Two calendars, forty kilometres apart
Tuesday, 7:20 in the morning. Forty kilometres apart, two men begin exactly the same working week.
The first runs a plastic packaging plant with 240 people. By 7:35 he has already walked line 3, greeted the twelve workers coming off shift and typed into his phone that the old sealing machine "is making a strange noise". His week, exported from Outlook, holds 21 meetings: almost all of them half an hour, almost all one-on-one, half of them booked with less than twenty-four hours' notice. Four walks of the shop floor. Not one meeting with more than three people. Asked why, he answers with a line that in his industry sounds like a credential:
"I'm not a meetings guy. I'm out on the floor."
The second runs a plastic packaging plant with 260 people. At 7:20 he is reading the material for a meeting that will happen on Thursday and that was booked three weeks ago. His week holds nine meetings, all of them an hour or longer, all with four or more people from three or more functions sitting at the same table. Two of them put outsiders —a large customer, the bank— alongside insiders. He walks the plant once, on Thursday, for forty minutes, with the operations manager beside him.
Both work around 55 hours a week. Same industry, same size, same country, the same hours of a life handed over to the business.
In the largest measurement of CEO days that exists, the shape of the second man's week —not the man: the shape— is associated with 7% higher sales for every standard deviation of an index that measures how closely a week resembles each one, controlling for labour, capital and the usual variables (Bandiera, Hansen, Prat and Sadun, Journal of Political Economy, 128(4), 2020).
A standard deviation means nothing in anyone's calendar, so here is the translation, stated openly as an approximation: it is roughly halfway between the first man's week and the second's. Going from twenty-one short one-on-one meetings to about fourteen, half of them long, planned in advance and with more than one function in the room.
You have already chosen which of the two. You chose it years ago, without looking at a single piece of data, and you have never revisited it.
Two shapes of the same week
Instead of sorting weeks into categories invented in advance —the operational CEO, the strategic one, the visionary, the whole bestiary of the conference circuit— the authors handed a machine 654 combinations of activities, defined by who was there, how many, which functions they came from, whether outsiders were present and whether the activity was planned or simply turned up. It is an algorithm nobody told what to look for: it knows nothing about management. They asked it for two pure shapes —and then ran the exercise again with three, with four, with eleven, with twenty— but which activities went into each of the two was decided by the algorithm, not by them. The two shapes that came out are the ones the authors christened "manager" and "leader".
The "manager" shape will feel familiar to anyone who has ever walked a Latin American plant: shop-floor visits, one-on-one meetings, production, suppliers, shorter activities. The "leader" shape is almost its photographic negative: long, planned activities, with several participants, from several functions, and with insiders and outsiders sitting at the same table. Read against the "manager", in the "leader" shape an activity is 1.9 times more likely to bring insiders together with outsiders and 1.5 times more likely to cross functions; and 0.11 times —a little over nine times less— as likely to be a shop-floor visit.
Here is the detail that changes the whole reading. The "leader" shape is not the one that spends more time out of the building: activities with outsiders only are 0.58 times as likely, clearly less. What sets it apart is not going out. It is bringing people together: the meeting where the large customer sits with the head of production and the head of logistics at the same time, and where nobody has to repeat the same conversation three separate times afterwards.
"Manager" and "leader" are labels the authors attached to two statistical patterns, not a ranking: they describe the shape of a week, not the person who has it.
The five dials
One second before the list, because two different ways of counting the same thing meet here and it is worth not confusing them. The figure above shows the result: which activities show up more in each shape —the C-suite, cross-functional work, production, suppliers, shop-floor visits. What produces that result sits one step further back, and it is five variables: the same five the authors used to describe every activity to the machine, and the same five that separate the first man's week from the second's. The seven activities in the figure are the consequence; these five variables are the cause, and they are the only ones you control from Monday morning.
Not one of the five is about quantity.
Duration. How long each block lasts.
Notice. Whether it was booked more than twenty-four hours ahead or simply turned up.
