Marketing: Das's Unit 3 Wrap-Up

Unit 3: Market-Facing Functions, taught by Professor Das Narayandas. The final session: a summary of Unit 3's concepts, with Starbucks Korea as optional reading.

Over this unit I wrote pre-class pieces on five of Das's cases: Olay, Belden, Bajaj Finance, Equitable, and Starbucks Korea. Each one stopped at a multiple-choice question.

In the last session he didn't teach a new case. He pulled three weeks of material into a single thread: first find your own market, then staple yourself to the customer's journey; make a promise, then get the whole organization — people, processes, AI and pay — to keep it together. In the final few minutes, he talked about himself.

Partway through the second half, he asked if we could give him an extra fifteen minutes. We said yes.

Here goes — enjoy:

Whoever you promote is who the organization becomes

The classroom we were in is called the Big Room at HBS. Das opened by explaining what the room is used for: this is where a hundred or so tenured professors decide who gets tenure and who doesn't.

The review report is written by a committee of three tenured professors. They spend four months interviewing dozens of colleagues and field experts, and what they write reads like a case — about twenty pages, often longer than the cases we read. Everyone reads it, discusses it in this room for ninety minutes, and then votes. Three options: yes, no, abstain.

He asked: which one is the worst?

The answer is abstain. Four months of preparation and ninety minutes of discussion, and you still can't decide — either the case itself has a problem, or something hasn't been said out loud.

He went on: in a professional services firm, the most important decision is whom to promote. Whoever you promote is who the organization becomes. Put the wrong person in this room and they will slowly strip away the institution's character.

When someone isn't promoted, the mood in the room is far heavier than when someone is. It takes roughly nine years to reach that review. Saying no means admitting that eight to ten years of investment have come to nothing. So it can't be driven by emotion; it has to be calm and objective, and it can't stray from your own culture.

This was his setup for the day's last topic: talent moves, and good talent only gets more expensive. At the very end of the day, he talked about pay.

Don't call yourself small — go find the biggest market that's yours

The summary began with MAM — My Addressable Market. Das said he's still looking for a better name, something like "my fair share."

His demand was blunt: never say you're a small player in a big pond. Saying that isn't modesty; it means you haven't understood yourself.

If someone in your organization says "we're not big enough," ask them to show you the market where we are the biggest. Then go fight there. Be the 900-pound gorilla in that patch, and you set the rules; the game gets played your way.

He said that, compared with some of the other people he teaches, what the people in this room lack is resources. He's also taught people whose cost of capital is very low and who have investors chasing them with money; for them he'd give a different talk. If your resources are limited, you need the maximum return from every unit of them. He doesn't buy the VC line of "go global, go fast."

Win on your own turf first, and you earn the right to make the turf bigger.

My own reading: MAM isn't asking how big your ambition is; it's asking where you get to write the rules.

Don't tell me what you do — tell me what your customers go through

To win on your own turf, Das said, you have to make every customer a hero. How? The line he kept returning to all unit: staple yourself to the customer's journey.

The phrase comes from a 1992 HBR article, Staple Yourself to an Order, written by several of his colleagues. Back then, what you stapled yourself to was an order. Today it's the whole journey: from the moment the customer first feels the pain, all the way through consumption and what they do after.

He told the story of Intuit.

Co-founder Scott Cook watched his wife manage the household accounts. She used a checkbook: whom she wrote a check to, how much, each one recorded by hand — a real hassle. This was the era of dot-matrix printers, and Intuit's first product was a small program to help households write checks.

Later they noticed that many of the people buying this home software were small businesses. It made no sense — a business needs a general ledger and financial statements. At first they assumed users had filled in the forms wrong.

Only when they stapled themselves to those owners' daily routines did they see it: the accounting software on the market was written by accountants for accountants. A small-business owner has a good idea and a lot of passion, but isn't a business person by training, let alone an accountant. When someone asked one owner whether he knew about the general ledger, he asked back: which World War II general was he?

If you don't staple yourself to the customer's journey, you'll miss the opportunity no one else can see either.

