Everything is green until it blows up: why your team doesn't report red flags in time

You would expect the best teams in a hospital to report fewer medication errors. The data on detected errors seemed to say the opposite.

In the nineties, Amy Edmondson, who earned her doctorate at Harvard in 1996, joined a study of errors in drug administration [2][3]. She compared eight units of the same university hospital [2]. The units whose nurse managers led and coached more, and that had better relationships, recorded more detected errors, not fewer [2].

Were they making more mistakes? Edmondson proposed another reading: a detected error rate depends on two things, the errors that actually happen and the ones people dare to report [2]. In some teams, whoever made a mistake said so and everyone looked for ways to keep it from happening again; in others, they kept it to themselves [1].

A nurse in one of those units put it this way: "You get put on trial! People get blamed for mistakes" [1].

Example scenario (it does not describe a real company): Monday, progress meeting for the new invoicing system. For six weeks in a row the project has come out green. On Thursday, ten days before go-live, the technical lead asks for fifteen minutes: the connection with the bank never passed testing. He had known for a month.

Nobody lied in those six meetings, but nobody told the whole truth either. And a red flag that arrives late tends to arrive when fixing it already costs more.

One clarification: Hábil did not take part in any of the cases or studies mentioned here, and has not worked with the organizations cited. We tell them using what those organizations, the researchers and the Diario Oficial de la Federación (Mexico's official gazette) reported publicly, with the date of each fact.

In 1 minute

• The risk: that your team, or your vendor, knows something is going wrong and does not tell you in time. The red flag reaches you when there is less room left to correct course.
• What inaction costs: go-lives that slip and decisions made on a false green. At a factory, a team that did not dare to question the assignment from management lost time in "blind alleys" [1]. And 75% of companies in Mexico do not see it as highly likely that their people will report a serious breach, according to a survey by the consulting firm KPMG of more than 90 decision-makers in Mexico and Central America [13].
• What has been measured in Mexico: in a survey of 2,500 medical residents at 105 hospitals, teamwork came out as the greatest strength and a non-punitive response to errors as the greatest weakness [15]. Teams that work well together, but where making a mistake still feels punished.
• What you will learn here: what the psychological safety that Edmondson described in 1999 is [1], what friction hierarchy adds, what breaks in your systems, six signals with thresholds and a seven-question test.
• Your decision for Monday: how you are going to respond to the next red flag, and who escalates it, to whom and how quickly.

In this article

1. Why did the best teams report more errors?
2. What is psychological safety, in two lines?
3. What friction does hierarchy add in Mexico?
4. What has been measured in Mexico?
5. What does a hidden red flag look like in a systems project?
6. What breaks in your systems when nobody speaks up?
7. What mistakes do executives make when they try to fix it?
8. How do you measure it at your next meeting?
9. Key takeaways
10. What does Hábil do?
11. Frequently asked questions

Why did the best teams report more errors?

A team that reports more errors may be the one that works best: what you see depends on what your people dare to tell you.

In the hospital study, the detected error rates rose almost in a straight line with how open each group was [3]. That is why Edmondson did not read them as proof that some made more mistakes, but as a mix of errors that occurred and errors that were reported [2].

The finding that is most useful to an executive is a different one: the differences showed up between units of the same hospital. It was not the institution's culture, but that of each group, usually of 4 to 20 people, and it had a lot to do with the direct manager [3].

Edmondson summed it up this way: when errors do not come to light, one person can learn, but the group does not [3].

The table is our reading of the study [2][3].

Keep this idea in mind, because it comes back in the Mexican measurement: a low number of reports is not, by itself, good news.

What is psychological safety, in two lines?

It is the confidence, shared within a team, that you can say "I made a mistake" or "this is going wrong" without paying for saying it.

In two lines: in 1999, Amy Edmondson, now a professor at Harvard Business School, defined it as the shared belief that the team is a safe place to take risks in front of others [1][4]. Where it exists, people ask questions, admit mistakes and ask for help, and that learning is associated with better performance [1].

