A story. Maya Sharma runs marketing at a D2C fashion brand in Mumbai with two million names on its list. Arun is one of them. Rekha is another. Tara is Arun’s daughter and has never opened an email from a brand in her life. Nobody in the story explains anything — the short notes after each part do that. Read it as a story.
1
The Invoice
The invoice was for ₹13.4 lakh and Maya had no argument with it.
That was the odd part. She had been reading invoices from marketing suppliers for nineteen years and the skill she had built was not arithmetic, it was translation. A retainer meant we will be busy on your behalf. A licence meant your team will have somewhere to log in. A media bill meant we reached some people, some of whom were going to buy anyway, and the sums that would tell you which are sums we control. She signed those and felt a small unresolved thing each time, like a stone in a shoe she had stopped shaking out.
This one was a share of a number she had watched being made. She had agreed the size of the group that would be left alone before the work started. She had agreed what ordinary looked like. Ninety-one days later the two numbers sat on the same page and the gap between them was not a matter of opinion.
So she signed it, and for the first time in a decade felt she had bought something rather than rented a hope.
Then Ashwin came in with the sheet.
**
Her CFO had a habit of putting one page on the desk and not sitting down, which she had eventually understood was not aggression but efficiency.
“Three lines,” he said. “You had one for years.”
“I know what I have.”
“Sending cost. Then the fee for the things your team runs inside the emails. Then this.” He touched the invoice. “Which is a share of an improvement, and which annualises to about four crore eighty.” He paused. “There’s a fourth. You showed me a slide in March. Other brands pay to reach people who open your mail.”
“I had a slide.”
“Where is it?”
Maya looked at the page. Ashwin saw a cost line, three deductions and a net figure that had fallen for two quarters. Maya saw the number that was not on the page: what it would cost to damage the attention that made those deductions possible.
“If I put another brand’s offer inside my emails, it will work,” she said. “Month one, better in month two, because we’ll get better at choosing. And somewhere in month five, people stop opening.”
“You don’t know that.”
“No. Here’s what I do know. That invoice exists because a hundred and fifty thousand people who were drifting away from us didn’t. If they stop opening, the improvement stops. If the improvement stops there’s nothing to take a share of.” She turned the page round. “Your third line is paid for out of the same thing your fourth line spends.”
Ashwin sat down, which she took as a concession.
“Then how much,” he said, “can this inbox earn before it stops being worth opening?”
She did not have the answer. She did notice that nobody in nineteen years had put it to her in that form.
**
“Eight weeks,” she told him that evening. “I’ll show you the fourth line. And I’ll be able to prove it hasn’t cost us the third.”
It was 7.39 p.m. She shut her laptop.
One minute later, in three different parts of the same city, three inboxes did three different things.
Arun, Rekha and Tara at 7.40 p.m.
What just happened
Maya’s programme now has four lines where it used to have one: the cost of sending, then capability revenue from what customers complete inside the message, outcome revenue from verified improvement, and media revenue from attention other brands will pay to reach. The fourth is the largest and the only one that can eat the other three, which is why she is frightened of it.
The inbox had one verb. Now it has four.
**
Key points
- Four lines, and only the first is a cost.
- The media line is last because it is the one that can destroy the attention the other lines depend on.
- The governing question: how much can an inbox earn before it stops being worth opening?
2
Twenty Seconds at a Stove
Arun Mehta was not a lapsed customer and would have been mildly offended by the suggestion.
He bought from them three or four times a year and had done for six years. If you had asked whether he was a loyal customer he would have said yes, and he would have been telling the truth as he understood it. What he did not know was that four times a year had quietly become two. He still opened their emails. He opened them the way you glance at a hoarding from a moving train — the eyes register, nothing happens, the train goes on.
It did not feel like drifting away from anything. It felt like being forty-six and busy.
On every dashboard his team used, he looked fine.
**
The mail arrived while he was standing in the kitchen with a chai he had made and forgotten to drink, waiting for the rice, half-listening to commentary from the other room where his daughter was doing something on her phone he had given up asking about.
He opened it because he was standing there and it was there.