Das added that insight isn't only about new products. Friction in the processes customers go through to deal with you is also an opportunity. Digitization lets you break the business down into processes: at each step, which parts are people and which are machines. Most friction comes from the manual steps. Automate them — you don't have to lay those people off; they can go do other things.

Then he made a request. From today on, when we meet again, he doesn't want to know what business we're in. He only wants to hear two things: what our customers are going through, and how they experience us.

Before class, in my Bajaj Finance piece, I wrote: "Customer-centricity means the customer can walk into the next product on their own, rather than waiting for a team they never see to push it." This class moved that sentence one step earlier — before customers can walk in on their own, someone has to walk the road with them once.

A brand is a promise; making it and keeping it are two jobs

Das recast Keller's brand pyramid in his own words: a brand is a promise.

First, look at what you've promised; then at how the customer responds to that promise; finally, at what kind of relationship they want with you — an emotional embrace, or an efficient business handshake. Different businesses, different answers. You have to walk both paths; they just carry different weight in different businesses.

He brought back the four-box chart from the Belden session for one point only: a leader has two jobs. One is to take part personally in making the promise; the other is to make sure your people keep the promises already made.

Along the way he put in a fair word for sales. He's done sales himself and loves it. He just doesn't want us to treat sales as the tip of the spear. A promise takes the whole organization to keep; sales is just one link. Always putting sales on the highest pedestal tells everyone else in the organization, over and over: you're not good enough. That isn't true.

When he got to the "bow tie" and "diamond" customer interfaces, he made a point of correcting us: the diamond isn't a sales team, it's a customer-commitment team — mobilizing the whole company's resources to deliver the experience promised to the customer.

This approach costs more. So don't do it defensively. If it's only to hang on to customers you already have, costs go up and revenue doesn't — that's a margin-cutting move. Only do it where revenue can grow faster than cost.

He said maybe 20 to 30 percent of business is still best done transactionally: each deal is just a deal, no past, no future. Being customer-centric doesn't mean being obsessed with every customer.

There was one more line from Unit 2 that he repeated: saying no to an order isn't saying no to the customer. An order you know you can't do well will simply become your reputation. Promises need boundaries. You wouldn't ask a neurosurgeon to look at your toe.

In my post-class notes on Olay, I wrote: the promise can change, but the condescension can't stay. This session filled in the other half — you can promise high, but the promise needs boundaries.

A new uniform doesn't make you a new role

From box-selling supplier, to value-added partner, to strategic advisor — in B2B, B2C and professional services alike, the climb is hard. Das said a new uniform and a few new talking points won't turn you into a solutions provider. You may need different people.

He asked a question he hadn't had time for in earlier sessions. Three account managers: one holds the bottom tier, selling standard products and competing on price and performance; one is at the top as a strategic advisor; one is in the middle, doing a bit of everything. The customer says only one can stay to manage the whole relationship. Whom do you keep?

The bottom one matters most — he earns today's money, and the whole pyramid stands on his shoulders. But when he looks up at the top, what he thinks is: too far, too complicated, I have to go out hunting every day. He'll protect today but can't build tomorrow. The top one loves dinners and conversations about where the industry will be in five years; when the grunt work comes, he'll feel it's someone else's job. The middle one isn't great at either end — which is why he's in the middle.

He described a real experiment he was part of. A diversified financial services firm assigned an account manager to each of its top hundred clients: a third from the bottom tier, a third from the middle, a third from the top. Two years later they looked at sales, and only one variable explained the results: where that manager originally came from. Wherever they came from, that's the category they sold more of. Cross-divisional, "one firm" selling didn't happen once. And the pay had already been adjusted.

If there's one person in the organization willing to step out of their comfort zone, it's you. Don't count on anyone else.

Many colleagues love to say you should turn every employee into an entrepreneur; he disagrees. If someone has the same risk appetite as you, they should start their own company, or walk in and tell you: no base salary, give me 35 percent commission. And by the way: the best people are often the ones who don't want to work for you.

