Manage for the Bad Case

The Core Insight

Loudcloud went public on 1.94 million dollars of trailing six-month revenue, against a 75 million dollar forecast for the next year. BusinessWeek called it the IPO from hell. The bankers cut the price to 6 dollars, and Ben Horowitz sold anyway. Neither underwriter held a closing dinner.

The company carried 477 employees and no private investor at any valuation. Horowitz sold the operating business to EDS for 63.5 million dollars in cash and kept the software. The rest became Opsware, trading at 35 cents, half the cash in the bank. Five years later Hewlett-Packard paid 1.65 billion dollars in cash.

Most management books describe a company that is working. Horowitz calls those peacetime books, written by consultants, and counts one exception in Andy Grove. He wrote the book for the case where the plan has failed.

A quarter forecast at 100 million dollars booked 37 million. Three layoffs removed 400 people. He offers no recipe for that. The discipline is to tell the truth at the worst moment and to decide alone against the crowd. Under it sits the machinery that keeps a company alive long enough to get lucky.

The Framework

The mode comes first, and it changes every rule under it. Peacetime means a large advantage in a growing core market. Wartime means an imminent existential threat and one bullet in the chamber. Horowitz counts his own tenure as peacetime for three days and wartime for eight years.

• Peacetime follows protocol, and wartime violates protocol to win.
• Peacetime pushes detail decisions down, and wartime cares about a speck of dust on the prime directive.
• Peacetime builds a recruiting machine, and wartime builds one that can run a layoff.
• Peacetime trains people for career development, and wartime trains them so they do not get shot.

The job splits in two under the mode. Knowing what to do is one half, and getting the company to do what you know is the other. Ones decide on thin data and find process design boring. Twos want clear goals and structured accountability, and thinking time makes them anxious. Ones left alone produce chaos, and Twos left alone fail to change direction.

A One CEO staffs with Twos and with Functional Ones, who run a function as Ones and sit on the executive team as Twos. A second true One splits the company.

Skipping the second half of the job accrues management debt, an expedient short-term decision with an expensive long-term consequence.

• Two people run one organization, so nobody owns the slipped schedule.
• A key employee arrives with an outside offer and gets paid above better engineers, and the secret always escapes.
• No feedback process exists at twenty-five people, so direction arrives without correction.

The larger payment is a silent tax, because people rarely repair a weakness they cannot see.

Key Ideas

Bad News Must Travel Fast

His largest improvement as a CEO came on the day he stopped being too positive. He had assumed that he absorbed bad news better than anyone, and he took it hardest. He hoarded the problems he had no way to fix.

Three mechanisms pay for the transparency. Trust comes first, because the communication a company needs is inversely proportional to the trust inside it. Brains come second, because a brain cannot solve a problem it does not know about. Culture comes third, and it runs as a routing protocol: bad news travels fast and good news travels slowly. Asking people never to bring a problem without a solution buys silence on every problem nobody has solved.

After the EDS sale he took 80 employees to a Santa Cruz motel, told them where the company stood, and issued new grants. Two quit. Of the other 78, all but two stayed until the Hewlett-Packard sale.

Managers Lay Off Their Own People

Doug Leone had never seen a company recover from consecutive layoffs to an outcome above a billion dollars. Horowitz ran three, and they removed 400 people.

The procedure runs in six steps. Get your head right, because the past overwhelms you. Do not delay, because a leak forces every manager to look stupid, to lie, or to go silent. Be clear about the cause, which is that the company failed to hit its plan and excellent people are leaving because of it. A layoff breaks the trust a CEO trades on daily, and owning the failure is what rebuilds it.

Then train the managers, because managers lay off their own people. The work does not go to human resources, to a more confrontational peer, or to an outsourcing firm. Each manager explains briefly that the company failed rather than the person, states that the decision is final, and answers every question about severance.

Address the whole company before the individual meetings. Campbell put the reason in one line: the message is for the people who stay. Then stay in the building, talk, carry boxes, and thank the people leaving.

Firing an Executive Is a System Failure

Horowitz frames a failed executive as a failure of his own hiring system. The job went to the candidate with no sharp weaknesses instead of the one with the strength the company needed. No great head of sales exists in the abstract, only the right one for the next twelve to twenty-four months.

The board learns individually and by phone, starting with whoever made the referral. Severance gets approved first, because an executive takes roughly ten times longer than an individual contributor to find a job.

Script the conversation, and read the written performance record first. A CEO who says he has decided ends the meeting, and one who says he thinks reopens it. Campbell set the standard: you cannot let him keep his job, and you can let him keep his respect. The direct reports hear it within hours, then your staff, then the company.

A demotion adds two costs. He carries embarrassment, because everyone he knows knows his title. He carries betrayal, because you are no better at your job than he is at his. Decide the new role before the meeting, because leaving him under the new boss invites sabotage. Be at peace with him quitting, since a CEO who cannot afford to lose him cannot make the change. Admit that a founder-CEO is equally underskilled, and where it fits, couple the demotion with a raise.

