Imagine joining a business as its new CRO. You have a mandate to improve growth, a board expecting progress and a team wondering what you’re going to change.
You also have experience. You’ve seen what good looks like, delivered results elsewhere and brought some ideas with you. That’s part of why you were hired.
The weekly forecast meeting looks like an obvious place to start. It takes too long. The same opportunities come up repeatedly. An experienced sales manager makes adjustments that seem to rely heavily on personal judgment. You’re reasonably confident that clearer rules and better technology could give everyone some time back.
The team pushes back. The forecast is usually close. Important deals get the attention they need. Why interfere with something that works?
Perhaps you’ve spotted an improvement they can no longer see. Perhaps you’re about to remove something you don’t yet understand. Both deserve consideration, and neither can be settled by who has the stronger opinion.
Before changing the process, I’d want to know what makes its results possible. I’d ask the same question about the approach you intend to introduce.
This is where Dunning–Kruger becomes useful to a GTM discussion. It raises a question about the basis on which we judge capability, including our ability to recognize what is working. Taken seriously, it should make us curious about our own assessment before we use it to explain somebody else’s.
What counts as success?
The team points to results from its current approach. The incoming CRO points to what she achieved elsewhere. Before choosing between them, we need to understand what each is calling success and what produced it.
Take a marketing campaign. It generated a lot of leads. Sales liked the leads. A few became customers. The quarter finished ahead of target. Somewhere along the way, the campaign became a success and a candidate for more investment.
Each observation is useful but they don’t establish the same thing.
Lead volume tells you about response. Bookings tell you that business was won. To understand the return, you also need to examine the cost of acquiring those customers, what it takes to serve them and whether they receive enough value to stay. Revenue attributed to a campaign isn’t necessarily revenue that would have disappeared without it.
A team can be right that a campaign generated demand and wrong about whether repeating it is the best use of the next pound or dollar.
The incoming CRO’s previous success deserves similar scrutiny. What did the approach contribute? How much depended on an established brand, a different buying process, strong product demand or people who knew when to depart from the rules? Which of those conditions exist here?
This isn’t a reason to discount experience. Experience gives you a valuable starting hypothesis. Examining its conditions makes it more useful.
The incumbent team has work to do as well. An accurate forecast is valuable, but it doesn’t tell us whether the process is efficient, whether errors cancel each other out, or whether the business depends on one manager to make sense of unreliable inputs. And repeatedly predicting disappointing results is hardly a complete account of commercial capability.
Before arguing about which approach is better, agree what better needs to mean.
The research is more interesting than the popular version.
In their 1999 studies, Justin Kruger and David Dunning found that low performers on tasks involving humor, grammar and logic substantially overestimated their performance. They proposed that some of the knowledge needed to perform well is also needed to recognize errors. A person may therefore lack both a skill and a reliable basis for evaluating it. Kruger and Dunning, 1999
High performers also misjudged themselves, but an important part of their error concerned their standing relative to others. The authors suggested that they assumed their peers could perform similarly well. In a follow-up phase, seeing other people’s work helped them revise that assessment. This wasn’t evidence that capable people generally lack confidence, or that self-doubt is a mark of expertise. Original study, particularly “The Burden of Expertise”
The explanation remains contested. Critics have shown how measurement noise and the way results are grouped and compared can produce or exaggerate the familiar pattern. Near the bottom of a scoring scale, there is more room to overestimate than underestimate; near the top, the reverse applies. That can produce part of the apparent pattern without establishing a psychological explanation. Grouping people by their test results and comparing their self-assessments doesn’t, by itself, establish why they differ. Gignac and Zajenkowski, 2020, Magnus and Peresetsky, 2022
Other work has supported reduced sensitivity to errors among low performers in particular tasks; another replication found a statistically significant but minimal effect. Neither a universal law nor a simple “debunked” verdict does the evidence justice. Jansen and colleagues, 2021, Dunkel and colleagues, 2023
For GTM, I would take this as a reason to examine how we form our judgments. It doesn’t tell us that a particular CRO is overconfident, that a cautious team is secretly exceptional or that a business has a psychological condition.