In professional services, the problem is harder. Clients at the top tier listen to only two kinds of people: those with gray hair, or rocket scientists.

A McKinsey partner sitting across from a client will often touch their hair like this — see, I've seen a lot. BCG took a different road and called itself the rocket scientists: we come up with ideas no one else can. The experience curve and the growth-share matrix both came from there. The middle tier is domain and process experts, who rely on deep industry knowledge.

He also said that the person who really made McKinsey the firm it is today didn't put his own name on the door. Once a brand is tied to one person, the promise is that person, and every client will wait for him to show up in person.

Many professional services firms complain that clients don't respect them. Das said clients don't respect you because you've never built credibility at that level. Making promises isn't enough; you have to keep them.

Before class, in my Belden piece, I wrote that beneath its three hard problems lay a single question: "how can a channel system designed to sell parts accommodate selling solutions?" This class gave a colder answer: the system has to change, and the people may not be the same people.

Build a staircase to the moon, and draw the guardrails in plain sight

Coming back to Equitable, he said he isn't selling any particular methodology. To keep an organization innovating you need five things; what you call them doesn't matter.

The first is goals. Lots of people talk about moonshots. Das said your job is to build the organization a staircase to the moon. OKRs or something else, everyone's goals have to roll up into the company's — otherwise it's like being pulled by several horses in different directions, and the organization gets torn apart.

Goals can stretch, they can carry pressure, but don't break people. Don't put moonshot goals into baseline evaluations and pay; that only demoralizes people. Give a realistic goal first, then ask: in your view, what's your moonshot? If they hit it, reward them publicly. People like to be seen.

The second is bringing in the voice of the customer. Design thinking, at its simplest, means listening to customers, not over-engineering, getting a minimum viable product out first, and keeping the customer in the room the whole time. Strategy grows out of execution.

The third is guardrails. The leader draws them, and draws them in plain sight. It's fine to say "this isn't open for discussion" — a family business saying "we handle the finances ourselves," for instance — perfectly fine, as long as it's stated clearly and applies to everyone.

He said that a week earlier, in this very room, he'd been discussing with colleagues how to bring AI more actively into HBS teaching. He spent the first fifteen minutes drawing the guardrails: here's my view, we can debate it; but once it's decided, we don't do anything outside the guardrails.

Inside the guardrails, shorten the distance between decision and action. He has seen plenty of bosses with a stack of documents on the desk, flipping through page after page without reading, signing at the bottom. Ask why, and the answer is: we need a system to stop the people below from going rogue.

That's exactly why he had us read Starbucks Korea. In that incident, seven people had signed off. Each one signed assuming someone else had looked.

One more signature doesn't mean one more layer of management.

The fourth: agility isn't just IT's business; the whole company should think about how to iterate quickly. The fifth: look at skills, not titles. Don't tell him you have three VPs and four AVPs; tell him what skills you have. He admitted this one is the biggest threat to HBS itself — an institution that has always taken pride in titles.

Before class, in my Starbucks Korea piece, I wrote: "Make reviews match real accountability. Everyone should know what they're judging, and whom to go to when they spot a doubt. Adding signatures can't replace that." In the Equitable piece, I wrote: "Who can be 100 percent accountable has to be spelled out." Looking at them now, the two sentences say the same thing — draw the guardrails in plain sight, and land accountability on specific people.

Sixty percent human judgment, thirty percent machine prediction, ten percent hand it to the machine

Next he talked about AI, and then about pay. Before getting into AI he made a point of saying: this is a business person's take, not a technical one.

He sorted a leader's decisions into three kinds.

The first is descriptive analytics: the machine provides historical data, and people make forward-looking decisions. We all grew up in Excel. He put it strongly: if you're still making decisions in Excel two years from now, you've fallen behind. Use AI, check it for hallucinations, but use it.

The real craft here is visualization. He urged us to watch our own people: how do they spot patterns? The next generation reads charts differently from us. Don't compete with AI; collaborate with it, and let it lay out patterns in a way everyone can understand. This kind of decision will still make up about 60 percent in the future — when uncertainty is high, theory is unclear and the world is changing too fast, you still have to rely on intuition.

