This is a composite story based on real incidents at multiple companies. The specifics are fictionalized. The patterns are exact.

A SaaS company with 800 customers and $40K MRR received a support ticket: "I was promised feature X during the sales demo but it doesn't exist in the product."

The agent checked. The feature didn't exist. It was on the roadmap but not built. The sales rep had oversold. The agent, following protocol, responded: "That feature is currently in development. I don't have a timeline for release."

The customer was furious. They'd purchased a $2,000 annual plan based on the demo. They demanded a full refund. The agent escalated.

The escalation sat in a queue for 36 hours. The manager was in back-to-back meetings.

The Cascade

During those 36 hours, the customer:

Posted on Twitter: "Bought [product] because they demo'd feature X. Turns out it doesn't exist. Been waiting 2 days for a refund. Avoid."

The tweet got picked up by a tech newsletter with 30,000 subscribers.

Two other customers who'd also been sold on the same non-existent feature saw the tweet and realized they'd been misled too. They requested refunds.

A Hacker News thread appeared: "Is [product] vaporware?" It reached the front page. Eight more customers who were on the fence about renewing saw the thread and cancelled.

In 48 hours, the company lost 11 customers. Total revenue impact: about $22,000 in immediate cancellations plus an estimated $15,000 in deals that went cold because prospects saw the HN thread.

$37,000 in damage. From one ticket that sat unanswered for 36 hours.

What Went Wrong

The sales team oversold. This is the root cause. Demoing features that don't exist (or exist only on a roadmap) creates a ticking time bomb that eventually detonates in support's inbox.

The escalation process was too slow. 36 hours for a customer who'd been sold a non-existent feature and wanted their money back. By hour 12, the customer had given up on the support channel and gone public.

Nobody had authority to act fast. The agent couldn't process a $2,000 refund. The manager was in meetings. No one else had authorization. The customer was waiting for a decision that nobody was making.

There was no crisis detection. A customer threatening to go public (implicitly or explicitly) should trigger an immediate response. The ticket wasn't flagged as high-risk because nobody was looking for risk signals.

What Should Have Happened

The agent should have had authority to issue the refund immediately. A $2,000 refund is significant but not catastrophic. The cost of not refunding (lost customer, public complaint, cascade) was 18x the cost of the refund.

The escalation should have gone to a person, not a queue. "Manager will review" means nobody is responsible. "I'm escalating to Sarah, who will respond within 2 hours" means someone specific is accountable with a specific timeline.

The response should have acknowledged the sales error. "You're right. The feature you were shown doesn't exist in the current product, and you shouldn't have been told it did. That's our mistake. I'm processing a full refund now, and I'm personally sorry for the frustration this caused."

That response costs the company $2,000 (the refund). The alternative cost was $37,000 in cascade damage. The math is obvious in hindsight.

The Crisis Protocol (For Next Time)

Every company should have a crisis ticket protocol. Here's what the company built after their incident:

Any ticket mentioning "refund" + "promised" or "demo'd" gets auto-flagged as potential sales misalignment. Routed to a senior support person within 1 hour.

Any ticket where the customer mentions Twitter, Reddit, HN, a review site, or "going public" gets routed to the founder or VP within 30 minutes. Not the support queue. A person.

Agents have authority to process refunds up to $500 without approval. For amounts over $500, they can commit to the refund ("I'm approving your refund now") and process it with manager confirmation after.

AI classification detects high-risk language patterns. Supp's classifier identifies intents like "demand refund," "was promised," "going to dispute," and "posting about this" and assigns elevated priority scores. These tickets skip the regular queue entirely.

The Sales-Support Connection

The root cause (sales overselling) is a cross-functional problem that support can't fix alone. But support can document it.

Every ticket where a customer says "I was told [thing]" and the thing isn't true is a sales process failure. Track these. Share them with sales leadership monthly. "This quarter, 7 customers reported being sold features that don't exist, resulting in $14,000 in refunds."

That data changes sales behavior faster than any training program.

The Lesson

The company survived. They learned. They built better processes. But the six months of growth they lost (from churn, from cold deals, from the lingering HN thread that showed up in Google results) took a year to recover.

One ticket. 36 hours of silence. $37,000 in damage.

The cost of responding to the first ticket within 2 hours with a full refund and a genuine apology: $2,000.

Speed in support isn't about efficiency metrics. Sometimes it's about survival.