What was odd about it — though he could not have said what was odd — was that it was about the thing he had been meaning to do. Not a sale. The two kurtas he bought, the same two colours every eighteen months, a fact about himself he had never said out loud and would have been faintly embarrassed to see written down. In his size. In stock, four left. At ₹2,400, which was what he had paid last time. His unused store credit of ₹300 already taken off. And a line at the top saying he had last bought them in February last year.
He had not known that. He had known a while ago. The month turned a vague intention into an overdue task, which is mildly annoying in the specific way that being shown your own postponement is annoying.
There was a button. He had a small sigh ready for the browser, the login he would not remember, the OTP arriving in a different app while he stood at a stove.
He pressed it and a panel opened inside the message and asked him to confirm ₹2,100.
He confirmed. Twenty seconds, from opening to done.
Then the mail asked one more thing: would he like to hear about these again in nine months, or fourteen?
He thought about that for longer than he had thought about the purchase. Nine felt keen. Fourteen felt like a long time to go without thinking about kurtas. He chose fourteen, because it was true, and because being asked a question with two honest answers is a small pleasure and he did not examine why.
Then the rice was done and he went and ate it.
He did not think about the mail again. He would have been surprised to learn that four things had happened, or that any of them had been paid for.
What just happened
The old mail was written on Monday and was true on Monday. This one was assembled at the moment he opened it, which is the only reason it could show a real price, real stock and a real credit balance. Inside those twenty seconds: a payment that used to leak somewhere between the tap and the checkout, a fact about him that nobody had known and that will be read the next time somebody decides when to write, and — quietly — his removal from the list the brand had been using to buy his attention back through an ad platform. He had been reachable for nothing all along.
The same customer, the same evening, before and after.
Key points
- A mail written at send is a prediction. A mail assembled at open is a decision.
- The click-through was never a step in the journey. It was the leak.
- Somebody who completes inside the inbox has proved they are reachable free, and should stop being bought.
3
The One Who Was Left Alone
Rekha Iyer had ordered from the same brand nine times in five years, which by every measure they used made her a better customer than Arun.
Her interval had stretched too. The same four months, the same slow lengthening, the same absence of any decision to go away. She and Arun had never met and never would. What they had in common was a shape: order value, order gaps, the line their opening rate had made since March. When the list was drawn in April there were three lakh people on it. Both of them were on it. Neither knew.
No analyst chose which of them would get the new treatment. A coin toss put Arun in one half and Rekha in the other. That single fact is the entire reason the comparison would be worth anything at all.
At 7.40 she was on the Western Express Highway, going nowhere, on a call with her sister about their mother’s cataract appointment, which had been moved twice and would be moved again. The phone was in the cradle. The mail arrived.
It was not a bad mail. Somebody wrote it, somebody approved it, somebody chose the picture, and it went to nine lakh people that evening including her, and it said what such mails say: new season, things she might like, come and look.
She saw the sender name go up the screen. She did not open it.
There was no moment of rejection. That is the thing — there never is. Her sister was talking. The traffic moved four feet. By nine o’clock she could not have told you the brand had written to her, because in the sense that matters they had not.
She did nothing wrong. Nobody in this story does anything wrong. The mail gave her no reason to interrupt her evening, and it had not been built to give her one, because it had been built for nine lakh people at once.
**
It would be convenient to say Rekha never bought from them again.
She did. Eleven days later, on a Saturday, she needed something for a wedding and went to the site directly, the way she had nine times before, and spent ₹2,100.
Nobody recovered her. Nobody did anything at all. She needed a thing and went and got it, the way a share of every drifting customer does, every month, in every category, entirely unassisted — a fact that has funded a great deal of confident reporting in the history of marketing.
Rekha’s ₹2,100 belongs to the brand. It does not belong on anybody’s invoice.
What just happened
Rekha is one person standing in for a hundred and fifty thousand. To know whether anything worked, somebody has to be left on the ordinary version at the same time — not compared against last year, which measures the season and the pricing and the weather. And she is the reason the group matters rather than merely being rigorous: a slice of the untreated group comes back on its own. Treat everybody and every one of those purchases gets counted as something you caused.
Two groups. Only the gap gets paid for.
**
Key points
- The comparison has to run at the same time, or it measures the weather.
- Some of the group you left alone comes back anyway. That is the baseline.
- Gross sales are not proof. The gap is.