Participants. One-on-one, or four or more around the same table.
Functions represented. A single one, or three crossed.
Who is sitting there. Insiders only, outsiders only, or both together.
Five dials. None of them costs money, none of them needs board approval, and all five are on the inside of your calendar today.
Where this comes from. The diaries belong to 1,114 manufacturing CEOs across six countries, collected through one phone call a day for a full week: 42,233 activities recorded in fifteen-minute blocks. The benchmark is Mintzberg, who in 1973 observed five executives and for forty-four years remained the largest direct record of a chief executive's calendar. Two caveats. First: the number of pure shapes is not discovered by the algorithm, it is set by the researcher; the authors chose two because two yield a single-dimension index that is easy to read, and they showed that the portrait holds up against models with three to twenty. Second: the two shapes are not boxes, they are the endpoints of a line that no real CEO occupies —each one gets a number between 0 and 1— and almost everybody lives in the middle. When this article says "the wrong shape", read "too far, along that line, from where your company is".
Killing meetings: the advice that does not survive the diaries
The textbook advice —kill the meetings, the clean calendar as proof of focus— has a local version here that is nobler and considerably more dangerous, because it comes with a moral credential built in: I'm not a meetings guy, I'm out on the floor.
The diaries say otherwise. Among the CEOs in the sample, the quarter that spends the least time in meetings gives them a little over 40% of their time; the quarter that spends the most, 65%. Put another way: even the CEO who meets least spends more than four hours in every ten in meetings. Meeting is not a deviation from the number one's job: it is the number one's job. Those two figures measure how much, not how, which is why they say nothing about the outcome. And the shape associated with higher sales is not the one with fewer meetings: it is the one with fewer meetings that are longer, planned and with more functions inside them.
Horizontal, not vertical
The authors did not settle for the correlation: they built a model of the assignment between CEOs and firms and estimated it on the data —a model that separates the effect of the CEO from the effect of the company that chose him. The result is what makes the whole piece of work honest: there is no superior type. Most firms run by "managers" are as productive as those run by "leaders". The damage is not in being one thing or the other: it is in being the wrong one for the company you happen to run. Anyone who read Competing in turbulence will recognise the structure: there is no good method, there is a method that fits the terrain. Here it is the same thing, applied to one person's seven days.
And that damage has a size: according to the study's own structural estimate, a little more than one in six firms ends up with the wrong type of chief executive.
And which one does yours need? The rule is short, and the study itself hands it over: the larger the company, the more functions that have to coordinate for an order to go out complete, and the more outside counterparts that weigh as much as partners —a large customer, a bank, a head office, a new market—, the more it needs the "leader" shape. The other way round: the more the business is won inside the plant, with few functions and few counterparts, the more the "manager" shape is the right one. And if nobody else shares the running of the company —if there is no second-in-command with real managerial responsibility—, the company is asking the owner for both shapes at once, which is the most common way of ending up with neither.
On that 7% in higher sales, the precision almost nobody gives: the performance difference appears only after the CEO is hired —it is not that the better CEOs end up at the better firms— and it does not appear straight away. It takes about three years to show.
Those three years are, for a board or a family council, the most actionable thing in the study. Replacing the number one every eighteen months —the most common reflex when the numbers do not add up— guarantees by design that the effect of either shape will never be seen: it is taking the cake out of the oven after ten minutes, three times in a row, and concluding that the oven is broken.
The inherited calendar
If the error is one of matching, there are two ways out and it is worth putting them on the table before you make the diagnosis of your own week. One: change the shape of the week, which starts on Monday with the five dials. The other: change who shares the running of the company —a second-in-command with real managerial responsibility, and a dashboard that lets the owner see without being there.
The second has modest support, and it is worth stating its size: in the working-paper version of this same study (NBER Working Paper 23248, not in the published one) the index is higher where a COO exists. It is a correlation, not a recipe, but it is enough to reframe the discussion: in the mid-sized company, the second-in-command is argued about as an org-chart decision —whose turn it is, where the nephew ends up— when it is a decision about the shape of the owner's week, in disguise.
