How Teams Learn Faster: Reflection, Feedback, and Reinforcement

In 2014, a group of researchers led by Giada Di Stefano ran an experiment at a call centre in Bangalore. New hires went through the standard technical training. One group then spent the last fifteen minutes of each day writing down what they had learned. That was the entire intervention. No extra instruction, no coaching, no additional practice, and fifteen minutes taken out of the working day rather than added to it.

After ten days, the group that reflected performed measurably better on the assessment than the group that kept working, by a margin the researchers put at roughly 23 percent. Their paper called it “learning by thinking,” and the uncomfortable implication for anyone running a team is that the fifteen minutes of production they gave up bought more capability than the fifteen minutes of production they would have kept.

What the Research Shows

Experience Is Not the Teacher. Reflection on Experience Is

The Bangalore result is a specific instance of something learning science has held for a long time: doing the work and learning from the work are separate activities, and one does not automatically produce the other. A team can run the same broken release process eleven times and get better at nothing except tolerating it. We wrote about the structural version of this failure in double-loop learning, where teams correct errors inside the existing rules without ever asking whether the rules are the problem.

The US Army arrived at the same conclusion by a different route with the After Action Review, built on four blunt questions: what was supposed to happen, what actually happened, why was there a difference, and what will we do differently. It is deliberately unglamorous. It is also the closest thing to a portable reflection protocol that exists, and it works because it is short and scheduled rather than left to whoever feels like thinking about it.

Forgetting Is the Default, and It Is Fast

Hermann Ebbinghaus mapped the forgetting curve in the 1880s using nonsense syllables and himself as the only subject, which sounds like a methodological joke until you notice the shape has survived 140 years of replication. Memory for new material drops sharply within days unless something interrupts the decay.

The interruption that works is spacing. The testing effect, documented by Henry Roediger and Jeffrey Karpicke, sharpens the point further: retrieving information from memory produces far more durable learning than re-reading it. Being asked beats being told. That finding has an obvious operational consequence which almost no corporate training programme acts on, because a single workshop is easier to schedule than six spaced touchpoints.

Feedback Only Teaches When It Is Safe to Be Wrong

None of the above functions in a team where being wrong is expensive. Reflection requires people to say what did not work, and retrieval practice requires them to attempt an answer they might get wrong, in front of colleagues. Amy Edmondson’s research, which we covered in psychological safety as infrastructure, found that better teams report more errors rather than fewer. They surface more of what happened, which is precisely the raw material learning runs on.

A team can run the same broken release process eleven times and get better at nothing except tolerating it. Doing the work and learning from the work are separate activities.

Why This Matters for Teams

Most organisations already spend real money here. LinkedIn’s learning data, which we covered in the AI skills surge piece, shows demand for capability outstripping what traditional programmes deliver. And McKinsey found 87 percent of companies reporting a current or expected skills gap. The money is being spent. The gap is not closing.

The reason is a design problem, not a budget problem, and we made that argument at length in the training ROI piece. A course is an event. Capability is a habit. Events do not become habits without a mechanism that carries them across the gap, and for most teams no such mechanism exists. People attend the workshop, return to a full inbox, and the material decays on exactly the curve Ebbinghaus drew.

The cost of getting this wrong compounds quietly. A team that does not learn from its own delivery repeats its own mistakes at full price, every time, and the repetition is invisible because each instance looks like a fresh problem. Nobody logs the fourth occurrence of a handoff failure as a pattern. They log it as a bad week.

The Gap the Data Reveals

Learning science is unusually settled. Spacing works, retrieval works, reflection works. These are not contested findings at the edge of a discipline. They are among the most replicated results in psychology.

The gap is that essentially none of it survives contact with a corporate calendar. Spacing requires touching the material six times over six weeks, and the training budget is structured around a two-day workshop. Retrieval requires being asked, which requires somebody to do the asking, and no role owns that. Reflection requires protected time, which is the first thing cut when delivery is behind, which is always.

The second gap is that learning gets treated as an individual activity when most of what a team needs to learn is collective. One person understanding why the release broke is not the same as the team understanding it. Individual learning transfers to the group only if there is a moment where it gets said out loud to the group, and that moment has to be on a calendar or it does not happen.

What Actually Installs

Fifteen Minutes, Written, at the End

The Bangalore intervention is the cheapest thing on this list and the one most likely to be dismissed as soft. Written matters. Thinking about the day is not the same as articulating it, because writing forces the vague sense that something went badly into a specific claim about what and why.

For a team, put it at the end of the week rather than the day, and cap it at three prompts: what worked, what did not, what we will do differently. Long enough to be honest, short enough that nobody starts skipping it in week three.

Space the Reinforcement, Do Not Repeat the Course

If your team has been trained on something, schedule three short touchpoints across the following six weeks rather than a refresher session in six months. Ten minutes in an existing meeting is sufficient. The purpose is not to re-teach the content. It is to force retrieval, which is the mechanism that makes it stick.

Concretely: ask someone to explain the thing to the group without notes. That single move converts a passive review into the testing effect, and it costs nothing.

Run the After Action Review on Wins Too

Teams that only review failures learn that the review is a blame ritual, and they start managing what goes into it. Running the same four questions after something went well breaks that association and surfaces the thing nobody writes down: which of our practices actually caused this, as opposed to which ones we performed while getting lucky.

Make One Person Responsible for the Asking

Retrieval does not happen spontaneously. Somebody has to ask the question, and if it is nobody’s job it is nobody’s job. Rotate it. The person running the weekly rhythm owns the ten minute reinforcement slot for that month, and the rotation matters because it distributes the practice rather than concentrating it in the manager, which is its own trap.

Write the Lesson Where the Work Is

A lesson recorded in a retrospective document that nobody opens has not been captured. It has been filed. Put it where the work happens: in the runbook, the checklist, the template, the definition of done. The test is whether someone doing the task next month encounters the lesson without having gone looking for it.

Di Stefano’s fifteen minutes still strikes most managers as an indulgence, which is the whole finding. The intuition that time spent not producing is time wasted is exactly what the experiment was built to test, and the intuition lost. Teams do not get better by doing more of the work. They get better by being made, on a schedule somebody owns, to say what the work taught them. Our Modern Manager Playbook covers the weekly mechanics in more detail, but the first version of this fits in a recurring calendar invite and three questions.

Related Reading

• Double-Loop Learning: Why Smart Teams Still Make the Same Mistakes
• The Training ROI Problem Is Not About Budget. It Is About Design
• LinkedIn’s Workplace Learning Report: The AI Skills Surge and What It Signals
• Psychological Safety Is Infrastructure, Not Culture
• Deloitte’s Human Capital Trends: The Shift From Jobs to Capabilities