4
Day Sixty-Three
The report came on the fifteenth and Maya read it twice before her nine o’clock.
She had learned to read these backwards from every other supplier report she had ever seen. The first number was not the result. The first number was what the untreated group did: a hundred and fifty thousand people, left alone, getting on with their lives. Eleven point two per cent of them came back. Some fraction of that eleven point two was Rekha, on a Saturday, buying a wedding gift.
The treated group came back at fourteen point six.
Three point four points. Across a hundred and fifty thousand people at an average order of ₹2,400, that was ₹1.22 crore of revenue in ninety-one days that would not otherwise have existed. Annualised across the whole base once the other half was treated, ₹4.8 crore. Against what she was spending to buy those same people back through a platform, more than that again.
“Nothing was recovered,” she said, when the analyst finished. “He hadn’t left.”
“No. Fewer of them became lost.”
“Which is an absence.”
“An absence with a control rate attached, and a reacquisition bill that won’t arrive in eighteen months.”
That was the moment she understood what she had bought, which was not a campaign and not a tool. She had bought a number for a thing that does not happen.
**
Ashwin’s objection was the good one.
“Three lakh eligible. You treated half. So for ninety-one days you deliberately did not do the thing that works, to a hundred and fifty thousand people you knew you could help.”
“Yes.”
“What did that cost?”
“About one crore twenty, if the lift holds.” She had made him do that arithmetic himself in April and he had not enjoyed it then either.
“And you’re comfortable.”
“No. I do it because the alternative is worse.” She turned her screen. “Rekha Iyer. Row four hundred and something. Didn’t get the treatment, came back anyway, spent ₹2,100. Now — if I’d treated everybody, no control group, and she’d got the mail and then come back on that Saturday, what does the report say?”
Ashwin looked at it. “It says we recovered her.”
“It says we recovered her. And I pay somebody a share of ₹2,100 for a wedding gift she was going to buy anyway.” She turned the screen back. “There are thousands of Rekhas in there. That’s not a rounding error, it’s the difference between a measurement and a story.”
“So the crore twenty is the price of knowing.”
“The crore twenty is the price of knowing. Once. Properly.”
He was quiet a while. Then: “Eleven years of approving marketing spend. First time anybody’s told me what they didn’t do.”
**
There was one line in the report Maya did not raise in the meeting.
Rekha Iyer had crossed into the dormant list on day sixty-three. Not because of anything the brand had done, and not because of anything it had failed to do beyond declining to do something it did not yet know would work. She had gone quiet for ninety days and the definition caught up with her.
Maya sat with that for a while. It is one thing to hold a group in the aggregate. It is another to be able to name a row.
What she did about it was undramatic and probably the most useful thing she did that year. Three rules into the operating agreement. Control windows finite and agreed in advance, never extended because a chart looked better with more weeks in it. Complaints and unsubscribes reviewed alongside revenue, with either able to stop a test regardless of what the lift was doing. And on the day a window closed, the control group gets treated.
“Window shuts on the thirty-first,” she said at the end. “First of next month, everybody gets everything.”
“That will muddy the—”
“It will muddy nothing. The measuring is finished on the thirty-first. After that they’re just customers I know how to help and haven’t.” She was already standing. “There’s no version of this where we keep them in the dark because the chart looks tidier.”
What just happened
The baseline is what would have happened anyway. The gap above it is the only thing anybody gets paid a share of — no gap, no fee. And the gap gets measured on the incremental completed outcome, not on opens or clicks or anything a supplier can flatter. What the essay does not show, because it is a spreadsheet and not a scene: every one of those interventions also wrote down what was tried, on whom, in what state, against which control, and what happened. That record is the part that does not get cheaper for a competitor to copy.
**
Key points
- No gap, no fee.
- Holdout windows are finite, agreed in advance, and treated the day they close.
- Complaints can stop a test regardless of what the revenue is doing.
5
Three Rooms
Tara Mehta was twenty-one and had never in her life opened an email from a brand.
It was not a position she held. She had no objection to brands and bought things constantly. Email, in her understanding of the world, was where institutions put documents — her college put results there, her bank put statements there, somebody at an airline once put a boarding pass there. Brand mail went into it the way junk post goes into a hallway: real, present, unread, cleared in batches with a swipe.
