Stripped Threads

Pressure: A number people can move will get moved, and the moving looks like the work.

The estimate was too high.

That was my read of it, anyway. We were putting a proposal together for a client, and the schedule the team handed me seemed inflated and did not match the scope I could see. I went through it looking for the place where somebody had misread the work, and I did not find one. I pushed on it in review and the team held their ground. Dan, the senior engineer whose team had built it, walked me through his reasoning without moving a number.

I read it as a team being conservative, but trusted their judgement. I sent it to the customer.

The feedback came back direct and hard to hear. Our proposal was twenty-five percent higher than every other one on the table, a margin wide enough that it had stopped being a question about methodology and become a question about whether our firm knew what it was doing.

There was nothing wrong with the estimate, though it took me a long time to see that. Everything in that number had been put there for a reason, and to understand the reason you have to go back a year and a half, to a standard I held tightly (and still believe now).

I believe integrity is central to being a great leader. Doing what you said you would do, especially when keeping your word costs you something, is close to the center of the whole thing, and it is what I aspire to. At that firm it was our reputation and it was my own standard, and I did not hold it because it was convenient. I held it because I believed it was most of the reason a client chose us over someone cheaper, and I still believe that today.

Which is what made this lesson so hard to learn. It is a good thing to care about. It is also the reason for everything that follows.

Six months before that proposal, we had finally shipped a different program: a ruggedized field radio for a defense customer, delivered three months late. And the reason we were late was a conversation in Amsterdam the year before, while we were traveling, just before a program review with vendors from around the world.

Dan found me there. The embedded software work on the core module that ran the radio was far more complex than anyone had estimated, and understanding it meant working through a four thousand page manual. He and the team had put a number on it. They needed three more months.

I said no. We had committed to prototypes in nine months. I had built that estimate myself, with the analysis behind it to price the work as a fixed bid, so the date was the economics too. A slip earned us nothing and cost the company money.

But the economics is not what made me certain. What made me certain was that holding the date was the honorable thing to do, and I was the person who held dates. That was not a calculation. It was closer to an identity, and it made the conversation feel already settled before he finished making his case. I walked him through the commitment, the analysis, the fixed bid. I told him the date was not one of the variables available to us and we would have to find another way.

Dan did not argue. He was disappointed, and he went back to work. He is a person of real integrity, so he took the message well and said he would do what he could.

We shipped exactly three months late. Exactly the timeline Dan had given me.

I filed it away as an estimation problem, something we could do better on the next program of that kind. What I had actually done was teach a capable engineer, and everyone who watched what happened to him, that telling me the truth about risk was expensive and pointless. The risk did not go anywhere. It went into the container that was still cheap to use, which was the estimate.

So when Dan’s team handed me an inflated schedule half a year later, they were not being cautious. Dan had told the truth once, been overruled, and then watched the program run exactly as long as he said it would. The second time, he put the margin in the number, where I could see it but could not overrule it.

Dan was not being dishonest. He was protecting himself and his coworkers from me.

I was the source of that problem.

And it did not stop at the estimate. What I had built, without ever deciding to build it, was a team that stopped pushing on how to deliver fast at the quality the work deserved, because I had settled that question for them. Fewer arguments, fewer inventive answers, less appetite for exactly the kind of risk that was the entire job.

I wanted a team that would think aggressively and tell me the truth. What I built was a team that was careful with me. That is the opposite of what the work needed, and nobody imposed it on me. I did it, and I did it while feeling justified. The conviction did not feel like a blind spot from the inside. It felt like standards.

I was holding a virtue, in public, with conviction, and the conviction is what made me deaf to the one person trying to tell me something I could not see.

A belief held too rigidly strips out the same way a screw does.

A screw holds because it is tightened to a point. Past that point the threads strip. The screw keeps turning, smoothly, easily, gripping nothing at all, and from above it still looks perfectly seated. You find out when the joint takes a load.

Nothing strips a thread except overtightening. And once it is stripped, the instinct that got you there, turning it a little harder, does not bring it back.

That is what I did to a commitment I was right to care about. I tightened it past the point where it held anything.

I should say what I cannot prove. I have wondered since whether that twenty-five percent was really the padding, or whether we were simply priced wrong for that class of work, and I cannot rule the second one out. Rate structure explains a lot of lost work and it is the duller explanation, which is usually a point in its favor. What I can say is what I saw inside the estimate and where the margin came from, because I had watched it get put there.

