Support: resolve and win back.

If someone is not happy with the experience, correct it for them, and then correct it for everyone who comes after. Yes, you cannot make everyone happy, and that is not the point. Albert Hirschman gave the point its best description in 1970: an unhappy customer has two options, exit or voice. Exit is a number in a churn report. Voice is information, and a brand that makes voice easy learns what a brand that makes it hard never will.

The research on what happens next is older than most of the brands reading this. TARP's studies for the US Office of Consumer Affairs in 1979 and 1986 found that most unhappy customers never complain to the company at all; they leave, which is Hirschman's exit. The ones who do complain and are satisfied come back more often than the ones who never had a problem. Hart, Heskett and Sasser named the effect in 1990: a good recovery can create more goodwill than if things had gone right in the first place. The Ritz-Carlton built a policy on it, letting any employee spend up to $2,000 on a guest's problem without asking. The number is the point: it tells the employee what a saved guest is worth, before the argument starts.

The lesson I carry into this stage came from a CEO I worked for, in a conversation about selling. He said: you get it. You find the need, and then you sell the solution. The solution is what your customers need, and if you do not have it, it is fine to point them in the right direction. Support is the stage where that is tested, because the person in front of you has a need the product did not meet.

The step: detractor

Reichheld's word, from the recommendation question in the previous stage. A detractor is someone who would tell a friend not to buy, and the negative branch out of Fulfilment delivers them here. The two mistakes are to treat them as a cost to be minimised, and to treat them all the same. The complaint about a damaged box, the complaint about a finish that did not match the sample, and the complaint about a reorder reminder that arrived while the last order was still full are three different needs, and only one of them is about the product.

At a pharmacy and health brand, the reminder meant to catch the second order fired near day 89 for almost everyone, while across 2,477 repeat buyers the median gap between first and second order was 111 days. The old trigger matched an observed reorder 3% of the time. Customers were not unhappy with the product. They were being reminded to buy something they still had, which is a small insult repeated monthly. The fix was not a better apology; it was a flow that times reminders by product category from 18,412 orders and stops the moment someone reorders. That is support done upstream, so the complaint never has to be made. The best support stage is the one the customer never sees.

The flow: winback

The winback flow is the red loop from the detractor step back to the start, and it carries the heaviest expectation in the map: bring back the person who left. Most winback flows are a discount sent at a fixed number of days. Two things are usually wrong with that. The day is a guess, and the discount answers a need the person may not have. Ehrenberg's finding about promotions applies here with extra force: a discount to a lapsed customer buys back the ones who were coming back anyway and teaches the rest to wait for the next one.

At a kidney-health subscription brand, the save series offers pause or skip before any discount. The brand added a pause option in July; 175 subscriptions paused, and only 2 of them had pressed cancel first. Pause had become a second way out rather than a save, so I recommended offering it only at the cancel step, where it meets someone who was leaving. The six subscription flows follow the months where people cancel: about 18% in month one, 35% by month three, half by month five, with checkpoints at day 90 and month six. A winback that fires on those days meets a person at the moment of the decision. One that fires at day 60 for everyone meets nobody in particular.

A save is not a cut

Deliverability has a support stage too, and it is the place brands reach for the axe. After a bad bounce day at the same brand, 7,393 bounces in a day after its domain went offline, the usual move would have been to suppress everyone who bounced. Instead, recent bouncers went on a rolling 14-day cooldown and came back automatically, never-engaged profiles stayed out for good, and bulk sends moved to people who had opened, clicked or ordered recently. The cooldown is a save: the person did nothing wrong, their mailbox did, and a permanent suppression would have turned a technical fault into a lost customer. Hirschman again: the bounce was voice from the mailbox, not exit from the person.

The objection: support is a cost centre

It is a cost centre the way the Ritz-Carlton's $2,000 is a cost: a number put next to what a saved customer is worth. At the subscription brand, median tenure was four to five months and the mean seven to nine, so a saved subscriber in month three was worth several more orders on average. A brand that knows that number can afford the first reply within the hour and the pause before the discount. A brand that does not know it treats every complaint as the same ten-dollar refund, and loses the ones that were worth a hundred.

What I check first

• Time to first reply. Hours, not days. Most of the save happens in the first conversation.
• The complaint piles. Product, delivery, expectation, price. The biggest pile is the next product decision, not the next refund.
• What the winback offers, and when. If the answer is a discount at day 60, check the day against the median reorder gap and the offer against what people complained about.
• Whether pause is a save or an exit. Count how many people who paused had tried to cancel first. At the subscription brand it was 2 of 175.
• What a saved customer is worth. Tenure times order value, from the store. Put it next to the refund budget.

If you take one thing from this stage: find the need, meet it fast, and fix it upstream so the next customer never has to ask. Some people leave anyway, and some arrive here without ever buying. The last stage is where the brand decides what to learn from both.