In a study of 1,048 major business decisions, process mattered more than analysis by a factor of six. Twenty-seven prompts that supply the process, not the answer.
Process beats analysis by a factor of six
Dan Lovallo and Olivier Sibony studied 1,048 major business decisions made over five years — new product investments, M&A, large capital expenditures — by asking managers which of 17 practices they had applied. Eight of those practices concerned the quantity and detail of the analysis: did you build a detailed financial model, did you run sensitivity analyses. The other nine described the process: did you explicitly explore and discuss major uncertainties, did you discuss viewpoints that contradicted the senior leader's.
They then controlled for industry, geography and company size and measured how much of the variance in decision outcomes each side explained. The answer: process mattered more than analysis — by a factor of six.
On the 673 decisions where return figures were available, moving a company from the bottom quartile to the top quartile on decision process was worth 6.9 percentage points of ROI. The same move on the quantity and detail of analysis was worth 5.3.
The asymmetry underneath those numbers is the part worth keeping. In the authors' words: one of the things an unbiased decision-making process will do is ferret out poor analysis. The reverse is not true; superb analysis is useless unless the decision process gives it a fair hearing.
That is the argument for this entire tag. Faced with a hard decision, the instinct is to reach for more analysis — another model, another week of data, one more opinion. The evidence says the scarcer input is a process: a defined sequence that surfaces the real options, exposes the assumptions, and gives the inconvenient view a hearing before the decision closes.
A prompt is a good container for that, because a process is exactly the kind of thing that is easy to describe and hard to remember to do at the moment you need it.
What these prompts have in common
This is not a topic pack. The prompts tagged here come from eleven different areas of the site — operations, startups, hiring, product, data, customer success, partnerships, nonprofits, AI engineering, and both personal and business finance. What they share is a shape, and four moves recur across almost all of them.
Name the real options, including doing nothing. Most decisions arrive pre-narrowed to one proposal and its absence. Costing the status quo honestly — what it actually costs to carry on as you are — changes the answer more often than any other single step.
Write the criteria down before you look at the options. In a noise audit at an insurance company reported in Harvard Business Review, Daniel Kahneman, Andrew Rosenfield, Linnea Gandhi and Tom Blaser found that the median difference between two underwriters pricing the same policy was 55%. Same file, same training, same firm. Judgement made in the presence of the options drifts; criteria fixed in advance do not.
Argue against it before you commit. Gary Klein's premortem is built on 1989 research by Deborah Mitchell, Jay Russo and Nancy Pennington showing that prospective hindsight — imagining that an event has already happened — increases the ability to correctly identify reasons for a future outcome by 30%. The mechanism is grammatical, and it is the whole trick: a standard risk review asks what might go wrong; a premortem states that the project has already failed and asks what did.
Say in advance what would change your mind. If no finding would alter the decision, you are not deciding — you are assembling a justification, and the analysis is decoration.
Deciding whether to do the thing at all
The most valuable decisions on this page are the ones that end with no,
because they are the cheapest.
Decide Whether You Should Raise Money at All tests whether venture funding is the right instrument for your business, rather than the default one — see the startup pack for why that question got sharper. Decide Whether to Build a Partner Channel at All does the same for a channel, against the readiness conditions that predict one working; partnerships covers the rest of that programme.
Decide Whether Your Task Actually Needs an Agent is the same question in AI engineering, and it is unusually concrete: it runs the compounding-reliability arithmetic and lays out the cheaper architectures — a single call, a fixed pipeline, an agent with a bounded loop — with what each costs to run and to maintain.
Decide What to Automate and What to Leave Alone scores candidates on volume, stability and failure cost, and Decide Whether to Do It In-House or Buy It In works make-versus-buy through real total cost rather than a gut call on price. Both come from operations.
Build the Case for a Big Purchase or Investment is the fullest expression of the pattern: three-year total cost, payback under a realistic downside, the single assumption that breaks the case, and what doing nothing actually costs. Think Through a Big Purchase is the household version from personal finance.
Deciding between options
Compare Options in a Decision Matrix and Weigh a Decision are the general-purpose pair, and both carry a deliberate constraint: they expose which inputs are estimates. A matrix converts judgement into numbers, and a number looks authoritative regardless of what went into it. Making the weights and the guesses visible turns the output into an argument you can examine rather than an answer you have to accept.
Compare Two Financial Options puts two things on a like-for-like basis — most of these comparisons are hard only because the options are quoted in incompatible units.