We discounted the price to close the gap, and the customer reluctantly went with us for that program. Then they did not come back for another one. The padding did not lose us the program. It cost us the relationship.

Pattern: When a number decides who is in trouble, it starts reporting on that instead of on the work.

Identify the Pattern

Most numbers inside an organization are doing two jobs at once, and the two jobs pull against each other.

The first is measurement. The number is a read on reality, and its value sits in how accurate it is. The second is accountability. The number is a commitment somebody is answerable for, and its value sits in whether it gets met. A schedule estimate is both at the same time, which is why it corrupts faster than most numbers. So is a forecast, a capacity plan, a defect count, a pipeline number, and much of what sits on a leadership dashboard.

Under pressure, the second job tends to win. Not because people are dishonest, but because the accountability consequence is immediate and personal while the accuracy consequence is distant and shared.

This is among the oldest findings in the field. Steven Kerr laid it out in 1975, in a paper about organizations that reward one behavior while hoping for another, and his point was that people are not confused when this happens.^[1] They read the reward system correctly and act on it. Donald Campbell put the same mechanism in sharper terms a few years later: the more a quantitative indicator is used for social decision-making, the more it will be subject to corruption pressures, and the more it will distort the process it was meant to monitor.^[2] Charles Goodhart had made a version of the observation about monetary policy a few years before that, and Marilyn Strathern later compressed his version into the line most people know: when a measure becomes a target, it ceases to be a good measure.^[3]

The mechanism has been documented at scale. Gwyn Bevan and Christopher Hood catalogued what happened in the English health system under a regime of published targets: ambulance response clocks that started at moments flattering to the number, waiting lists reclassified, patients held in ambulances outside emergency departments so the four-hour clock would not begin.^[4] The reported figures improved. Bevan and Hood were careful to say the extent of the gaming could not be measured, which is part of the point, because the data that reports the improvement is the same data the gaming acts on. Wells Fargo is the familiar one. Regulators fined the bank $185 million in September 2016 over accounts opened without customer authorization, and the cause the investigations kept returning to was a cross-sell target that branch staff could not realistically meet unless they opened more accounts.^[5]

Then there is the inversion, which is the part that catches good leaders. Amy Edmondson’s study of hospital nursing units found that the units with the better team climate recorded more errors, not fewer.^[6] Her follow-up work pointed to reporting climate rather than error incidence as the explanation, though the design could not establish that the underlying rates were the same. Which is the useful shape of it anyway: a metric that improves under pressure is genuinely ambiguous, because the work may have changed or the reporting may have.

A number moving in a good direction is not evidence on its own. It is consistent with two different stories, and it cannot tell you which one you are in.

What It Costs

Four places this lands, none of them in the register of morale:

Decision quality downstream. Most of what you plan, price, sequence, and staff runs on numbers somebody handed you. When those numbers carry hidden margin, you are not making conservative decisions. You are making confident decisions against inputs you have misread, and you will likely attribute the gap later to execution.

The earliest signal, which is also the cheapest one. A risk raised months before the date costs a conversation. The same risk discovered in integration costs the schedule. When raising risk becomes expensive for the person raising it, the risks continue. The early warnings are what stop.

The engineering conversation itself. Once the estimate becomes the place where risk gets stored, it stops being an argument about how to do the work well and fast. My team stopped asking how to deliver aggressively at the right quality, because that question had been settled for them: aggressive answers got overruled and then vindicated at their expense. Padding was the rational move, and it replaced a conversation that was worth more than the margin it protected. That is the cost I would most like back, because a team that has stopped proposing the bold version of the work does not announce it, and you cannot miss what was not brought to you.

Trust, outward, where you cannot see it happening. The number left the building. A customer comparing proposals does not see a team protecting itself. They see a firm that either does not know how long the work takes or is charging them for the uncertainty. We fixed the price. The price was not what they had read.

Reframe the Mindset

One reading gets in the way here, and it is available to any competent leader.

The character reading: the team is sandbagging. It leads somewhere specific, which is pushing harder on the estimate. That is tightening a stripped thread. You will get a number closer to what you wanted and further from what is true, and you will find out on the schedule.

