It happens on a sales call, about twenty-five minutes in. You've explained the same thing forty times this year, and this time — because the person across from you isn't getting it — you invent a sentence you've never said before.
"Your hero image isn't competing on quality. It's competing for a half-second against nineteen thumbnails, and right now yours is the fourth white bottle in the row."
And they stop. And they say: oh.
Two weeks later you write a post about the same topic. You publish the conclusion — differentiate your creative, thumbnails matter, here's a framework. The sentence that actually moved someone is nowhere in it. You published the destination and threw away the road.
That sentence had a home in none of your content systems. The proof bank catches numbers. The story bank catches moments. The question log catches what your market asks. The objection ledger catches why they walk. Nothing in that stack catches the translation device — the thing you built on the spot to get an idea from your head into someone else's.
That's the explanation bank.
Why the best explanations are the ones you lose
Explanations evaporate for a specific reason: they're produced in speech, under pressure, in response to one person's confusion — and they don't feel like content while you're making them. They feel like talking.
Everything else in your capture stack has a moment where you notice it. A number arrives in a report and you think that's a proof point. A customer asks something twice and you think that's a post. But an explanation arrives mid-sentence while you're concentrating on the person, not the language. By the time the call ends you remember that it went well. You do not remember what you said.
Then, when you sit down to write, you reach for the version of the idea you've rehearsed — the tidy one, the one in your deck. That version is smooth, complete, and completely inert. It's a summary of a thing you understand rather than a device for making someone else understand it.
This is why so much founder content is technically correct and totally unpersuasive. The claim survives. The mechanism that made the claim land does not.
What actually goes in it (three types)
An explanation is not a claim. A claim is what you believe. An explanation is the thing that makes a person who doesn't already agree with you see it. Three types are worth capturing.
The analogy. You borrow a structure the reader already understands and map your idea onto it. "A SKU isn't a lottery ticket, it's a monthly subscription — you pay for it in forecasting, storage, reorders and ad budget every single month whether it sells or not."
The reframe. You rename the thing so its real nature becomes visible. "A coupon without an end date isn't a promotion. It's a permanent price cut that you also pay fees to run." Nothing new was reported. The category changed, and with it the decision.
The smallest concrete case. You shrink an abstract mechanism to one scenario a person can picture end to end. Not "attribution is complicated" but "a shopper asks a question on their phone, gets three answers, taps one, buys — and your report shows a direct visit."
All three do the same job: they move an idea from your experience into someone else's, using material the other person already owns.
The six fields
Keep it lightweight or you won't run it.
- The idea being explained. One line. "Why a stale hero image caps revenue rather than reducing it."
- The explanation, verbatim. Exactly as you said it, including the clumsy parts. Do not tidy this field. The tidied version is the dead version.
- Who you were talking to. Role and sophistication level. An explanation built for a $200K/mo operator and one built for a first-time seller are different tools, and using the wrong one is most of what makes content feel condescending or over-technical.
- What they said back. The evidence it landed. "So we've basically been paying rent on forty products that don't work" — that's a confirmed hit. If you can't fill this field, you have a phrase you liked, not an explanation that worked.
- Where it happened. Sales call, supplier meeting, podcast, a reply to a comment.
- Used yet. So you're not republishing the same device six times a quarter.
Optional seventh field, and it's the one that matures the system: where it breaks. Every analogy is wrong somewhere. Knowing where is what lets you use it confidently — and it's a second post.
Four capture triggers
You need tells, because you won't remember to look.
- You said "think of it like…" — you're about to build one. That phrase is the sound of a translation happening in real time.
- They interrupted you. People interrupt at the moment of understanding, not the moment of boredom. Something you said in the previous ten seconds did the work.
- You heard yourself say something new. There's a distinct feeling of surprising yourself mid-sentence. That's an explanation getting invented. It has a shelf life of about four minutes.
- Someone repeated it back later, in their own words. The strongest signal available. An explanation that gets re-transmitted by someone who isn't you is the only one you have proof about.
Capture takes sixty seconds after the call. A voice memo is fine. The rule is speed and fidelity, not polish — you are preserving evidence, not drafting.
The filter: not every analogy earns a slot
Founders who like this system immediately over-collect. Three tests keep the file useful.
Does it survive its own edges? An analogy that collapses the moment a smart reader pushes on it costs you more than it buys. If someone can answer "well, no, because…" in one sentence, you have a nice line, not a working device.
Does it change a decision? Some explanations only make people feel clever. The ones worth keeping end with the reader doing something differently on Monday. If you can't name the action, it's decoration.
Is it yours? Stock business metaphors — icebergs, Swiss army knives, building the plane while flying it — are dead on arrival because everyone has heard them and none of them are welded to your experience. The explanations that work are the ones grounded in something specific you've actually done. That's also what makes them impossible to counterfeit.
How an entry becomes posts
The mistake is dropping the explanation into a post as a garnish. It's the spine, not the seasoning. Build the post as a delivery vehicle for it: set up the misunderstanding, deliver the explanation, then prove it with one number and one example.
One good entry yields three posts, spaced weeks apart:
- The explanation itself, with the receipt that makes it credible.
- Where it breaks — the boundary. This is the post that separates you from everyone repeating your line, and it's usually the more sophisticated piece.
- The misapplication — the case where someone took the idea and ran it off a cliff. You'll have this material within a month of publishing the first one.
Why this is the hardest asset for AI to replicate
You can ask a model to explain almost anything, and it will produce a competent analogy on the first try. It will also produce a generic one — a comparison assembled from every explanation of that concept ever written. Readers recognize this instantly, even if they can't name what they're recognizing. It reads as filler.
The explanations in your bank are welded to a specific account, a specific number, a specific afternoon. They're not retrievable from the general corpus, because they were manufactured live in a room, once, for one confused person. A model can write around your idea. It cannot produce the sentence you invented on a Tuesday to get through to a skeptical operator with a returns problem.
That's the whole moat, and it costs sixty seconds a call to bank.
How you'll know it's working
The metric isn't reach. It's hearing your own language come back at you. A prospect opens a call using your phrase. A competitor's post uses your reframe without attribution. Someone DMs to say a colleague sent them "the subscription thing."
That's positioning arriving. An owned phrase does work in rooms you're not in — which is the entire objective, and the reason a strong explanation outperforms a strong statistic over a twelve-month horizon. Statistics get quoted. Explanations get used.
Run it for a quarter and you'll have fifteen to twenty entries, of which maybe eight are genuinely good. Eight durable explanations is more than most operators build in a career, mostly because they invent them constantly and write down none of them.
FAQ
How is this different from the story bank? The story bank holds what happened — a moment with a beginning and an end. The explanation bank holds a device for making an idea understandable. A story can carry an explanation, but the two get used differently: stories build trust, explanations transfer understanding. Keep them separate or the file becomes a diary.
Isn't this just inventing jargon? No — jargon compresses meaning for insiders. An explanation expands it for someone standing outside. The test is simple: if a smart person outside your niche needs a definition to follow it, you made jargon. If they follow it immediately and feel like they learned something, you made an explanation.
What if the analogy came from someone else? Use it and say where it came from. Attribution costs one clause and buys you the reputation of someone who reads. What you can't do is silently adopt another operator's signature reframe — that's the fastest way to get noticed by exactly the wrong audience.
How many entries before it pays off? Faster than the other banks, because the first entry is publishable immediately. You already have three or four good ones in your head right now from the last month of calls. Write those down before you build the file — that's your starting inventory.
If you're an ecommerce founder producing this material constantly on calls and losing all of it by Friday, that's not a discipline problem — it's a missing capture step. We build the systems that catch it, and then we write from them.