For those who want the backing:

• Where it comes from. Edmondson studied 51 teams at an office furniture manufacturer with some 5,000 people [1]. 427 of the 496 team members invited responded, and 135 managers from outside the teams rated their work [1].
• What she found. Where there was psychological safety there was more learning behavior, and that learning explained the better performance. Two things that depend on management also mattered: that the leader coached the team and that the company gave it support [1].
• Who else found it. Google studied 180 of its teams in its Project Aristotle, whose launch it places in 2016, and reports that psychological safety was the first, in order of importance, of the traits of its effective teams [6]. That is what Google reports; we found no other open source.
• What blame costs. In Harvard Business Review, in April 2011, Edmondson reported that the executives she asks estimate that only 2 to 5% of the failures in their company deserve blame, but acknowledge that 70 to 90% are treated as if they did [4][5]. The result, according to her: many failures go unreported and their lesson is lost [4][5].
• What it is not. In the same magazine, in May–June 2025, Edmondson and Michaela Kerrissey made clear that it is not being nice to avoid arguments, it does not mean accepting every idea, it does not protect anyone from being fired and it does not get in the way of accountability [7][8].

Translated to your meeting: psychological safety is not a comfortable team; it is a team where bad news travels up fast.

What friction does hierarchy add in Mexico?

Hierarchy alone does not explain your team's silence, but it adds friction: in a culture that expects the boss to give orders, bad news needs more help to travel up.

One well-known measurement of that distance is Geert Hofstede's. On his "power distance" scale, Mexico scores 81 points and the United States, 40 [11][12]. According to The Culture Factor Group, which publishes the model's scores, in Mexico subordinates expect to be told what to do and the ideal boss is a benevolent "patron" [11].

Take it as context, not as a diagnosis. Those scores come from measurements made in the seventies [12]. A national average does not describe your company or prove that people there keep quiet more. Remember the hospital: the differences were between units in the same building, and they depended on the direct manager [3]. The country's culture carries weight; the team's culture is up to you.

Silence is not only a Mexican matter, either. In 2007, in Harvard Business Review, James Detert and Edmondson reported that at a leading high-tech company, self-censorship seemed common from the rank and file up to senior management [9]. Three facts for your committee:

Look at the 25%. One in four people mentioned, among their reasons, that speaking up would make no difference [10]. That silence is not fixed with a suggestion box; it is fixed by what happens after someone speaks up.

About the KPMG survey, from a consulting firm: it was conducted in July 2026 with more than 90 decision-makers (senior management, boards and shareholders) in Mexico and Central America, without saying how many are from Mexico [13]. It measures perceptions of ethics hotlines, not project alerts.

Within those limits, it gives a useful signal. 84% of the companies surveyed in Mexico said they have an ethics hotline, but only 23% have it operated by an independent third party, and 43% of those that have one receive 0 to 10 reports a year [13]. Having the channel does not ensure it gets used.

And the law? The NOM-035-STPS-2018 (Mexico's official standard on psychosocial risk factors at work), in four points [14]:

• When: it was published in the Diario Oficial on October 23, 2018, came into force one year later and some obligations, two years later.
• For whom: every workplace, with different obligations for up to 15, 16 to 50 and more than 50 workers.
• What it requires, among other things: safe and confidential mechanisms for receiving complaints, and actions that promote participation and communication among workers.
• The point that concerns an executive: among those actions, which apply from the smallest workplace up, is promoting that the employer, the supervisor or the immediate manager communicate directly and frequently with workers about any problem that prevents or delays the work (section 8.2) [14].

Complying with it matters, but it is not the same as having psychological safety. The NOM does not prescribe a dashboard or the escalation of red flags to management. That part is yours to design. (Our reading.)

What has been measured in Mexico?

A Mexican measurement found teams that get along well and where making a mistake feels punished: cordiality does not ensure that people tell you about errors.

We did not find a Mexican company that has published, with figures and a sequence, how it measured and improved its reporting culture. The most solid evidence comes from hospitals, the ground where Edmondson started.