The research itself deserves the scrutiny we’re asking leaders to apply to their plans.
Recognizing a weakness changes the conversation. It doesn’t finish the job.
Suppose the CRO and the team agree that qualification is poor. They can examine what’s already being done and whether it addresses the cause. If compensation rewards bookings regardless of customer fit, another qualification workshop may have limited effect. If sellers lack evidence about which customers achieve value, better questions on a call will solve only part of the problem.
Now suppose everyone believes qualification is strong, but reviewing recent deals shows inconsistent buying evidence and repeated exceptions. The leadership task includes understanding why the existing standard felt adequate. Otherwise, a new process may inherit the same assumptions.
A third possibility is that qualification works well for one customer group and badly for another. Different accounts of the business may then be accurate descriptions of different experiences. Averaging them together, or persuading everyone to agree, would lose something useful.
This is why Acuity considers maturity, perception and alignment together. Maturity concerns how developed and consistently operated a capability is, as indicated by structured assessment answers. Perception compares that detailed assessment with the participant’s overall view. The alignment lens examines how closely participants’ assessments of operating practice agree, and where they diverge. That divergence may reflect different experiences, definitions or visibility across the business; it doesn’t necessarily mean people are in conflict.
These are signals to investigate. The answers aren’t an independent audit of the business, and Acuity doesn’t diagnose Dunning–Kruger or assess individual intelligence. A team can agree and still be mistaken. Its overall assessment can match its detailed answers while describing a weak capability. Where evidence is missing, the conclusion needs to remain open.
The combination helps frame a more useful next step: improve a recognized weakness, challenge a favorable assumption, locate uneven practice or investigate a strength that hasn’t received enough attention. Acuity’s guidance connects that investigation to operating evidence, decisions, ownership and a later check on what changed.
Commercial importance still matters. An acknowledged onboarding failure threatening renewals may deserve attention before a smaller, unrecognized problem elsewhere. Discovering a gap in perception doesn’t automatically move it to the top of the agenda.
Some of the value lies in recognizing what you should keep.
A mandate to improve creates pressure to identify things to change. Preserving an existing practice can feel like a less impressive contribution, particularly when its value is hard to explain.
In our CRO example, the forecast meeting might expose customer commitments that would otherwise be accepted at face value. Perhaps the manager knows which implementation dependencies make a promised start date unrealistic, or notices that a deal requires product work nobody has agreed to fund. Those judgments affect delivery plans and spending decisions beyond Sales.
If that is what the meeting achieves, removing it has consequences. If most of the meeting is spent reading information already available elsewhere, there is a good case for redesigning it. Understanding the useful work gives you a better basis for doing so.
There may also be effective practices worth extending. A team that consistently wins customers who adopt successfully and remain profitable to serve may have something valuable to teach the rest of the business. Dismissing its approach as ordinary could leave a worthwhile growth opportunity underfunded. Calling it a competitive advantage would require further evidence about alternatives and competitors, but you don’t need that claim to examine whether it deserves more investment.
I don’t think the objective should be to make leaders less confident. Where the evidence supports conviction, we should be prepared to back it.
That brings us to a question I think every growth business should ask:
Does the capability belong to the business, or does the business currently benefit from a capable person?
The distinction matters because the same result can be produced in very different ways.
Our sales manager may deliver a reliable forecast by maintaining relationships across Product, Finance and Customer Success, checking exceptions personally and correcting the system’s omissions. Another business may capture the relevant evidence as part of ordinary work, give people clear decision rights and need far less intervention.
Both can report a similar number. They have different exposure when that manager is absent, the volume of work increases or new people join.
A skilled operator may also struggle to explain a judgment that has become second nature. “You can tell this deal won’t close” isn’t much help to someone who cannot yet tell. What did the manager notice? Was the buyer avoiding a decision, was funding conditional, or was the proposed implementation date impossible? Which observations can others learn to use, and which still require experience?
There is valuable work in making that judgment more accessible. Some of it can become shared criteria. Some belongs in training or better access to information. Some should remain an explicit escalation to someone with the necessary expertise.