The second is prescriptive analytics: the machine makes the decision itself. His example was his wife's Tesla. He sits in that car and never touches the wheel — his wife won't let him. So what's left for the human? Giving direction. People set the rules, or the objective function: maximize profit today, or minimize inventory. The machine executes.

What suits the machine are decisions that are each tiny but each not quite precise enough — repeated enough times, the gap gets big. Dynamic pricing on retail shelves is one example. Once the machine has decided, don't second-guess it after the fact.

He has a new HBR article arguing that many companies just put an AI shell over the existing organization and then measure only usage: how many tokens were consumed. An employee using AI to look up where to eat dinner counts as usage too. Usage is not adoption.

The third sits in between: predictive analytics. When past OPM participants showed him their dashboards, nine out of ten featured customer churn. Churn is a lagging indicator — the customer has already left. A better metric is "customers at risk of churning": when a customer that makes up 20 percent of revenue starts flashing red, everyone mobilizes to save it.

Prediction has another trap, and it has to do with whom you hire. Data scientists take pride in accuracy, and the easiest way to raise accuracy is to strip out the fuzzy gray zone. Business people see two kinds of cost instead: false positives bring wasted effort, and false negatives bring missed opportunities. Don't fall in love with accuracy; watch the wasted effort and the missed opportunities. Models should be built by business people and data people together.

He left us homework for next year: hand 10 percent of decisions to prescriptive analytics, use prediction for 30 percent, and keep human judgment for the rest — but use AI to see more clearly.

Once AI steps onstage, go back to the start of the journey

Das said this was the other reason he had us read Starbucks Korea: AI is moving from backstage to the front of the stage.

Think of the business as a performance. Descriptive and predictive analytics are both backstage; if something goes wrong, there's still time to catch it. Dynamic pricing is already at the edge of the curtain. Generative AI and agents stand right in front of the customer: writing content for you, talking to customers for you.

Onstage, a single small mistake by AI can wipe out all the credibility you've built in the market. Many people think the logic of using AI over the past two years will hold for the next two. He says it won't. Onstage AI needs someone watching it 24 hours a day. The letting-go described earlier shouldn't be applied to decisions that directly affect customer relationships — there, you don't get a second chance.

A classmate asked: as more and more of our customer interactions are with AI agents, how do we keep our promises?

Das answered quickly: go earlier in the journey. Be with the real customer before they hand instructions to their AI agent — closer to the trigger point.

When he was in sales, he used to tell his team: if we can't get ourselves written into the customer's spec, the deal is over. That's even truer today. While human emotion is still part of the decision, write yourself in. Once the decision is handed to an agent, the agent uses cool, rational evaluation, and we may not get a chance.

Staple yourself to the customer's journey, not the AI agent's — by then, it's too late.

If the machine makes the call, don't make the user take the blame

While discussing prescriptive analytics, he first told a story and said he'd come back to it at the end.

When UPS rolled out its routing algorithm, ORION, it first piloted it with drivers. The drivers said: it's unbelievably good — it only showed up yesterday, yet it knows my area better than anyone.

Das added: have you ever tried telling a London cab driver he's taken the wrong route? He'll ask you to get out. To get a London cab license you have to memorize tens of thousands of streets and places and prepare for years. When people that proud say an algorithm beats them, they mean it.

But after the full rollout, adoption wouldn't climb. Most drivers were overriding the algorithm's routes.

The problem wasn't the algorithm. When the algorithm succeeds, who gets the credit? Not the driver. When it fails, who gets penalized? The driver. If I'm the one who takes the hit when it goes wrong, I'd rather keep my own hands on the wheel. So they quietly used the system "with the brakes on."

According to Das, UPS's eventual fix was simple: stop evaluating on-time rates and customer satisfaction and look at only one thing — how often you follow the machine. Late? Fine. Something went wrong? Fine. Adoption happened overnight.

People are afraid of being held responsible for someone else's mistake, even when that "someone else" is a machine.

