The German Bakery That Sold More by Removing Paperwork

At a German bakery chain, employees were required to sign off on whether they had made eye contact and smiled at customers. Another instruction concerned handling Berliner doughnuts without damaging their sugar coating. These were ordinary parts of the job. The operational checklist made them daily acts of certification.¹

There is a recognizable management theory inside that signature. A business promises customers a consistent experience. Headquarters cannot stand behind every counter. Instructions translate the promise into behavior; signatures make the behavior accountable. Somewhere between the promise and the paperwork, however, an assumption enters: requiring employees to document good work will make good work more likely.

The bakery gave researchers an unusual opportunity to test that assumption. What followed was neither a wholesale liberation of employees nor an embarrassing collapse in standards. It was a controlled experiment in removing two documentation requirements, followed by a decision to bring one back.

That last detail is essential to understanding the first.

Two checklists, two different jobs

The company, which the researchers do not identify, was a family-owned regional chain with 145 stores and annual revenue exceeding €100 million. Its shops finished bread and other products supplied by a central plant, prepared sandwiches, served customers, cleaned equipment, and managed displays and deliveries. Quality depended on repeated coordination across people and shifts.¹

Employees had reasons to value the business. In surveys, they praised its products, reliable wage payments, colleagues, and customer relationships. They also complained about extensive checklists. Management was concerned about losing vocationally trained staff and asked whether its approach to control might be contributing.²

The question divided executives. Some saw procedures as indispensable structure for a dispersed business. Others worried about wasted time and the message conveyed by repeatedly asking competent adults to certify routine actions. Both positions were plausible. Neither established what the checklists actually accomplished.

Guido Friebel, Matthias Heinz, Mitchell Hoffman, Tobias Kretschmer, and Nick Zubanov began with diagnosis. They examined the firm's checklist system and interviewed employees and store managers about usefulness and time costs. Two forms emerged as feasible candidates for removal. The exercise deliberately targeted suspected low-value practices, rather than drawing random rules from a hat.¹

The operational checklist required employees to affirm that they had performed prescribed tasks. The daily protocol recorded changing conditions: cash balances, products that had sold out, IT problems, and information for the next shift. Both consumed time. But one largely asked people to confirm familiar behavior; the other could carry information another person did not yet possess.³

In April 2021, the firm removed both forms from 75 randomly selected stores. The other 70 continued using them. Assignment accounted for geography and several existing store characteristics. The experiment ultimately ran for about ten months, through January 2022.¹

The underlying work requirements continued. Employees still had to follow company standards, and mystery shoppers still assessed performance. This was a reduction in documentation and monitoring, not permission to abandon food quality or customer service. The chief operating officer's announcement also expressed trust in employees and encouraged attention to customers and colleagues.¹

That combination matters. The intervention changed paperwork and communicated something about the employment relationship. Its effects cannot be assigned cleanly to either component alone.

What changed when the paperwork disappeared

The researchers estimated that sales increased by approximately 2.7 percent in treatment stores relative to the comparison group. The estimate deserves the qualification often lost when a finding becomes a headline: statistical evidence for the average effect was modest. The main sales result had a p-value of about .07, rather than clearing the conventional .05 threshold. Adjustments for testing multiple outcomes weakened the full-sample evidence further.¹

The result is promising, rather than a precisely settled productivity law. Random assignment makes the comparison much more informative than a company announcing that revenue rose after an internal reform. It does not make sampling uncertainty disappear.

Measured operational safeguards offered some reassurance. The researchers detected no deterioration in shrinkage, which combined product losses and theft, or in mystery-shopping scores. Google review text became more likely to mention favorable service speed and shop appearance. Average star ratings did not significantly improve.¹

A sales increase accompanied by stable measured quality is a useful finding. It is a narrower finding than proof that every aspect of the customer experience improved, or that uncommon failures became impossible.

Employee turnover complicated the story further. Overall attrition did not significantly change. Among vocationally trained employees, including store managers, the estimated monthly reduction was 0.44 percentage points, approximately 35 percent relative to the control rate. Among employees without that training, estimated attrition increased, which the authors summarize as roughly 20 percent. Both subgroup estimates were individually imprecise, but the difference between them was statistically clearer.¹⁴

Training status is important here. The study did not simply divide the workforce into people who had been employed for a long time and recent arrivals. Its categories concerned vocational qualifications, although qualifications and tenure were related.

One interpretation is that experienced, trained employees found the paperwork unnecessary or demeaning, while employees with less formal preparation benefited from its orientation. That explanation fits the pattern. The experiment did not establish why any particular employee left. A higher exit rate could reflect confusion, frustration, changing expectations, or other responses to the intervention. Calling the result evidence that less-trained workers “need control” would reach beyond what was measured.

The distinction is nevertheless consequential. A procedure can be redundant for someone who has internalized a task and useful for someone still organizing it. Removing the same form can therefore change two people's working environments in different directions. An average turnover metric would have concealed that possibility entirely.

The managers who saw it coming

Another distinction had been recorded before the treatment stores were selected. Regional managers, who knew the shops they supervised, were asked where checklist removal would help. Their predictions were unusually informative. In the 76 stores where they expected benefits, the estimated sales effect was about 5 percent. In the remaining stores, the estimate was close to zero. The positive subgroup findings survived the paper's multiple-outcome adjustment.¹

The managers' explanations pointed toward team competence, organization, the absence of operational problems, and employees' likely reactions to reduced monitoring. Their knowledge was richer than a simple measure of store size or average tenure.¹

This offers a disciplined role for managerial intuition. Ask for a prediction before the result is known, preserve the reasoning, then compare it with evidence. Experience becomes a testable source of information. It gains credibility through what it anticipates.