This is where a system view earns its keep. A forecast depends on what Marketing attracts, what Sales qualifies and promises, what Product and delivery can support, and what Finance needs to plan. Incentives influence which problems people surface. Ownership determines who can resolve them. A reliable result at the end doesn’t tell you which of those relationships are sound and which someone is repairing by hand.
Better growth decisions start with seeing the system.
AI makes this a particularly consequential distinction. Return to the CRO’s proposal to reduce the time spent on forecasting. An AI tool may summarize calls, update records and flag patterns much faster than the current process. Those improvements could be worthwhile.
But what information does the manager use that the tool cannot see? Does “implementation ready” mean the same thing to Sales and delivery? Is a customer’s enthusiasm being treated as evidence of approval? Does anyone record the exceptions the manager currently resolves through a conversation?
Automating around those omissions could make the output easier to produce while leaving its basis unreliable. Equally, capturing the missing evidence and making exceptions visible could help more people exercise better judgment. We need to establish which is happening.
A recent study of AI-assisted reasoning offers a useful caution. Participants using AI performed better while continuing to overestimate their performance. Its randomized comparison also found overestimation among participants without AI. The defensible lesson is that better task performance need not bring equally good self-assessment. These were reasoning tasks, not studies of CROs or commercial forecasts. Fernandes and colleagues, published online 2025, journal issue 2026
For the business, the test has to reach beyond whether the tool produces a plausible answer. Does it help identify errors early enough to change a decision? Does time saved survive the checking and rework? Can another person use the process effectively? Does performance hold for the customers and situations that matter?
That investigation should help us move faster. It shouldn’t become an excuse to study the business indefinitely.
Decision velocity depends on knowing what needs to be resolved.
The CRO doesn’t need a perfect model of the company before making a useful change. She does need to be explicit about the decision. “Improve forecasting” is too broad. “Reduce the time spent producing the forecast while preserving its reliability for decisions about spending and delivery capacity” gives the investigation a purpose.
Start with a manageable selection of recent deals, including expected wins, surprises and exceptions. Look at what was known at the time, who changed the judgment and what evidence prompted the change. A final result alone can hide a forecast that became accurate only when it was too late to be useful.
Bring in the people who can explain the relevant dependencies. RevOps can trace records and changes. Sales can explain customer evidence. Finance can identify when a revised forecast would have altered a spending decision. Delivery or Customer Success can test assumptions about readiness and timing. Use actual operating records and shared evidence, rather than trying to identify who gave which diagnostic answer.
Then choose a bounded test. For one team and one forecast horizon, automate preparation and routine updates while retaining review of material exceptions. The CRO owns the decision; RevOps runs the comparison, with Finance and delivery agreeing what the output must support.
Compare the existing and proposed approaches using only the information available at each forecast date. Track time spent, corrections, missed exceptions and accuracy at the horizon that matters. Check whether the apparent saving has simply moved work to another team.
Before starting, set a review date and agree what would justify expansion or stopping. For example, a new process shouldn’t earn wider rollout merely because it saves preparation time if it repeatedly misses commitments that affect delivery or spending. Early evidence may justify extending the test; replacing the established process requires evidence across enough operating cycles to support that decision.
If the new approach meets the agreed standard, expand it. If it fails, establish whether the problem is missing data, unsuitable rules, an unclear handoff or the tool itself. That gives the next decision a better foundation than defending the original proposal.
A small, reversible experiment should be easier to authorize than a change that removes experienced capacity or commits the business to a new operating model. The amount of investigation should reflect what is at stake and how readily the decision can be reversed.
We can act with conviction while remaining willing to be wrong.
For the incoming CRO, that means holding the inherited process and the preferred replacement to a fair test. The board has a role too: if it asks only what has changed, it may miss the value of understanding what deserves to stay.
An uncomfortable truth should have more influence than a comfortable hierarchy. That applies when the evidence challenges the new leader’s plan, and when it challenges the team’s account of a process they’ve spent years building.
The question to take into the next consequential GTM decision is simple:
Before you change or scale part of your GTM system, how well do you understand what is producing its current results?