After talking about AI stepping onstage, he came back to this story and placed it in a chart from his HBR article: for any decision, ask two things at once — who makes the decision, and who is accountable for the outcome. Miss either one and things go wrong. The UPS drivers landed in the most troublesome box: the machine decides, the human takes the blame.

Das said this isn't a new problem. ERP was supposed to let headquarters see the whole organization clearly, but plant managers were still evaluated on their own plant's performance. They threw headquarters' production schedules in the trash and did things their own way. Sales force automation was the same. You want good data, and people give you numbers that fill the boxes.

If the machine decides and the machine is accountable, it has to be like walking a dog: if the dog does its business on the sidewalk, the owner cleans it up and can't walk away claiming not to know whose it is. Machines make mistakes, so each machine needs an assigned owner — the people who designed it, who hold the keys to update it. They clean up, not the people using it.

Everyone loves to say "human in the loop." He said: don't make the user that human; make the designer that human. The learning loop and the usage loop are two different loops.

He mentioned Amazon CEO Andy Jassy, who visited campus a few weeks ago: there are a few numbers on his dashboard that even he can't touch. All he can do is tell the people who own the system: I did my best, it didn't work, something's broken, go fix it. Das described this approach to the CEO of a consumer goods company, who said it was impossible — my whole reason for existing is to turn these knobs every day. Das said many people in this room are the same.

There's another box that's the reverse: a person makes the decision but lets the machine take the blame — the dog ate my homework. Just last week he'd met a group of Wall Street traders complaining that their quant models were inaccurate. When they went back and checked, it turned out the traders themselves had been overriding the models all along.

At the end of my pre-class Starbucks Korea piece, I left a question: "When AI makes output faster and cheaper, where are we going to put human attention?" This class gave me a concrete answer — on the rules and the exceptions, and on who cleans up after something goes wrong. Not on adding one more human signature to every single result.

You set the rules; the organization will game them

The last topic was pay. He said there are only two clean approaches: pay for behavior, or pay for outcomes.

Paying for behavior: his example was the person behind the McDonald's counter. You wouldn't hold them responsible for store traffic; they can't control it. They just need to do two things well: smile, and when a customer orders a burger, ask if they'd like fries. Cross-selling is the opportunity. Do it, and you get paid. There are two preconditions: you know which behaviors lead to results, and you can see those behaviors — hence the camera above the counter. Salespeople like this look after both the company and the customer, and think about long-term satisfaction. They'll say, "Let's do this together."

You pay for outcomes when you can't spell out the causal link between effort and result. There are many paths to a close; you can only measure the result, so you pay commission on the result. Salespeople like this look after themselves first, often side with the customer and push the company to give ground, and focus on the short term. They'll say, "I delivered the result — leave me alone."

Then came the two-axis chart: the horizontal axis, can you measure effort; the vertical axis, can you link effort to outcomes. The four boxes were blank on the handout; he walked through them one by one in class.

High on both: give a high base and low commission, or even none. Many financial services firms in this country now say "we don't take commissions," meaning: I'm not maximizing my own income; I'll take care of you.

Low on both — you don't know which behaviors lead to a close, and people are out in the field where you can't see them — pay for outcomes.

The hard ones are the other two boxes.

You know what works, but you can't see what people are doing. Most companies choose a hybrid here: a bit of base, a bit of commission. Das said that's not a good idea. He also joked: it's the last class — don't let my parting impression be that I'm always setting traps for you.

His answer: manage the effort you can't see with training first, then motivate it with pay. Build your own sales methodology, run a boot camp, and don't let anyone into the field until they pass. Once people are out there, you can't manage them. Teach them how to fish and how to bait the hook first, then let them live by their skill. Many companies fear that "people leave once they're trained," so they don't train. But as long as someone is still working for you, one trained person beats an untrained team running around out there.

The last box: you can see the effort — in a digital world, almost every action is recorded — but you don't yet know which actions work. He said this is where many people go wrong. The task here is learning. Don't leave "what works" for salespeople to figure out on their own; use AI and the data you have to learn how customer behavior is changing. Once you've learned it, move up or to the left.