Her father, she had noticed, read emails. She found this endearing and slightly archaic, like his habit of printing things.
Her diagnosis of why she ignored them was sharper than most of the industry research on the subject, and she delivered it at dinner without any sense of having said something interesting. “A sale isn’t an event just because they put ends tonight in the subject line.”
So the fact that she had, by that July, opened the same thread eleven days running needs explaining.
**
It came in through the group.
Nine people from her course, in the same WhatsApp thread for three years and probably good for thirty more. Nikhil, who finds things, put a link in it in early June with a message that said only this one’s decent, need one more. Four joined. Then two more, because four had.
What they had joined was a Circle — six people inside something much larger, with a thing to do every day during the cricket. MyToday is the app it all sits under; Muniverse is the world; WePredict is the part where you say what you think will happen. You spend Mu to do it. Mu is earned by turning up — you cannot buy it, and you cannot cash it out, which is the whole reason it is interesting. Get it right often enough and your Predictor Score goes up, and that is yours and cannot be bought either.
Tara turned out to be unusually good at one thing, and it was not cricket. It was not being confidently wrong. Her score climbed while people who knew far more about the game stayed flat, and she found this deeply satisfying in a way she was not willing to describe to anybody, including herself.
**
The mechanism, on her phone, was one thread from one sender that never spawned a second thread.
This is the part her father would not have understood, and she tried once, briefly. Every day or two it came back to the top of her inbox with something new. But it was not a newsletter and it was not a series, because if she opened Tuesday’s message on Friday, Tuesday’s message showed her Friday. The score in it was the current score. The deadline was the deadline as it now stood. Her Mu was her Mu as of that second.
She found this by accident, scrolling back for a friend’s total from an earlier match. She opened the old one, then the newest one, and they showed the same evening.
“That’s either clever or slightly creepy,” she said.
“Did it show you something you hadn’t given it permission to show?” her father asked, from behind a newspaper, and she considered this seriously because it was a better question than she expected.
“No. It just refuses to be old.”
No app to install. No store. No permission dialogue she had declined in 2023 and never revisited. It was simply there, in the place she already looked eleven times a day, and it was always today.
Tara’s inbox at 9.14 p.m., and the three reasons she opens it.
**
On the ninth day she was at a friend’s birthday on a rooftop in Bandra, holding a drink she did not want, and it came back at 9.14 p.m. It did not tell her a streak was at risk. It named the five people who had gone and the one place still open.
Three minutes later Nikhil wrote in the WhatsApp group: you’re the only one left.
Tara had, at that moment, no interest in cricket at all. She had a mild interest in the drink and a strong interest in a conversation happening two feet to her left. What she had was five names on a screen and one gap where hers went, and a person she would see on Monday who would point at it.
She opened it and did it standing up, ninety seconds, one thumb.
It would be wrong to say she enjoyed it. More accurate: not doing it was not available. A private streak can be broken in silence. Five people and an empty place are an obligation.
**
Here is the part that took Maya’s team eight months to understand, and it did not come from Tara.
Arun was in Muniverse too.
Not in cricket. In gold. He had checked the rate every morning since 2011, the way a certain kind of forty-six-year-old does, and one of the rooms simply attached a question to a number he was going to look at anyway: closing level, over or under. His Circle was his brother-in-law and two men from his old office. They said almost nothing to each other. He opened it at 8.15 every morning with his chai and it took forty seconds and he would have told you, if asked, that he was not using an app at all.
And Rekha was in one as well, though she would have described it differently again. Cousins across three cities, seven of them, and what they predicted was mostly nonsense — match totals, the monsoon, whether a film would open well. She could not have told you her Mu balance. What she could have told you is that the thread is where her sister is at ten o’clock at night, and that on the evening of the cataract appointment four people asked how it went before she had thought to tell them.
One world. Three rooms. Three completely different reasons to be there.
Standing, the number, and the people. Nobody returns for the same reason.
**
Two things went wrong that summer and they are more instructive than the eleven days.