The useful reading is different. A metric that has drifted is reporting on the conditions around it. Read it that way and it becomes informative again, just about a different subject than you thought.

A padded estimate is an accurate measurement of what a team has learned about raising problems with you.

Transform the Tension

Three moves. None of them ask you to break a commitment. A screw is supposed to be tight. The question is how tight, and whether you can still tell.

Ask for the aggressive number, and make it safe to miss. The team should be reaching for a date that is hard to hit, and you should want them to. But a stretch you punish is not a stretch. It is a deadline, and people will price it like one. If nothing on your list slips, you are not finding the edge of what the team can do. You are finding the edge of what they are willing to promise you.

Renegotiate when the conditions change, and do it early. I had this backwards. I thought holding the date no matter what was integrity. But a commitment is made against a set of facts, and when the facts move, the honest thing is to go back to the customer while they still have room to decide. Renegotiating late, with excuses, is a breach. Renegotiating early, with the options and the cost laid out, is what your word meaning something actually looks like. It costs you something either way. It cost me more to stay silent and let the schedule tell them nine months later.

Go back once the outcome is known. The cheapest move on the list, and the one I missed. Dan was right to the month and I never told him so. He drew a conclusion from my silence, and it was a reasonable one. People do not pad because they are withholding effort. They are pricing your future behavior off your past behavior, accurately.

Shift: Which of my numbers can people move, and what have they learned from me about the cost of not moving them?

Reflection (Questions to Ponder)

• Which numbers in my week are a forecast, which are a commitment, and which are being asked to be both at once?
• The last time someone brought me a schedule I thought was too long, what did I do with it, and what did they take from that?
• Have estimates around me drifted in one direction over time, with nothing about the work to explain it?
• When did someone last bring me a risk months early, while it was still cheap? If I cannot remember, that is the finding.
• Who on my team has turned out to be right about something after I overruled them, and did I go back and say so?
• Which of my standards am I holding so tightly that it has stopped being possible to argue with me about it?

Connection (Questions to Discuss)

• Where in our process does a number get adjusted before it reaches the next level, and what is the adjustment protecting against?
• What would it take for someone here to give me a schedule they believed and expected to be held to?
• Which of our numbers do customers and partners see, and what do they read into them that we have not considered?
• What risk are we carrying right now that has not been said out loud because saying it would not change anything?

Subtraction (Things to Stop)

• Stop treating a conservative estimate as a negotiating position. It is data about your own conduct before it is data about the work.
• Stop asking for one number when you need two. The date you want the team reaching for and the date you promise a customer are different objects, and forcing one figure to carry both means you lose the reach.
• Stop reading an improving metric as an improving reality without checking what changed about the measurement. The two look identical in the number itself.
• Stop closing a risk conversation by removing the variable. That does not remove the risk, it relocates it somewhere you have less visibility.
• Stop taking the absence of argument as agreement. A team that has stopped disagreeing with you is cause for concern.

Action (Things to Try)

• Pick one number your team reports upward. Ask the person who produces it what they would have written if it carried no consequence. The difference between the two figures is the margin you have been planning against without knowing it.
• The next time you hold a date against someone’s objection, say the cost out loud in the same conversation: here is what we are accepting, here is who absorbs it, here is what I will do if you turn out to be right. Then do that thing.
• Find one instance in the last year where someone on your team was right and you were not, and go tell them. Not as an apology. As evidence, for the next time they are deciding what to bring you.
• Before your next proposal, plan, or commitment goes to a customer, ask one question about the numbers in it: what does this figure tell someone outside this company about us? The number is going to answer that question whether or not you asked it.

Years later, the arc of this one bothers me more than the late delivery did. One engineer carried both halves of it: Dan told the truth and was overruled, then padded and held his ground. Same person, taught by one conversation, and a team that learned it by watching what happened to him.

I was not careless, and I was not under anyone’s thumb. Good values do not protect you from this. Held hard enough, they can become a mechanism for disconnection. You only get there by turning it too far.

What I do differently is small. When somebody brings me a number I do not like, I find out what the number is protecting before I argue with it, because it is protecting against something I did more often than I would like. And when I hold a line over someone’s objection, I write their name down, because the outcome is going to arrive eventually and I want to be the one who brings it up.