The measurement was carried out in 2014 by the Comisión Nacional de Arbitraje Médico (CONAMED, Mexico's national medical arbitration commission) and the graduate studies division of a Mexican public university [15]. CONAMED published it in its journal in 2017 [15]:

• Whom: 2,500 medical residents in 78 specialties, at 105 public and private hospitals [15].
• With what: an online questionnaire of 42 questions in 12 dimensions, in line with the one developed in 2004 by the United States health care quality agency (AHRQ) [15].

Source: [15]. In private hospitals, non-punitive response was the only one of the 12 dimensions below 50% [15].

The survey also asked how many adverse events each resident had reported in the last 12 months. More than a third (34.2%) had not made a single report, and 58% had reported between one and five [15]. In the private sector, those who had reported nothing were fewer: 24.9% against 33.7% in the public sector [15].

What can an executive take from this measurement? Three readings of ours:

• Getting along is not the same as being able to say "I made a mistake." Teamwork came out as the greatest strength in both sectors, and non-punitive response as the greatest weakness [15]. A team can be cordial and still hide its errors.
• A zero is not always good news. That a resident reported nothing in a year does not prove they kept quiet; they may not have seen any event. But a third at zero is a question worth asking [15].
• It can be measured with simple questions. Further on we leave you a seven-statement test.

What does a hidden red flag look like in a systems project?

In a project, the hidden red flag is almost never a lie: it is a green that nobody dared to question.

Edmondson saw it at the same furniture manufacturer, with two teams that were developing new products [1]. One tried many things quickly and often simultaneously. The other perfected one solution after another before seeking feedback, and got stuck in "blind alleys": "Each path represented time wasted," its members said [1].

Why? One of them explained that many just wanted to do the project without questioning it: they were worried about "getting their hands slapped" by management [1]. A senior manager summed it up: "They did learn, but not fast enough" [1].

You may not make furniture, but you have projects with a deadline, a committee and a weekly report.

Example scenario (it does not describe a real company; all figures are invented for illustration): a distributor with 140 people is replacing its invoicing system. The plan is 16 weeks.

The same project, with two different decisions. In week 4 the review is not a presentation but a demonstration with the system running, and the connection with the bank cannot be shown. And there is a written rule: every risk to the date is escalated to the sponsor within 48 hours. The credentials are applied for in week 5.

These are the signals that red flags are not reaching you:

The thresholds are indicative and ours, not the model's. They are for asking questions, not for accusing anyone.

What breaks in your systems when nobody speaks up?

If progress is reported by word of mouth, your company only knows what your people dare to say; the decision worth making before things blow up is that progress comes from evidence.

What breaks, in order of appearance (our reading):

1. Progress becomes an opinion. "We're at 80%" cannot be verified. What can be verified is a part of the system running, with its tests passed.
2. Defects show up late. They are discovered in acceptance testing or with the customer, when correcting them tends to cost more and move the date.
3. Incidents are fixed "off the books." Someone fixes the failure at night and does not log it. That person learned; your company did not [3].
4. The vendor manages its image, not its risk. If a red flag can cost it the contract, it has an incentive to report it as late as possible.

Which decision to make, depending on your size? The NOM-035 already distinguishes three workplace sizes [14]. We use them as a reference:

The columns "What usually breaks" and "The decision" are our reading; the NOM-035 does not talk about projects.

What mistakes do executives make when they try to fix it?

One of the costliest mistakes is to punish, even with a gesture, the first person who brings a red flag: the rest take note.

The column "What it can cause" is our reading, supported where indicated by the sources.

How do you measure it at your next meeting?

Measure it with an anonymous seven-question survey, and turn it into a rule with an owner, a date and an escalation path.

The seven-statement test

Ask each team member to rate, anonymously, from 1 (strongly disagree) to 7 (strongly agree). The statements are adapted and paraphrased from the scale Edmondson published in 1999 [1]; the way of scoring and reading it is ours.