Whenever you think about pay, always think about control at the same time. He calls these the two Cs: control and compensation.

He had no patience for complicated compensation plans. Compensation consultants charge a lot; the more they charge, the more complicated the plan; the more complicated the plan, the harder it is to execute. A pay system is best explained on one page: this is how we pay you. Don't write it as a book, and don't imagine a contract can cover every situation.

He recommended five questions, which he said a colleague taught him (the handout credits Ian Larkin), enough to design any incentive: What good behaviors will this system produce? What bad behaviors will it produce? Do the good far outweigh the bad? How can you reduce the bad — through culture, through training? How much will reducing them cost, and is it worth it?

On the second question, he repeated a line from the previous two units: your job is to design the rules of the game; your people will game the rules. Your organization is smarter than you; never assume you're smarter than it. Cold-calling in his class works the same way: if students know they were called on last session, they won't be called on next session, so they don't prepare. So he cold-calls using game theory.

He closed with three assumptions about people: people take care of themselves first; they're not like you and have different risk appetites; and they're always calculating which takes less effort — going along with you or working against you.

Before class, in my Belden piece, I wrote: "People follow the comp plan, not the strategy speech." What I saw then was the SAMs maxing out their product quotas first and only then getting to solutions. This class completed the thought — people follow the pay, and they also follow the loopholes you didn't design for. So list the bad behaviors as carefully as you list the good ones.

To whom much is given, much is expected

When he finished, he said this was the first time he'd taught the whole unit in this version. He started teaching this unit with OPM 61; this was the third or fourth time. He'll teach it once more in the spring, and then write the book.

In the remaining minutes, he said he wanted to shift his relationship with us from "yelling at you, being tough on you" to something more like friends. He talked about himself.

He was smart from childhood; no exam ever stumped him. Having it come so easily brought other things too: a bit of arrogance, and no sense of purpose. People said he should go to IIT; he'd never even been there, and got in without preparing. Graduate school was the same. He always assumed everything he got was simply his due.

Then his elder son was born. It was the first adversity in his life that he couldn't handle. His wife carried everything, searching everywhere for help for their child. Within a year they decided to leave India to find a better environment for their son to grow up in.

Going abroad for a PhD wasn't about wanting to be a scholar; it was the family of three's way out. That's why he calls himself an "accidental academic." He did research in windowless rooms, never met a customer, and wrote papers that are still considered groundbreaking. When he graduated, people advised him to go to a research university; they chose Massachusetts, because it was the best place for their child to grow up.

In his early years in America, his only thought was getting tenure and giving his family a sense of security. For that, his wife raised their children almost single-handedly. He said he made many mistakes. If he could do it over, the one thing he'd most want to change: to be a more present father and a more present partner. He's learning to be that now.

Finally he said: life is hard. You're all very fortunate; life has given you a great deal. He quoted a line —

Much is expected from those to whom much is given.

Those who have been given much are expected to give much.

He said his two sons have taught him a great deal.

After Class

If I had to compress this whole class into one question, it would be roughly: the customer's journey, the promise you make, the interface that faces the customer, AI's decision rights, the way you incentivize — are these aligned?

Each of the five pre-class pieces in this unit stopped at its own multiple-choice question. This wrap-up session put them back on the same map. The brand makes the promise, but processes, decision rights and pay have to keep it together. If any link is out of alignment, the customer will feel it.

I'm taking these questions back to Arcus42: in which market are we the first name customers think of; does the promise we make have boundaries; which decisions can go to machines, and who cleans up when something goes wrong; and what behavior do our incentives actually reward.

The answers aren't finished yet.

Das said that after teaching it once more in the spring, he'll turn this course into a book. Our cohort got to hear it before the book was written.

---

The class content is compiled from my notes of 2026-10-26; Professor Das's words are my paraphrase, and some wording may differ. Professor Das noted that the handout material comes from the book he is writing; this piece records only my understanding from class and does not reproduce the handout.