Showing depth without lecturing

The idea

When someone asks you to explain something you know well, you have a ladder in front of you. Rung one is a single sentence a smart outsider could repeat back. Rung two is a paragraph a peer would nod at. Rung three is one specific place the thing broke on you — the rung nobody can climb from reading a blog post.

Climbing earns credibility. Climbing also spends attention. Depth offered past the interviewer's interest doesn't read as expertise; it reads as poor judgment about what the listener needs. So the ladder has a top, and the top moves depending on who is listening.

The ladder, live

How it works

1. Rung 1 — one sentence, no jargon. Name the thing, the problem it solves, and the outcome, in a form a smart outsider could repeat back to a colleague. If you can't do this, you don't own the concept yet.
2. Rung 2 — one paragraph a peer would respect. The mechanism plus the choices inside it: what you keep, where you keep it, what you rejected. Peer vocabulary is welcome here; this is where you sound like a practitioner rather than a reader.
3. Rung 3 — one specific where-it-breaks detail. A failure mode with a consequence, how you detected it, and what you changed. This is the rung that proves you operated the thing, so it must be true and it must be yours.
4. Hand the wheel back, then stop. "I can go into the storage layer or the client retry policy — which is more useful?" Depth invited is not depth imposed. Waiting for the invitation is the seniority signal.
5. Re-read the room every rung. Note-taking stopping, "sure, sure", a glance at the clock, a follow-up on something else entirely — all of these mean the top of the ladder just moved down.

The two meters in the visual are a simple model, not a law. Attention and credibility are multiplied rather than added, because either one at zero kills the answer: nobody is convinced by something they stopped following, and nobody is impressed by something that never got past the headline.

net signal = listener attention  ×  expert credibility  ÷ 100

listening: peer engineer
  rung 1   95 × 30 / 100 = 28.5  ->  29
  rung 2   86 × 72 / 100 = 61.92 ->  62
  rung 3   72 × 94 / 100 = 67.68 ->  68   <- stop here
  rung 4   34 × 96 / 100 = 32.64 ->  33   <- the fall

what each climb costs and buys (peer engineer)
  1 -> 2   attention  -9   credibility +42   net +33
  2 -> 3   attention -14   credibility +22   net  +6
  3 -> 4   attention -38   credibility  +2   net -35

listening: hiring manager (non-specialist)
  rung 1   98 × 30 / 100 = 29.4  ->  29
  rung 2   74 × 72 / 100 = 53.28 ->  53   <- stop here
  rung 3   46 × 94 / 100 = 43.24 ->  43
  rung 4   14 × 96 / 100 = 13.44 ->  13

Notice the shape rather than the digits: rung 3 buys much less credibility than rung 2 did, and rung 4 buys almost none while spending most of the attention you had left. That is the whole lesson — diminishing returns on depth, accelerating cost on attention.

When to use it

The trade-off: the ladder is a shape, not a script. Three rungs delivered as an uninterrupted ninety-second monologue still fails — the point of each rung is that it is a place you can stop.

Watch out for

• An invented rung 3. Borrowed war stories collapse on the first follow-up: what did the graph look like, what did the fix cost, who paged you. If you haven't operated it, climb only to rung 2 and reason forward out loud instead.
• Starting on rung 2 with no landing. Opening in jargon leaves the non-specialist on the panel with no handhold, and they are often the one writing "communication: unclear" in the debrief. One plain sentence first, always.
• Trivia dressed as depth. Naming an obscure flag or a config default is not rung 3. A real break has a consequence, a detection path, and a decision. If nothing changed as a result, it isn't the rung.
• Missing the hand on the wheel. "Right, right — and how would you test it?" is a redirect, not encouragement. Stop mid-sentence, answer what was asked, and leave the rest as an offer.
• One ladder for every listener. The same rung 3 that delights a staff engineer buries a founder. Pick the height from the room in front of you, not from how much you happen to know.
• Treating rung 4 as forbidden. If they ask for the storage internals, the internals become rung 3 and you should go there gladly. The mistake is never the depth itself — it's the depth nobody requested.

Worked example

Panel: a hiring manager who ships product, plus a staff engineer who owns payments. Question: "Walk me through how you'd make a payment endpoint safe to retry."

You land rung 1 in a sentence — the client stamps each payment attempt with an ID, so a retry after a timeout charges the customer once, not twice — and you can see the hiring manager relax, because she now has the whole idea. You climb to rung 2: one key per checkout rather than per HTTP call, a durable record written before you call the processor, a fingerprint of the request body so a reused key carrying a different amount is rejected instead of silently replayed, keys scoped per merchant and expired after a day. The staff engineer starts writing.

Then one rung more, and only one: the write-then-call gap. You commit the in-flight record, call the processor, and die before storing the result — the key now reads in-flight forever, retries get a 409, and the money may already have moved; so you pass the same key downstream as the processor's own idempotency key and run a sweeper that asks the processor what really happened and heals the record. You add the human half — an iOS client that regenerated the key on app resume, found by joining double-charge refunds against client key age — and then you stop and ask: "I can go into how we stored and expired the keys, or the client-side retry policy — which is more useful?"

That question is the move. It says you have another rung available and the judgment not to climb it uninvited. The staff engineer picks one, and now every further detail you give is depth they asked for.

Check yourself