With the exits in plain sight, the map of where each shape shows up can be read without stinging.
| Characteristic of the firm | What happens to the "leader" behaviour index | What it means for the number one's week |
|---|---|---|
| Larger | Higher, and it survives the controls | The shape that worked with 80 people stops working with 300, even if the owner is the same |
| Multinational or part of a group | Higher, and it survives the controls | External counterparts who have to be seated next to the internal ones, not in separate meetings |
| A COO exists | Higher, in the working-paper version and as a correlation | It is a decision about the design of the week, not a prize for a manager |
| Family-owned and run by a family CEO | Lower (significant without controls; it stops being so once you control for size and complexity) | The owner-manager's default shape is the "manager" one, and the company usually needs something else |
| Brazil and India, against France, Germany, the United Kingdom and the United States | Significantly lower | It is not local folklore: it shows up in the data, with Brazil as the study's second-largest national sample |
Where each calendar shape shows up (and what it means for your week). The drop in the index with a family CEO is significant on its own and stops being so once you control for size, multinational status and group membership —the same happens with stock-market listing—: much of the "family effect" is a size and complexity effect. Source: Bandiera, Hansen, Prat and Sadun, JPE 128(4), 2020.
A caveat before importing anything: Argentina is not in the sample. The six countries are Brazil, France, Germany, India, the United Kingdom and the United States. The proxy is Brazil, the study's second-largest national sample, and the bet —and it is a bet, not a fact— is that a middle-income economy with mid-sized, closely held family firms looks more like that half of the sample than like the German one. What travels is the matching logic, not the percentages.
Caveat made, here is the regional finding. Estimated demand for "leader"-type leadership is practically the same in low- and middle-income countries as in rich ones —if anything, slightly higher in the poorer ones. Supply is not: among the firms that need a "leader"-type number one, 35.6% fail to get one in low- and middle-income countries, against 5.4% in rich ones. It is not that our companies need less of that shape of leadership; it is that they get less of it.
From the same team and the same diaries: family CEOs work 9% fewer hours than professional ones, controlling for qualifications and firm characteristics (Bandiera, Lemos, Prat and Sadun, Review of Financial Studies, 31(5), 2018). It is not a reproach, it is a labour-supply fact with a practical consequence: if the number one is going to put in fewer hours, the shape of those hours matters more, not less.
Put the three things together —owner-manager, mid-sized firm, region— and you have the portrait. The calendar was built when the company was half its current size and had no markets abroad; it worked so well that the company grew. And nobody revisited it, not out of negligence but because there is nobody whose job it is: it is not on the org chart, it is not in the budget, there is no committee that audits it. That is the inherited calendar.
You do not let go of what you cannot see
Suppose the reading comes out badly: the shape of your week is not the one your company needs today. The objection arrives on its own, and it is always the same sentence.
"I would delegate. But there is nobody here to delegate to."
It is the same class of sentence as "my calendar runs me", which in Part 1 did not survive the stopwatch. The evidence suggests that the obstacle is usually not the supply of managers, but that the owner has no way of knowing whether the manager is doing the job well. And nobody lets go of what they cannot see.
The experiment is the same one we cited when we wrote about the structure a company outgrows, in article 02, and that time we kept the headline figure. Bloom, Eifert, Mahajan, McKenzie and Roberts (Quarterly Journal of Economics, 128(1), 2013) gave away free management consulting to randomly chosen textile plants. What we left out is what matters today: the owners ended up delegating hiring, investment and pay decisions because better information allowed them to monitor. The people available in the Maharashtra labour market did not change. What changed was what the owner could see from his desk.
That turns a problem of character —"people here do not commit"—, unsolvable and therefore comfortable, into a problem of design, solvable and therefore uncomfortable. It is not trust that is missing: it is information, and information is built out of a metric, a cadence and a format. None of the three is assembled in a calendar chopped into short one-on-one blocks.
