You spent three weeks on the pricing analysis. It is right. You presented it, everyone agreed it was interesting, and six weeks later the company is still pricing the old way.
Nobody rejected it. A conclusion that nobody runs has the same value as a conclusion that was wrong, and it cost more to produce.
Why these three models
The decision is how to get a correct piece of work acted on, and how much of your effort to spend on that rather than on being more correct. The features that fire are a proposal that sits some distance from what people currently believe, a set of people who influence each other rather than deciding independently, and a choice where the better option does not automatically win.
Three lenses. Position produces an equilibrium answer about how far your proposal sits from the prevailing view. Spread produces a cycle answer about how a conclusion actually travels. Noisy choice produces a complex answer about why being right by a margin does not translate into being chosen. The first is the one you can act on today and the one nobody treats as part of the work.
1. Your proposal has a position, not just a truth value
Imagine the range of views your colleagues hold on the question, laid out on a line. Somewhere on that line is the current consensus. Your proposal occupies a point, and the distance between those two points does more work than the quality of your reasoning.
The picture is borrowed from models where options sit at positions and distance from a decision-maker’s current position affects what gets chosen.3 It is an analogy here rather than a measurement, and it earns its place because it makes one thing visible that the usual framing hides: your proposal has a location, and you chose that location without noticing.
Two things follow from thinking about it this way, and both are uncomfortable for anyone who believes correctness should be sufficient.
The first is that a proposal far from the consensus is not adopted in one step, however well evidenced. The people who would have to act on it are being asked to revise several beliefs at once, and each revision has its own cost in credibility with their own colleagues. A proposal that is a long way out gets called interesting, which is the word an organisation uses for a thing it is not going to do.
The second is that you can choose your position. Most analyses are presented at the point the analysis reached, which is the furthest point the evidence supports. That is the correct place to end the thinking and rarely the right place to start the conversation. Presenting the nearest version that still changes the decision is not dishonesty, provided you say where the full analysis goes and why you are proposing a step rather than the whole distance.
Distinguish distance from the consensus from distance from a person. A proposal three steps from what the company believes is a slow adoption. A proposal one step from what the company believes and directly contrary to something a specific colleague designed is a different problem wearing the same clothes, and moving your position closer does nothing about it. The tell is whether the objections change when you moderate. If the same person produces a new objection each time you concede, the distance was never the issue, and the next move is a private conversation rather than a revised slide.
There is a further reason distance predicts reception better than evidence does, and it is not about persuasion at all. The interventions that would change a system most are the ones it resists hardest, because they are the ones that require the most existing structure to be given up.4 The resistance you meet is often proportional to how right you are, which is why being more right does not help.
The practical test is one question asked before you present: what is the smallest change that would still be worth making. If the answer is much closer to the current practice than your headline recommendation, lead with it and keep the rest as the direction of travel.
The near version has its own failure mode, and it is the common one. The small change gets adopted, it works, and the question is then treated as settled. The remaining distance disappears, not because anyone argued against it but because the problem stops feeling open. The defence is to attach the next step at the moment of adoption: name the measurement that would trigger the second move, and the date somebody will look at it. A step with no stated successor is not a sequence. It is a concession that closed the conversation.
2. Conclusions travel person to person, not from the front of a room
The second lens explains why the presentation itself does so little.
A conclusion spreads through an organisation the way anything else does through a network of people who talk to each other: by transmission between individuals who are already in contact, at a rate that depends on how many contacts each has and how receptive they are.1 A meeting is one simultaneous exposure to everyone at once, which is a poor way to start a spreading process, because everyone hears it in the same room in the same tone and their first reaction is formed in public.
The room is where a proposal is judged, and the corridor is where it is adopted.
Which suggests an ordering that most people run backwards. Before the meeting, have the conclusion in three or four separate conversations with people who will each carry it to different parts of the company. Not to build a coalition in the political sense, but because that is the only mechanism by which a conclusion becomes something people have already partly accepted before they are asked to accept it publicly.
The people to pick are not the most senior. They are the ones whose contacts do not overlap with each other, because a conclusion that reaches four people in the same team has reached one place. This is the same reason a diverse ensemble beats a strong single source: what you want is coverage, not endorsement.
Sequence those conversations as deliberately as you choose the people. Start with whoever is most likely to find the flaw rather than whoever is most likely to agree, because a conclusion that has already survived its hardest reader arrives in the second conversation stronger, and you still have time to repair it. End with the person who talks to the most people, so that the repaired version is the one that circulates.
Expect the conclusion to reach the decision in a form you did not write. That is not a failure of the mechanism, it is the mechanism, and a version compressed by three retellings is what the person who acts will hear. So give the carriers the sentence you want repeated: short enough to survive the compression, with the number already in it. If you cannot write that sentence, the analysis is not finished, whatever the deck looks like.
One caution that belongs here. Pre-socialising a conclusion is legitimate when you are genuinely open to it changing and manipulative when you are not. The difference is observable to everyone involved, usually within one round, and once you are read as running a process rather than asking a question, the mechanism stops working for everything you propose afterwards.
3. Better does not mean chosen
The third lens is the one that stops you from treating adoption failure as evidence of stupidity.