1. Around here, if someone makes a mistake, it ends up being used against them. (reverse-scored)
2. In this team, problems and uncomfortable topics can be brought up.
3. Around here, people who think differently are sometimes pushed aside. (reverse-scored)
4. In this team, you can propose something risky without paying for it.
5. It is hard to ask teammates for help. (reverse-scored)
6. Nobody here would do anything on purpose to get in the way of my work.
7. Around here, what each person knows how to do is put to good use.

How to read it. Reverse statements 1, 3 and 5 (a 7 counts as 1, a 6 as 2, and so on). Take the average by team, not by person. More than the exact number, look for two things: the team with the lowest average and the statement with the lowest score across the whole company. That is where the conversation starts.

Three indicative clarifications of our own: the team's boss should not collect the answers; a team with fewer than five responses should not be reported, so that nobody can guess who answered; and the result should go back to the team in less than a week.

After the survey: the rule

Example of our own; the time frames are indicative.

When the next red flag arrives

Matrix of our own.

Key takeaways

Three questions for your next executive committee:

1. When was the last time someone brought us a red flag before it was obvious, and what happened to them? If nobody remembers a case, or if things went badly for whoever raised the alert, you already have a first clue.
2. Where does the progress we review come from: from what someone says or from something we can see working? If it is the former, your green is an opinion.
3. Can our systems vendor tell us "this is going wrong" without fear of losing the contract? If the answer is "I don't know," ask them directly this week.

What does Hábil do?

We work in a way that lets the red flags in your systems project be seen in time.

Hábil has worked on critical systems since 2006, in Mexico and the United States. In a systems project, a red flag that arrives late costs money. This is how we work so that it is seen in time:

• You see the project in real time, not in a report at the end. Each task, its owner, its status and its date are on a board you have access to, in your tools or in ours.
• You talk to the people who build. A channel shared with your team, where the people who build are present, with the response time we agree on at the start.
• Progress is seen working. Depending on the project, we move forward in stages with a deliverable that you review, or in short cycles where every two or three weeks you see a finished, tested piece of progress.
• A review that does not depend on whoever wrote the code. Quality testing is run by someone other than the person who built it, and then your people validate in your own environment.
• Everything is put in writing. Decisions, agreements and scope changes, with the date and who approved them.
• For management: a single person accountable on our side, who answers for the result, with regular follow-up, clear escalation and risks in plain view.

See the How we work page. If you want to go deeper:

• Waterfall or Scrum: how to choose the way to run your project
• Your software project fails for the same reasons a surprise party does
• Tests that actually catch errors

If your project has been green for weeks and something does not add up for you, let's start there. Request your first conversation, at no cost. It leads to an entry deliverable: a concrete map of your case, with its scope, who takes part on your side, what evidence is reviewed and what decision it lets you make. Write to us on WhatsApp.

Frequently asked questions

What is psychological safety?

The belief, shared by a team, that it is safe to take risks in front of others: asking a question, admitting a mistake or warning of a problem. Amy Edmondson defined and measured it in a 1999 article in Administrative Science Quarterly [1].

Does psychological safety mean nobody is held accountable?

No. Edmondson and Kerrissey made this clear in Harvard Business Review in 2025: it is not being nice to avoid arguments, it does not mean accepting every idea and it does not get in the way of accountability [7][8]. What is worth distinguishing is what gets sanctioned: breaking a rule, not warning of a problem. (Our reading.)

Is it enough to comply with the NOM-035 or to have an ethics hotline?

They help, but they are not enough. The NOM-035 requires confidential complaint mechanisms and promoting direct communication about the problems that delay the work [14]. And in the KPMG survey, 43% of companies in Mexico with an ethics hotline receive 0 to 10 reports a year [13]. A channel without a response teaches people that speaking up is pointless.

Does this apply to my systems vendor?

Yes, as our reading: a vendor that fears losing the contract has an incentive to soften bad news. Ask them how they would warn you of a red flag and when they last did so.

Sources opened and consulted on October 22, 2026. Quotations from and paraphrases of texts in English are given here in their original language or in our own words. There are no direct quotations from the Harvard Business Review articles: everything is paraphrased.