Your week is a capital allocation decision
Three moves. Not one of them needs a project or a budget.
One: the twenty-minute read. Export a single week of your calendar to Excel —last week, exactly as it turned out— and build one column per dial: duration, more than twenty-four hours' notice, participants, distinct functions and who was sitting there. You do not need to remember anything: almost all of it is in the invitation. Then, three calculations: what share of your meetings had three or more functions at once, what share was booked with notice, and how many times there were insiders and outsiders in the same room. The higher all three, the closer your week is to the "leader" shape. Do not expect a number that is comparable with the study's: the index is estimated over an entire sample, not over a single calendar. What you need to know is not your percentile, it is which way you fall.
The long version, for anyone who gets hooked: two weeks logged in fifteen-minute blocks —the study's own unit— and completed at the end of each day. It gives a better picture. The twenty-minute one already decides.
Two: the matching question. Not "is my calendar any good?", which is a question with no answer, but "what kind of company am I today?": how many people, how many functions have to coordinate for an order to go out complete, how many markets, whether or not there is a second-in-command with real managerial responsibility. And only then: does the shape of my week look like the one that company needs? Today's company, not the one from five years ago.
Three: the three columns. Every block falls into one of them. (a) What only the number one can do, and the test is simple: if someone else does it, the outcome changes, not just the speed. (b) What gets delegated with a dashboard: delegating without the data that makes monitoring possible is not delegating, it is abandoning, and three months later it comes back more expensive. (c) What should not exist at all: that column is best looked at last, once you can no longer negotiate it with yourself.
The number one's week is the only asset in the company that gets allocated every single week without anyone asking for a return analysis.
Questions for Monday
The first five can be answered from memory, right now, without opening anything.
Five from memory, two with the calendar open
- Of the decisions that crossed your desk last week, which came to you because they are important and which because nobody else has the data?
- How many of last week's meetings brought three or more functions to the same table —and how many were the third one-on-one this month with the same person?
- What kind of company is yours today —functions that have to coordinate for an order to go out complete, markets, outside partners— and what kind was it when you built, without noticing, the calendar you are still using?
- If the shape of your week and the shape of your company do not match: which three blocks do you change this week, and who takes the place you leave open?
- When was the last time anyone —the board, the family council, you yourself— assessed in writing the shape of the number one's calendar against what the company needs today?
And the long version, for anyone who is game:
- Export last week and mark it up with the five dials. Looking at the result, which way does your week fall?
- For every decision that climbs to your desk only because nobody else has the data: what information would have to exist, at what frequency and in whose hands, for you to be able to let it go next month?
Go back to the two men at 7:20 on Tuesday. Using the rule just given —size, functions that have to coordinate, outside counterparts, second-in-command—, and as a reading rather than a verdict, the one who is probably mismatched is the first: 240 people, several functions that have to coordinate for an order to go out complete, a large customer who already weighs as much as a partner, no second-in-command, and an entire week built in half-hour, one-on-one blocks. His company stopped being the one he was running when he built that calendar; the calendar never found out. Nor is the second man "right" simply for being the second: he would be just as mismatched if his business were still won on line 3. The error is paid for in both directions and it is corrected with the same five dials.
That leaves the old sealing machine. If the owner stops walking the line at 7:35, who hears the strange noise? The answer cannot be "nobody", and it cannot be "let him keep walking the line every day" either. It is that somebody else hears it and that the fact reaches the owner's desk the same day. That is a dashboard: it costs less than twenty-one meetings a week, and it is the only thing that keeps letting go from meaning going blind.
That closes "The CEO's time": Part 1 handed you back the pen to your own calendar; this one showed you that the calendar has a shape, that the shape can be measured, and that the expensive error is not the shape but the mismatch. There are two doors out of a mismatch and there is no third: change the shape of the week, or change who shares the running of the company. Which leaves on the table the question the study does not answer: if matching explains that much, how do you choose the person? The usual answer —"I have a good eye for people"— is an autobiography, not a method. That is what the next article is about.