Training Returns 200 Hours for Twelve

McDonald's trains its people, and startups with far harder jobs do not. Productivity comes first, because companies count hires and never count the employees who became fully productive. Performance management comes second, because an untrained employee gives you no basis for a firing. Product quality comes third, because hiring fast without training turns a clean architecture into a Frankenstein. Retention comes fourth, and every exit interview at Netscape gave two reasons: they hated their manager, and they were not learning.

Andy Grove supplies the arithmetic. Four lectures at three hours of preparation for each hour of class costs twelve hours. Ten students then work about 20,000 hours in the following year. A 1 percent improvement returns 200 hours against your twelve.

Functional training runs from a one-page expectations document up to a multiweek boot camp. Management training is where you set expectations for one-on-ones, feedback, and objectives. A manager gets no new headcount until a training program exists for the people he wants to hire. The CEO teaches the management course himself, because being too busy to train is the moral equivalent of being too hungry to eat.

The Big-Company Executive Has the Opposite Job

An executive from a large company is interrupt-driven. Most of his work arrives, and more than three new initiatives in a quarter is too many. A startup executive originates eight to ten new initiatives a day, because nothing moves on its own.

He waits for the email, the call, and the scheduled meeting, and your people start asking what he does. Large scale rewards prioritization and process improvement, and a startup rewards constant initiation.

Three screening questions separate them. Ask what he does in his first month, and a plan built on learning and incoming work fails. Ask how the new job differs, and a candidate who expects his experience to transfer fails. Then ask why a small company at all, and equity as the motive fails, because one percent of nothing is nothing.

Integrate aggressively. Give monthly, weekly, and even daily objectives, so he creates rather than waits. An executive with no questions is a candidate for firing. One who is not up to speed in thirty days gets fired.

Horowitz learned what he wanted by doing the jobs, as acting head of human resources, acting chief financial officer, and acting head of sales. The decision itself is lonely, because consensus decisions about executives drift toward the candidate with no weaknesses. Every executive but one and every board member voted against Mark Cranney, on his school, manner, and appearance. Under him, sales rose more than ten times over and the market cap rose twenty times.

Politics Is Something the CEO Causes

Politics is any advance of a career or an agenda by means other than merit and contribution. The least political CEOs often run the most political organizations, because the generous answer creates the incentive.

An executive brings an outside offer and asks for a raise, and you grant it. You have now paid for asking rather than for performing. Others agitate, the money becomes first come first served, and the quieter and often more competent people pay for it.

Two mechanisms hold it down. Hire for ambition aimed at the company, with personal success as a by-product. Then put strict processes on the activities politics runs through. Compensation moves on a structured review with the board involved in executive pay. When someone lobbies for scope, say as little as possible, because anything you say becomes cannon fodder and leaks. Run the reorganization immediately, which leaves no time for lobbying.

Promotions run on a formal, visible, defensible process, because peers read each one as merit or as favor. The Law of Crappy People is the deeper reason. Talent at any title level converges to the worst person holding that title, because everyone a level below benchmarks against him. Define the skills at each level crisply, and name a person rather than a generic competence. Run a promotions council across groups, so one chain of command cannot mint five vice presidents against one in engineering.

Practical Applications

Tell the company your worst current number this week, with the plan and the risk attached. The routing protocol is built from what you did with the last piece of bad news.

Write the layoff script before you need one: one page for the company meeting, one page for each manager, and the severance detail attached. The manager delivers it to his own people, in person, on the same day.

Audit the management debt on one sheet. Count the organizations with two people in charge and the salaries you raised because somebody produced an offer. Add the people who never had a written performance conversation.

Put a training program in front of the next requisition. The manager who wants a hire writes the one-page expectations document first, and you teach the management course yourself.

Before hiring an executive into a job you have never done, do the job. Write the strengths you need and the weaknesses you accept, then check the references yourself.

Who This Is For

Founders running a company in trouble get the most from it. The material assumes employees, a board, a plan you are missing, and a decision you have avoided for two weeks.

A solo founder or a team of five can skip most of it. The machinery starts to matter where you stop knowing everyone's work at first hand.

Three limits are worth naming. The advice comes from enterprise software between 1999 and 2007, where large companies were the customers. Horowitz now runs a venture firm, and the book works partly as a calling card, which shapes which stories get told. The wartime frame also gets borrowed to excuse ordinary bad management. A CEO who declares war gets to skip the explanation, and this book supplies the vocabulary.

The Decision

Name the hardest conversation you are avoiding right now. It is usually a demotion, an executive who is not working, or a number the company has not heard.

Then write the first three sentences you plan to say. The first names the company failure rather than a personal one. The second says that you have decided, rather than that you think. The third holds the settled part: the severance, the new role, or the date.

Say them within the week, to the person, in the room. A conversation you cannot write down is a decision you have not made.