Part three of four on the assumptions under your Decision Stack. Part one, Assumption #1: If our strategy is coherent, it's right., argued that every layer has assumptions underneath it. Part two, Assumption #2: We're a data-driven company. We test everything., argued our tools only test the bottom one. You don't need to have read either.
In 2011 Netflix split its DVD and streaming businesses, called the DVD half Qwikster, and lost 800,000 subscribers in a quarter.
Reed Hastings' strategy assumption was right. Streaming would grow, DVDs would decline, and separating them made sense on paper. The assumption underneath it was the one that broke: that customers would accept two accounts, two bills and two websites to get what they'd been getting from one.
Afterwards, one of his own managers told him: "I knew it was going to be a disaster, but I thought, 'Reed is always right,' so I kept quiet."
That sentence was meant as a compliment. It cost the best part of a million subscribers.
Decisions flow down. Evidence has to climb.
And almost nothing in a normal company is built to help it climb.
That's why naming your assumptions isn't enough on its own. You can write down the belief holding up your strategy, put a name and a date against it, do everything part one asked, and still never hear the thing that would tell you it's failing. Because the person who can see it failing is four levels below you, and they've already done the maths on what saying so would cost them.
Nokia's middle managers knew Symbian was losing. Timo Vuori of Aalto University and Quy Huy of INSEAD interviewed 76 of them after the collapse and found people who understood exactly what was happening and said nothing, partly because the executives above them were known to shout. Senior leadership held the line because a new operating system would take years and, as one of them put it, somebody had to keep the faith.
Stephen Elop was the outsider Nokia hired from Microsoft to run the company in 2010. Five months in, he wrote it down in the burning platform memo, and the sentence was: "Our competitors aren't taking our market share with devices; they are taking our market share with an entire ecosystem."
Everybody in the building already knew. The memo wasn't news. It was too late.
Three reasons evidence doesn't climb
The messenger pays and the assumption doesn't. If you're right, you've embarrassed somebody senior. If you're wrong, you've marked yourself as the person who doesn't get it. The expected value of speaking up is negative, and everyone in your organisation can do that sum in about four seconds.
The assumption is somebody's identity. Andy Grove described Intel's core assumption as the "self-evident truth that Intel was a memory company." Not a belief. A truth. Nobody contradicts a self-evident truth in a meeting. They just get quieter, and the quiet reads as agreement.
Quite often there's simply no route. This is the one people underrate, because it isn't about courage at all. A finding at the bottom of the stack needs somewhere to go that can actually invalidate a commitment at the top, and in most organisations no such channel exists. Boeing's engineers had the test data that contradicted the bet. There was no path from that data to the person who'd signed the contract.
The afternoon that saved months
At Huddle, a long time ago now, we were experimenting with empowering cross-functional teams, which tells you how long ago it was. A few of us had spent weeks researching and planning a new initiative.
Then we did something deliberate. Instead of handing the team a spec, we set up the kick-off to share everything we knew about the problem, the customer, and what we were actually trying to achieve, and then invited the whole team to help design how best to tackle it.
A junior developer put his hand up. "Oh, if that's what we're trying to do, why don't we just build X?" It was obvious once he said it, but none of us had seen it, because none of us had been looking from where he was standing.
I've told that story for years as a lesson about the power of clear objectives, and it is one. But it's also a story about a route. His answer existed before that meeting. What didn't exist was a room where a junior developer was expected to contradict a plan his seniors had spent weeks on. We'd built that space, on purpose, and it paid for itself in an afternoon. I've worked with plenty of organisations since where he'd have kept his hand down, and they'd never have known what it cost them.
What actually makes evidence climb
Ask for the specific thing, not for feedback. "Any concerns?" gets you nothing, and it deserves nothing, because it puts the whole burden on whoever is brave enough to go first. Try this instead: what have you seen in the last month that doesn't fit our story about X? That's a request for data, and people will hand you data all day long.
Make dissent somebody's job for twenty minutes. Gary Klein's premortem needs no training and no consultant. Tell the room the plan has already failed spectacularly, have everyone write down why in silence, then go round the table. The silence matters, because it stops the first confident voice setting the frame for everybody else. And it works because nobody has to attack the plan. They've been handed a diagnosis to perform instead.
Watch for the careful version. This point is David Bland's. Where there's no safety in the room, teams don't stop doing the work, they do it carefully. They map the safe assumption, run the safe test, and look at the data least likely to embarrass anyone. Everything gets documented, nothing gets risked, and the map comes out confirming what the room already believed. Going through the motions is a different failure from staying silent, and it's a great deal harder to spot.
Give every assumption an owner who is obliged to be told. In part one the name against the assumption was about accountability. Here it's an address. A name against an assumption isn't about blame, and it isn't about who gets to decide. It's about there being one specific person whose actual job is to receive the bad news, so that a junior team member with an uncomfortable data point knows where to send it.
And when you're stuck, borrow Grove's question from part one. If the board brought in a new CEO tomorrow, what would they change? It lets people say the unsayable thing by pretending to be somebody else.
If you're the one at the top
The uncomfortable version of all this is that "Reed is always right" was a compliment.
Your team isn't withholding evidence because they're disloyal or timid. They're withholding it because you've been right a lot, and every time you're right the cost of contradicting you goes up a little. A long track record is an asset on the balance sheet and a tax on everybody below you. Nobody sends you the tax bill.
Hastings' answer was farming for dissent. Not waiting for disagreement to show up, but going out and harvesting it deliberately, as a standing management practice. It's a good phrase because farming is work. You don't farm by leaving the gate open and hoping.
The practical test is embarrassingly simple. When did somebody last tell you something you didn't want to hear? If you can't remember, that isn't evidence you've been right. It's evidence the route up is broken.
If you're anywhere else
Then you're probably already sitting on something.
In Seeing Around Corners, Rita McGrath borrowed a line from Andy Grove for this: snow melts from the edges. The first evidence that an assumption is failing turns up at the periphery, with the people closest to customers, and it moves inward slowly if at all. When Grove and Gordon Moore finally looked at Intel's memory business in 1985, only one of their eight fabs was still making memory. The edges had already moved. Only the top of the company hadn't noticed.
So you've noticed a number that doesn't fit the story. You've heard the same thing from three customers in a row. You've watched a rule get applied to a case it was obviously never written for, and you've assumed somebody more senior must have clocked it.
They probably haven't. And the ones who have are waiting for someone else to go first, which is the whole reason nobody ever does.
So say it out loud, and say it as a specific observation rather than a challenge: I'm hearing something that contradicts our assumption about X. You're handing them a piece of evidence and letting them do what they like with it. That framing costs you almost nothing, and it's the version that actually travels.
Somebody already knew
Nokia's middle managers knew. Netflix’s managers knew. Intel's fabs had already re-based themselves on reality while the top of the company was still calling itself a memory company. Boeing's engineers had the data.
In not one of these cases was the information missing. It just had nowhere to go.
So here's the question, and it works whichever end of the stack you're standing at. Who in this building already knows we're wrong, and what would it cost them to tell me?
This is part three of four on the assumptions under your Decision Stack. Part one is Assumption #1: If our strategy is coherent, it's right.. Part two is Assumption #2: We're a data-driven company. We test everything.. Part four is Assumption #4: The board wants to see the roadmap., on taking the assumptions to the room that owns them (coming next!).
