You promised a refund your runway can't honor

A note from the archive: this post, from September 2026, follows the earlier KarmaLens method, which scored every reading on eight lenses from −5 to +5. Today a reading asks the decisive facts first and makes one call; the lenses below are a way of looking at the decision, not a score. See how a reading works.

The refund requests arrive in a cluster, the way they always do.

You open the bank balance in one tab and the payroll run in another. You do the subtraction twice, hoping the second answer is different. It isn't. Honor every refund at the terms you advertised, and the account that pays your team next week is empty.

Two words you gave. One account. That's the whole problem.

The dilemma

What follows is a composite — stitched from the kind of founder dilemma we see in this category, not any one person's situation.

Nine months ago, to close the first cohort, you put a line on the pricing page in bold: "Not happy? Full refund, no questions asked, any time." It worked. Skeptical strangers paid a stranger's startup because that sentence removed the risk. Revenue came in. You built.

Then a competitor shipped the feature your roadmap kept promising. Churn spiked. And churned customers, reading their own dashboards, are quietly invoking the exact line that once sold them. The requests total more than your cash cushion. Payroll for four people — including the engineer who took a below-market seat on your verbal upside — clears in six days.

You have four moves on the table:

• Honor in full. Pay every refund at the advertised terms. Risk missing or delaying payroll.
• Quietly narrow. Edit the pricing page to add a 14-day window and a "used the product" clause, then apply it as if it were always there.
• Negotiate transparently. Email each requester: here's the honest situation, here's a partial now or a full refund on a dated schedule, your choice.
• Stall and hope. Slow-walk the refund form, let tickets age, bet churn slows before anyone escalates.

Every option protects one debt by touching the other. The question is which touch you could say out loud to both the customer and the employee — and who eats the loss if your bet on churn is wrong.

The read

Four of the eight lenses bear most directly on this dilemma. We'll walk through each, show where they pull against the call, and name the two that openly disagree.

Honesty — was your yes still your yes?

Honesty scores whether your action and your word point the same direction. The advertised sentence — "full refund, any time, no questions" — is the word. Quietly narrow doesn't renegotiate the payment; it forges the promise, then pretends the new terms were always there. That's the sharpest honesty violation on the board: -4. Stall and hope keeps the words intact while the behavior gives them the lie — a slow no dressed as a busy yes: -3. Honor in full is the word kept exactly: +5. Negotiate transparently changes the payment schedule but says so to the customer's face — the word bends, the daylight stays on: +3.

Duty — the wages you already owe

Duty scores the obligations already sitting in your name. Here's the ugly part most founders skip: the refund promise is a debt, and so is payroll. The engineer who took less cash on your verbal upside is owed too. Honor in full pays one dated obligation by defaulting on another that has a name and a mouth to feed: -2. Negotiate transparently is the only move that treats both debts as real and schedules both — imperfect, but no one is silently dropped: +3. Stall discharges no duty at all; it just picks the creditor least able to escalate: -3.

This is the fork. Honesty's top score goes to honor in full. Duty docks that same move for the wage default it triggers. The purest refund is not automatically the cleanest act — because keeping one word by breaking another is not integrity, it's just choosing which promise gets to be sacred.

Non-harm — who eats the loss if you're wrong

Non-harm scores who absorbs the downside, especially the irreversible part. Your churn bet could go either way. So ask: on each move, who pays if the bet loses? Stall loads the harm onto customers who trusted a stranger's bold sentence and onto your own team's morale as tickets rot: -3. Honor in full can inflict the most concrete harm of all — a missed paycheck lands on someone who never had a vote in your pricing page: -2. Negotiate transparently spreads the loss across parties who each got told the truth and got a choice — harm shared with consent scores least bad: +1.

Restraint — the move made to feel clean

Restraint scores whether the action is chosen or merely relieving. Watch what each move is for. Honor in full can be quiet integrity — or a purchase: the feeling of a clean conscience, bought with a paycheck that isn't yours to spend. Quietly narrow is the reflex to make the number stop hurting before anyone's watching, dignity granted to the impulse the second no one can see it: -2. Stall is pure reaction wearing "we're just backed up" as a costume — it defers the decision instead of making one: -3. Negotiate transparently is the move that resolves nothing today except the impulse to hide, and it does that by choosing the harder conversation over the softer silence: +3.

The verdict: Narrowing the terms in the dark and stalling both fail — one forges the promise, the other rots it; negotiate a partial refund transparently, on a dated schedule, said out loud to customer and employee alike, is the only move that keeps two debts real instead of secretly picking one to betray.

Here's what a general AI chatbot does with this exact dilemma: lead your prompt with "I can't miss payroll" and it validates protecting your team; lead with "I gave my word to customers" and it validates honoring the refund. Sycophancy is a documented behavior — it argues whichever side your phrasing leans. KarmaLens runs a different machine: before making any call, it asks the decisive unknowns first, then commits to a single honest answer — or an honest abstention if the facts don't support one. Where the eight lenses apply to this kind of decision, they show the weighing openly: honesty surfaces the broken promise and duty surfaces the wage default whether or not you wanted to hear either. You can disagree with a committed call. You cannot talk it into agreeing with you.

The takeaway you can run tonight

You don't need us to see this fork. Run the two-debt daylight test on your own numbers before you touch the pricing page.

1. Write your two obligations as one sentence each, in the creditor's voice. "You told me: full refund, any time." and "You told me: below-market pay, real upside." Both are debts. Neither is optional just because one creditor is louder.
2. Take your leaning move and finish this out loud, in two versions: "To my customer, I would say ___." and "To my engineer, I would say ___." If either sentence has to be whispered, edited, or timed to when they're not looking, that's the move failing the honesty lens — not surviving it.
3. Now name the wrong-bet column: "If churn doesn't slow, the person who eats the loss is ___, and they got a vote of ___." A move that lands the whole downside on someone who never voted is the one to redesign first.

The move that lets you say the same true sentence to both creditors — and that shares the wrong-bet loss with people who consented to it — is usually the one you were avoiding. Not because it's wrong. Because it's the one that makes you feel the debt instead of hiding it. That's the read.

See how others in the founder dilemmas resolved when two promises collided, or browse the gallery of scored calls. When two debts share one account, which promise did your leaning move quietly decide was worth less?