Choose Software or a Vendor Without a Six-Month Evaluation is built for the real constraint, which is time: fast enough to be used, structured enough to still catch what goes wrong in year two. Choose the Right Model for a Task applies the same discipline to model selection by defining the hardest case first and then designing the head-to-head test that settles it on your own data.
Qualify a Partner Before You Sign Them scores a prospective partner into sign, limit, pilot or decline — a four-way output rather than a yes/no, which is usually the more honest shape for a judgement made on incomplete information.
Two prompts from customer success decide about people rather than things, and are the harder for it: Decide Which Customers Get a Human and Which Get a Playbook and Decide Which Customers Aren't Worth Keeping. Both work from unit economics, and both are decisions that get made by default — badly, and without anyone owning them — if they are not made deliberately.
Deciding with other people
This is where process earns its factor of six, because the failure mode is social rather than analytical.
Run a Hiring Debrief and Make the Decision is the clearest instance of the noise finding in practice. It collects ratings before discussion begins, so the panel is not quietly converging on whoever spoke first, and it examines disagreement rather than averaging it away — the disagreement is the information. More on that in hiring.
Write a Product Decision Memo With a Recommendation is for the decision that keeps getting reopened, which is nearly always a decision that was never actually written down. See product management.
Split Equity With a Co-Founder Without Wrecking It forces the scenarios that break partnerships later — what happens if someone leaves, what vesting applies — while the conversation is still easy. Recruit a Board That Fills Your Actual Gaps builds a skills matrix first, so the seat is filled against a gap rather than by whoever was available; the nonprofit pack has the rest.
Handle a Raise That Isn't Coming Together belongs here too. It is written for the point at which the options are still real, because the defining feature of that decision is that waiting removes them.
Work out what is actually going on first
A good process applied to the wrong question produces a well-documented mistake.
Find the Bottleneck That's Limiting Your Output locates the single constraint capping throughput, so you stop optimising the parts that were never the problem. Fix a Problem That Keeps Coming Back separates the fix for this instance from the fix for the condition producing instances — run it the third time you see the same issue.
Work Out Your Real Capacity Before You Promise a Deadline counts the interruptions and overheads that never appear in a plan, which is the arithmetic most missed deadlines were missing.
From data analysis: Write an Analysis Plan Before You Pull the Data commits to the method and the decision rule before any data is touched, which is the only reliable defence against an analysis quietly becoming a hunt for a preferred answer. Estimate a Number You Have No Data For decomposes an unknown into knowable parts and identifies which assumption the answer actually depends on. Sanity-Check an Analysis Before You Share It is the last step before anything reaches a decision-maker.
Build a Theory of Change That Survives a Funder's Questions is the same discipline in a different vocabulary — activities to outputs to outcomes, with the assumptions stated and the weak links named.
The research pack covers the enquiry that feeds all of this, and holds the line that matters most: nothing produced with a prompt is verified merely because it is well-organised.
Where this stops
None of these prompts decide anything. They structure a decision and then hand it back, and they work only from the facts and numbers you supply — they will not source a market figure, a competitor's price, or a rate of return, because anything they produced there would be invention. A framework applied to made-up inputs is worse than no framework, because it launders a guess into a table.
They also cannot see what you have not told them: the politics, the relationship, the thing everyone knows and nobody has written down. Where a decision carries legal, financial, employment, medical or safety consequences — an equity split, a redundancy, a lending agreement, a contract you are about to sign — the output is preparation for a conversation with a qualified professional, not a replacement for one.
Sources
- Dan Lovallo and Olivier Sibony, The case for behavioral strategy, McKinsey Quarterly, March 2010 — 1,048 major decisions over five years scored against 17 practices; process mattered more than analysis by a factor of six, and top-quartile process was worth 6.9 percentage points of ROI against 5.3 for top-quartile analysis, on the 673 decisions with ROI data available
- Gary Klein, Performing a Project Premortem, Harvard Business Review, September 2007 — prospective hindsight increases the ability to correctly identify reasons for future outcomes by 30%, per 1989 research by Deborah J. Mitchell, Jay Russo and Nancy Pennington
- Daniel Kahneman, Andrew M. Rosenfield, Linnea Gandhi and Tom Blaser, Noise: How to Overcome the High, Hidden Cost of Inconsistent Decision Making, Harvard Business Review, October 2016 — a noise audit found the median difference between underwriters pricing identical policies was 55%