A Circle died. It had been assembled through a promotion rather than out of anybody’s existing group — people who had responded to the same offer and consequently had no relationship with each other whatsoever. It ran nine days. Nobody waited on anybody, because nobody would have noticed an absence, and an obligation to strangers is not an obligation. The instinct was to prop it up with more prompting, which is what every product does at that point. They shut it instead, and wrote down the rule they had just paid for: mechanics amplify a reason to care; they cannot manufacture one.
And in late July the thread got talkative. A build lowered the bar for what earned an interruption and Tara’s inbox got something from it four days running that did not deserve to be there. She stopped opening for four days — not deliberately, with no decision at all, in exactly the way Rekha had not decided anything on the Western Express Highway.
The bar went back up, and then higher than it had been. She came back inside a fortnight, which she would not have done if it had gone on a month. The world was not allowed to claim that every update mattered.
What just happened
No individual brand can do this. A bank, an airline, a retailer has two or three useful things to say to a specific person in a week, and for most people in most weeks, none — and raising frequency manufactures sends rather than attention, which is why click retention falls while the campaign dashboard looks fine. Daily attention has to come from something with its own reason to be interesting, and then be made available. Note also what carries the return: something live on a clock, a place you are further along than you were, named people who will notice, and a record that is yours. Three people, three of those four in different proportions.
**
Key points
- Frequency without value destroys attention faster than silence would have.
- The consumer world is not a side project. It is the supply side of the media line.
- Mechanics amplify a reason to care. They cannot manufacture one.
- The scarce thing is not content. It is the right to interrupt.
6
The Unit That Failed
The first one failed, which Maya had not expected and was afterwards grateful for.
August. One category, adjacent and non-competing, a single unit inside the Thursday digest, twelve per cent of the engaged base, and a group who saw none of it. The offer was from a footwear brand and it looked right — her customers bought kurtas, her customers had feet, somebody had made a slide with a Venn diagram on it.
The response was strong. Two point one per cent took the action, better than her own promotional units managed.
Six weeks later the cohort that had seen it was opening at 31.2 per cent. The group that had not was at 33.8.
Two point six points. By the standards of the industry she had worked in for nineteen years, nothing — well inside the range where a reasonable person shrugs, and comfortably inside the range where an unreasonable person calls it noise and books the revenue.
She stopped it on the Monday.
**
“The slot worked,” said Ashwin.
“The slot took money out. The surface lost value.”
“Two and a half points.”
“On twelve per cent of the base, for six weeks. Run it across the whole base for a year and tell me what it does to the third line.”
He did not need paper for that. “More than the media.”
“More than the media. And the media was ₹3.8 lakh.” She let it sit. “This is what I couldn’t explain in July. The fourth line is real and it can get big. It’s also the only line on that page that eats the others. Revenue arrives now. The damage arrives six weeks later. And the only way anybody grants permission is by continuing to open.”
“So what was wrong with the shoes?”
“Nothing was wrong with the shoes.” She had spent the weekend on this. “It was in the wrong mail. Thursday’s digest is the one people open because it’s useful. It’s the one doing the work of keeping them. I put a stranger inside the thing that was earning the trust.”
The decision surprised the advertiser and reassured her own team, in roughly equal measure and for the same reason. The attention rule was not a line in a governance document. It could veto revenue.
**
October was smaller in every dimension: one unit, one category, six per cent of the base, placed inside a mail people opened because they had a specific reason to.
The response was 0.9 per cent, less than half the first attempt, and her team was visibly disappointed, which she had to work quite hard not to correct too sharply.
Six weeks later: exposed 33.6, unexposed 33.7.
That was the number she took to the board. Not the 0.9. The nought point one.
Two attempts, and the number that decided both.
What just happened
The metric is not yield per open, it is yield net of the effect on attention that will or will not be there in six weeks. A programme optimising the first will always over-monetise, because the revenue lands now and the cost lands in a future quarter. And the reason Maya could stop a working line is not unusual virtue: her fee depended on those opens surviving. A market that consumes its own supply is not a network, it is a liquidation — and the protection against that is not a policy but an arrangement in which spoiling the surface is immediately expensive to whoever holds the dial.
**
Key points
- Ad load is set by what happens next, not by what fills today.
- The party holding the dial should be the party whose income dies if attention dies.
- The best result in this section had less than half the response of the worst one.