People do not deterministically select the highest-value option. They choose better options more often than worse ones, in proportion to the size of the gap, with noise that never disappears. A small advantage translates into a small increase in the probability of being chosen, not into certainty.2
So a proposal that is fifteen percent better than the status quo will lose regularly, and losing does not mean the analysis was ignored. It means the margin was inside the noise of how humans and organisations actually pick.
Two consequences. First, if your recommendation is only marginally better, the honest thing is to say so and to stop spending effort on adoption, because the effort exceeds the value of the difference. Much organisational energy goes into fighting for changes whose full benefit is smaller than the fight.
Second, where the advantage is large, say the size out loud rather than leaving it implicit. A proposal presented as better gets treated as a coin flip with a slight bias. A proposal presented as three times better on a number the audience already tracks moves the choice probability much further, and the number is usually available in the analysis you already did.
The noise is not a constant, and one part of it is yours to move. Choice runs closer to random when the options are hard to compare, and comparability is a property of how a thing is presented rather than of the thing. A recommendation stated in a unit the company already reports every week can be set against the status quo in one glance. The same recommendation stated in a unit the analysis invented cannot, and it will be treated as a matter of opinion by people who are not being obtuse. Convert into their unit even when yours is the more accurate one, and state what the conversion assumes. The precision you give up is smaller than the probability you gain.
What the three say together
- Name the smallest version that still changes the decision, and lead with that rather than with the furthest point your analysis reached.
- Have the conversation three or four times before the meeting, with people whose contacts do not overlap.
- State the size of the advantage in a unit the audience already tracks.
- If the advantage is small, say so and drop it. The adoption effort is a real cost and it should be justified by the size of the gain.
Where they disagree
The position argument and the size argument pull against each other.
The position argument says move your proposal closer to the prevailing view so that it can be adopted at all. The size argument says the probability of being chosen rises with the size of the advantage, and the nearest version that still changes the decision is by construction the version with the smallest advantage. Moderate the proposal and you improve its reception while weakening the very thing that makes people choose it.
The resolution depends on whether the change is reversible. Where a small first step can be built on, take the near version, because adoption of a small change buys you a new consensus position to argue from next quarter and the sequence gets you further than the single attempt would have. Where the change is all or nothing, a partial version is worse than useless, because it spends the credibility and delivers a fraction of the benefit, and you should either make the full argument or wait until conditions make it easier.
What none of them contain
None of the three prices the possibility that you are wrong. All of them treat your analysis as correct and the organisation as the obstacle, which is the flattering assumption and not always the right one. Sustained resistance from people close to the work is evidence, and the same techniques that get a correct analysis adopted will get an incorrect one adopted just as effectively.
None of them handles the decision that is not yours to influence. If the person who must act has already committed publicly to the opposite, you are no longer running an adoption problem, you are asking someone to bear a reputational cost, and that is a different negotiation with a different currency.
And one property the ensemble will not produce: organisations remember who was right. None of these lenses tracks the value of a proposal that fails now and is recalled in eighteen months when conditions change. Writing the analysis down and dating it costs nothing and is the only mechanism by which that value is ever collected.
The one action that survives the ignorance: before your next recommendation, write the smallest version that would still change what the company does, and have that conversation with two people separately before it reaches a meeting. If both react the same way, you have learned something about the position of your proposal that the meeting would have cost you a month to discover.
Who has to move
Whoever did the analysis, and the effort has to be budgeted at the start rather than found afterwards. The instinct on finishing a piece of work is to present it, which is one exposure at the furthest point from consensus, delivered to a room whose members will form their reactions in front of each other. The cheapest first test costs two conversations and tells you whether the resistance is to the conclusion or to its distance, which are different problems with different fixes.
Sources and notes
- Albert-László Barabási, Network Science, Cambridge University Press. Spreading processes on networks, and the dependence of transmission on contact structure rather than on broadcast, are developed in the chapters on spreading phenomena. Used in section 2 for the mechanism and for the recommendation to choose carriers whose contacts do not overlap. No claim is made about the degree distribution of any real organisation, and none is required: the argument needs only that people talk to some colleagues more than others.
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Quantal response equilibrium, in which players choose better options with higher probability rather than with certainty and the sensitivity of choice to payoff differences is finite, is developed alongside the experimental evidence on deviations from best response. Used in section 3. The claim used here is qualitative: choice probability rises with the size of the advantage and does not become certainty, so a small edge produces a small shift.
- Jeffrey Carpenter and Andrea Robbett, Game Theory and Behavior, MIT Press. Spatial models of choice, in which options are located in a space of positions and the distance from a decision-maker’s current position affects selection, are treated alongside voting and location models. Used in section 1 as a framing for the distance between a proposal and the prevailing view. The application is analogical rather than measured; no position is being estimated numerically here.
- Donella H. Meadows, Thinking in Systems: A Primer, Chelsea Green Publishing. The observation that an intervention’s effect depends on where in a system it is applied, and that the highest-leverage interventions are also the ones most resisted, supports the framing in section 1 that a proposal’s distance from consensus predicts its reception better than its evidence does. Used for framing rather than for a specific claim.
A note on a number this article does not give. There is no threshold advantage above which a proposal gets adopted. It depends on the switching cost, on who has publicly committed to the current approach and on how much slack the organisation has. What transfers is that the size should be stated in a unit the audience already tracks, and that a small advantage rarely justifies the adoption effort.
Joshua Agonya Pi’Rwot, Founder.