Has your calendar gone more than three years without changing while your company changed?
32sur works both doors. In ninety days what changes is the owner's week: fewer and longer blocks, planned two weeks ahead, with the functions that have to decide together sitting together —and the decisions that climb to your desk today because nobody else has the data, moving down with a dashboard behind them and an owner in front. The first step is a ninety-minute session with the last two weeks of your calendar exported and open: you walk out with the shape of your week measured, the shape of the company you have today, and the three blocks you change the following Monday. If it has been more than three years since your calendar changed and your company changed anyway, let's talk.
References
- Bandiera, O., Hansen, S., Prat, A. and Sadun, R., "CEO Behavior and Firm Performance", Journal of Political Economy, 128(4), 2020, pp. 1325-1369 (earlier version: NBER Working Paper 23248) — diaries of 1,114 manufacturing CEOs in Brazil, France, Germany, India, the United Kingdom and the United States; 42,233 activities = 225,721 fifteen-minute blocks; an unsupervised topic model (Latent Dirichlet Allocation) run over 654 combinations of activities (98,347 blocks, 78% of interactive time) characterises two pure behaviours: the number of behaviours is set by the authors —who compare the two-behaviour model against models with three up to twenty and choose two for interpretability— and the algorithm determines which mix of activities defines each one. Odds ratios from Table 2, p. 1342 (C-suite 33.90×, communications 1.90×, insiders+outsiders 1.90×, cross-functional 1.49×, shop-floor visits 0.11×, production 0.46×, suppliers 0.32×, outsiders only 0.58×); 7% higher sales per standard deviation of the index; three years until the difference shows up; correlations of the index with size, multinational status and group membership, which survive the controls, and with family CEO and stock-market listing, which do not. In the data description (section II.C), 40% against 65% of time devoted to meetings between the bottom and the top quartile of the distribution of meeting time itself: sample dispersion, not quartiles of the behaviour index. The correlation of the index with the existence of a COO does not appear in the published version: it is in NBER Working Paper 23248, and as a correlation. It is the backbone of this article.
- Bandiera, O., Hansen, S., Prat, A. and Sadun, R., ibid., structural estimation section and results by region — horizontal differentiation with matching frictions: 17% of firms end up with the "wrong" type of CEO; among the firms that require a "leader"-type CEO, 35.6% of those in low- and middle-income countries are badly matched, against 5.4% in high-income ones (the paper itself describes the ratio as "more than six times"); 21.6% against 49.5% of "leader"-type CEOs by region, with estimated demand practically identical; the loss from misallocation is equivalent to 13% —more than an eighth— of the labour productivity gap between rich and poor countries. Supplies the matching thesis and the regional angle.
- Bandiera, O., Lemos, R., Prat, A. and Sadun, R., "Managing the Family Firm: Evidence from CEOs at Work", Review of Financial Studies, 31(5), 2018, pp. 1605-1653 — on the same diaries: family CEOs work 9% fewer hours than professional ones, controlling for qualifications and firm characteristics, and that difference accounts for roughly 18% of the performance gap between family and non-family firms. Supplies the figure that puts a price on the owner-manager's calendar.
- Bloom, N., Eifert, B., Mahajan, A., McKenzie, D. and Roberts, J., "Does Management Matter? Evidence from India", Quarterly Journal of Economics, 128(1), 2013, pp. 1-51 — randomised experiment providing free management consulting to textile plants in Maharashtra, the textile belt around Mumbai: +17% productivity in the first year and more plants opened within three years; the mechanism was that owners delegated hiring, investment and pay decisions once better information allowed them to monitor. Supplies the mechanism (not the figure, already cited in article 02): you delegate when you can see.
- Mintzberg, H., The Nature of Managerial Work, Harper & Row, 1973 — for forty-four years, the largest direct-observation exercise on a chief executive's calendar: five executives. Supplies the scale of the prior ignorance and is the benchmark for the diary method. It is discussed in detail in Part 1 of this series.