7
Both Sides of the Same Market
There is a detail from that autumn Maya did not learn until much later and has since used, with permission, in two conference talks.
In September, between the failed unit and the second attempt, her own brand appeared inside somebody else’s surface. Not a platform. A world — one of the cooperative arrangements her team had joined mostly to understand how the routing worked, in which brands with earned attention made a small amount of it available to brands in adjacent categories, and everybody paid for completed actions rather than impressions.
A twenty-one-year-old in Bandra, in a thread she opened every day because five people she knew were waiting in it, saw one unit from a D2C fashion brand. It did not ask her to buy anything. It offered something adjacent to a thing she had already said she was interested in, and she saved it for after her exams.
Maya was, in the same quarter, the host and the advertiser. In her own mail she sat in the publisher’s seat and another brand paid her for access to attention she had built. In Tara’s thread she was the advertiser, paying somebody else for access to attention she had not built and could not have built — because she did not have, and would never have, a reason to be interesting to a twenty-one-year-old every day during the cricket.
The two transactions were the same transaction from opposite ends. The only difference was who had done the work of being worth opening.
**
In December Ashwin put the sheet on the desk and sat down first.
Four lines. The cost. The fee for what her team ran. The share of the improvement, which had grown, because two more mandates had been added in the autumn and one was working better than anyone projected. Then the fourth line: ₹22 lakh for the year against a media budget of ₹9 crore, and it had cost her, in the failed attempt, most of what it earned.
The net figure at the bottom had fallen again.
Maya did not open it first.
This was not a gesture she planned and she would have been embarrassed to have it pointed out, but the order in which a person opens two documents is the most honest available statement of what they believe. She opened the six-week attention report. Return stable. Complaints flat. The smaller October unit holding its nought point one through a second window. The customers recovered in the autumn still opening, three months after buying, which was the number she had come to care about more than any conversion figure anybody had ever put in front of her. The control group from the summer treated in August, as agreed, on the day the window shut.
Only then did she look at the P&L.
She said nothing about it and neither did he, because there was nothing to say. It was a consequence, not an achievement. Nobody in the room had done anything in December. The number was what four quarters of not over-monetising a surface looks like when you write it down, and it could reverse — if attention weakened, if the slots got louder, if the control discipline softened, if somebody started counting activity as value.
“Same again next year,” Ashwin said eventually. “Slowly.”
“Slowly.”
**
On the first of August, back in the summer, a mail went out to a hundred and fifty thousand people who had spent ninety-one days receiving the ordinary version of everything.
Rekha Iyer opened hers in a queue at a chemist in Chembur, waiting for her mother’s drops, on the day the cataract appointment finally happened. It knew what she had bought in April. It knew her size.
She did not buy anything.
She answered one question, because it had two honest answers and answering took four seconds, and she saved something to look at later. That was all. Not a rescue, and no scene. A channel that had been shut for four months becoming, in about ten seconds in a chemist’s queue, slightly open again.
She did not know she had been in anything, or that a number had been established at her expense, or that establishing it was the reason the mail in her hand was any good.
There is no version of this where she needed to know.
What just happened
In adtech the brand pays a platform to reach people who are frequently its own customers. Keep every part of that and change one thing — the brand takes the publisher’s seat — and the money runs the other way. A brand can sit on both sides of the same market in one quarter, and the only thing determining which seat it occupies is whether it did the work of being worth opening. The falling number at the bottom of Maya’s page has a name, and it is a scoreboard rather than an offer: it arrives when the three revenue lines have covered the cost line, and it cannot be bought from a supplier by negotiating a send rate down.
**
Key points
- Same components as adtech. The brand changes seats and the money arrow turns round.
- Which seat you sit in depends only on whether you earned the attention.
- The order in which you open two documents is the most honest thing you will say all quarter.
**
Nothing in this story is a technology. Twenty seconds at a stove needed a mail that knew the price. Eleven days needed a thread that refuses to be old. Both are engineering and both will be ordinary within three years. What will not be ordinary is a marketing team that will leave a hundred and fifty thousand people alone for ninety-one days to find out what it is worth, stop a working revenue line over two and a half points, and look at a falling cost line in December without claiming credit for it. The inbox was never the hard part. The hard part was being willing to know.
