Five lessons, about fifty-five minutes. Everything in module 2 assumed the queue was manageable. This module is for when it isn't calm — the customer who is genuinely, sometimes justifiably, angry, and what you actually say to keep them without giving away more than the brand can afford to.
An angry message rarely arrives out of nowhere. Almost always, something happened before it that you didn't see: an earlier reply that felt dismissive, a second problem after the first was supposedly fixed, or a wait that outlasted the customer's patience. Understanding that the anger has a history, even when you can't see it, changes how you read the message in front of you.
Three complaints in support work are almost never actually about the thing named in the subject line.
"This is broken, again." Often not really about the product defect — about not being believed the first time, or not being told it was a known issue. "I want a refund, now." Sometimes genuinely about the money, but just as often about wanting to feel in control of a situation that's felt out of their hands. "Nobody is responding to me." Rarely just about speed — usually about feeling unimportant to a brand they've spent money with and trusted.
"This is the SECOND bottle that's arrived with a broken pump. I don't understand how this keeps happening."
Surface reading: a product defect complaint — offer a replacement. Reading underneath: a person who told you this before, was presumably given a fix, and is now watching that fix fail — which is really a complaint about not being believed or properly followed up, wearing a product complaint's clothes.
The reply that only replaces the pump answers the surface. The reply that also says "I can see this is the second time — I'm sorry the first fix didn't hold, and I want to make sure this one actually does" answers what's underneath, and is the one that keeps the customer.
Take an angry message — real or made up — and write one sentence on what it might really be about, underneath the surface complaint.
There is a reliable shape to a reply that calms a genuinely upset customer down, and it is not "I'm sorry you feel that way" — a phrase that sounds like an apology while actually blaming the customer's feelings rather than the situation.
"I can see your order arrived five days late and the pump was broken on top of that — that's genuinely not the experience we want anyone to have. I'm sending a replacement today with express shipping, no charge, and it'll be with you by Thursday. I'll check it myself before it goes out."
A timeframe or gesture you can't actually deliver buys ten calm minutes and costs a much angrier customer later, when it doesn't happen. Only commit to what you can genuinely stand behind — this is the same discipline module 8's "the default is a real commitment" will apply to founders in Course 01, and it applies here too.
Take a real or made-up angry message and write a reply using the three-move shape. Time yourself — the goal is under five minutes once the shape is familiar.
Policy exists because most cases are routine, and a written rule lets a brand answer them fast and fairly. It stops being useful the moment applying it exactly, to one specific customer, would be technically correct and obviously wrong.
Before applying a policy strictly, ask: is this customer's history with the brand worth more than this one exception costs? Is the exception genuinely rare, or would granting it set a pattern you couldn't sustain? And — the one people skip — is there a version of "no" that still leaves the customer feeling heard, even if the answer doesn't change?
A first-time customer asks for a refund on a used, opened product forty days past the 30-day window, with no prior contact. Holding the line, explained clearly, is defensible — there's no relationship yet to protect, and the request is well outside policy.
A three-year repeat customer, several past orders, asks for the same thing four days past the window because a family emergency delayed her return. A one-time exception, clearly logged as one, protects a relationship worth far more than one refund.
The distinction isn't "be generous" versus "be strict" — it's reading the actual situation each time, which is precisely the judgement a chatbot enforcing policy literally cannot do.
Write one situation where you'd hold a policy firmly, and one where you'd grant an exception, using the same policy for both. Say why in one line each.
A complaint left as a public review, a comment, or a tagged social post is the same underlying situation as a private message, with one real difference: an audience is watching how it gets handled, and the reply is doing two jobs at once — resolving this customer's problem, and showing everyone else what happens if something goes wrong with this brand.
The acknowledge-own-act shape from 3.2 still applies. What changes is the public reply should be shorter and move the specifics to a private channel fast — "I'm really sorry this happened, I've sent you a DM to sort it out properly" — rather than litigating the whole situation in public, which reads as either defensive or like airing the customer's private order details for an audience.
"Hi [name], I'm really sorry your order arrived damaged — that's not the experience we want anyone to have, and I'd like to make it right properly. I've sent you a message directly so we can sort a replacement without you having to share order details here. Thank you for flagging it."
A defensive or corrective public reply reads badly to everyone watching, regardless of who's technically right. Move the disagreement private; keep the public reply warm, brief, and visibly taking the problem seriously.
Write a public reply to a one-star review, following the worked example's shape, then the private message that would follow it.
Occasionally, no reply you write is going to work — the customer needs to hear from the founder directly, or the situation has moved past what support judgement alone should decide (a legal threat, a request touching someone's safety, a sum of money beyond what you're authorised to offer). Recognising this moment is itself a skill, and it is not a failure to hand off.
The customer has explicitly asked to speak to someone more senior. The situation touches money, safety, or legal exposure beyond your authority — the same tiers Course 01's escalation lesson names. Or you've tried the acknowledge-own-act shape twice and the customer's anger is increasing rather than settling, which usually means the real issue hasn't been named yet, and someone with more context needs to find it.
To the customer: "I want to make sure this gets handled properly — I'm bringing in Ada, who can make a call I'm not able to on this one. She'll be in touch today." To Ada: "This customer's had two failed fixes already and is understandably frustrated — I think this needs a gesture beyond what I can offer alone. Full history attached."
Write the two messages from the worked example — to the customer and to whoever you're escalating to — for a situation of your own choosing.
Three threads are supplied: one email, one live-chat transcript, one public review. Each is a genuine risk to the relationship if handled badly.
Whether the judgement from this module holds up against real, specific anger rather than a calm hypothetical — the same test module 3's whole subject is built around.
You should now have: a way of reading past a surface complaint, the acknowledge-own-act shape, a policy-versus-relationship test, a public-reply template, and a sense of when to hand a conversation off. Module 4 turns to the mechanics underneath all of this — the five channels support actually happens on, and the different rules each one runs by.