours, telemedicine setup, and what to bring to a wellness check-up.
The bigger lesson: self-service only works when patients trust the source. We build that trust by being specific (our actual hours, our actual address at 412 E 18th in Weslaco, our actual services) and by never letting an article drift from what happens in the exam room. If the website says it, the staff lives it. That alignment is what turns an article into a real ticket-deflector.

Tickets are the signal that it's time to update the article. If the same question keeps coming in, that's a telltale sign that there is a missing or broken article. At Slite we cross-reference support patterns against what's in the KB and ask: does this doc still match what the team is actually doing? If not, update it. If nobody's ever asked about it and it's 18 months old, archive it. The question isn't "is this doc old?" It's "is this doc wrong?"

What worked to improve the article quality the most was: one named owner per doc, not a committee. When everyone owns it, nobody does the updating. We set a verification cadence per article (high-traffic ones get checked more often) and the review becomes a 5-minute task instead of a quarterly cleanup nobody volunteers for. The moment a doc is visibly stale, people stop trusting the whole KB and go ask in Slack or support forums instead. One wrong guide poisons the well.

We treat the knowledge base like a search portfolio instead of a document archive. We review each article based on demand from support tickets, chat logs, site searches, and common queries. We also look at friction, which shows where customers get stuck or repeat steps. We study decay, which appears when an article gets views but no longer solves the problem.

This helps us decide whether to create, update, combine, or remove content. We also track language drift as customer words change over time. When titles and headings do not match how people ask for help, self serve use drops. Content is often removed because of unclear wording, duplicate paths, or outdated assumptions.