There is still a limit. The managers knew about the experiment and could influence stores. The researchers found no treatment-related change in regional-manager visit frequency or duration, weakening one straightforward alternative explanation. That does not exclude every subtle change in attention. Nor does one successful set of forecasts establish that middle managers can reliably predict every organizational intervention.

The psychological explanation is similarly compelling and incomplete. Survey respondents in treated stores reported greater trust between headquarters and employees and stronger commitment to their own store. Commitment to the wider firm did not improve. Analyses of when employees previously completed forms, and when sales increased, offered little support for a simple account in which recovered minutes alone generated the gains.¹

But trust itself was not independently randomized. The employee survey had a response rate of about 35 percent, and the authors described their mediation analysis as inconclusive. Removing paperwork, expressing confidence, reducing irritation, and changing local interactions could have worked together. The evidence supports trust as a plausible contributor, rather than a proven causal explanation for the sales increase.

When a rule helps, and when it gets in the way

The idea that control can undermine performance has a research history. In laboratory experiments, Armin Falk and Michael Kosfeld found that imposing requirements sometimes reduced people's voluntary effort; participants interpreted some controls as signals of distrust.⁵ Self-determination research distinguishes being motivated by pressure from willingly taking responsibility for a task.⁶ Neither proves what caused the bakery's sales increase.

Research in other workplaces also warns against treating visibility as an unqualified good. In a Chinese phone factory, Ethan Bernstein found that allowing workers some privacy improved line performance.⁷ Amy Edmondson's research linked psychological safety with team learning, although the bakery study did not establish an increase in psychological safety.⁸ Being watched, being trusted, and feeling safe to speak up are different experiences.

But there is equally serious evidence in favor of structure. A randomized management-improvement program in Indian textile plants increased productivity by introducing better practices; a call-center experiment showed that monitoring can deter some forms of opportunistic behavior.⁹¹⁰ Surgical checklists illustrate the stakes of getting this wrong. An early before-and-after study found fewer complications after a safety-checklist program, while a later Ontario evaluation found no significant improvement after mandatory adoption. Those mixed findings are not an invitation to discard safety controls.¹¹¹²

The useful distinction is between a rule that carries information or prevents a failure and one that mainly records compliance. Herbert Simon's concept of bounded rationality reminds us that workers have limited time and attention.¹³ Organizational scholars Paul Adler and Bryan Borys called useful support enabling formalization, as opposed to controls experienced mainly as coercive.¹⁴ The bakery made the contrast tangible: knowing which machine failed helps the next shift; knowing whether yesterday's colleague smiled may not.

The checklist the bakery kept

The company ultimately abolished the operational checklist across its stores and reinstated the daily protocol. Staff and manager feedback supported its coordination value. Because the experiment had removed both together, this final choice was informed judgment, rather than an experimental demonstration of each form's separate contribution.³⁴

It is an appropriately untidy ending. Management learned enough to change the system without learning everything about it.

For another organization, the transferable method begins with the work a rule is supposed to accomplish. Which failure does it prevent? Who uses the information? Does completing it trigger a decision, or merely leave evidence of compliance? What assistance would disappear if the requirement vanished?

Those questions should reach the people completing the process and the people relying on its output. Complaints identify candidates for investigation. They do not settle the verdict. An unpopular control may protect a customer, a colleague, or a future shift from consequences the person completing it rarely sees.

Where safe and permissible, a trial can compare current practice with a carefully specified reduction. Forecasts should be recorded beforehand. Analysts should measure quality, coordination, employee outcomes, and meaningful failures alongside sales or time savings. If several rules change together, they should be explicit about what the experiment cannot disentangle. A narrow pilot also cannot establish the safety of removing protection against a rare catastrophe.

This is decision hygiene applied to management's own instruments. It asks the organization to examine its controls with the same skepticism it applies to employee performance.

It also helps to ask what could be removed, not just what should be added. In eight experiments, Gabrielle Adams and colleagues found that people often overlooked helpful changes involving subtraction.¹⁵ That does not explain why this bakery accumulated its forms. It does suggest a useful question for any process review: what are we requiring people to do that no longer helps the work?

The bakery's evidence supports a demanding standard for rules: judge them by their contribution, including their effects on the people expected to follow them. Preserve information that helps work travel across a shift change. Test whether repeated certification still prevents a meaningful failure. Revisit the answer when staffing, tasks, or risks change.

The signature was visible. The value it added had to be investigated.

References

Footnotes

1. Friebel, G., Heinz, M., Hoffman, M., Kretschmer, T., & Zubanov, N. Is This Really Kneaded? Identifying and Eliminating Potentially Harmful Forms of Workplace Control. Full paper dated December 2025, distributed as RFBerlin Discussion Paper 065/26 in March 2026. Sections 2–6; Tables 2–5; online appendices, especially A8, A20–A23 and B7–B10. Full paper and appendices. Related NBER Working Paper 34122 and NBER PDF. Working paper, not established here as a peer-reviewed publication. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

2. Friebel, G., et al. “Detecting potentially harmful workplace practice.” VoxEU, September 2, 2025. Researchers' account. Its broad statement about higher Google rankings is superseded here by the full paper's more specific review-text findings and null star-rating result. ↩

3. Friebel, G., et al. Is This Really Kneaded? ECONtribute Discussion Paper 304, August 2025 distribution, manuscript dated July 2025. Sections 2 and 6. Paper and appendices. See also the CEPR discussion-paper abstract. ↩ ↩²

4. Friebel, G., et al. “When Checklists Backfire: Too Much Control Can Harm Workers and Sales.” RFBerlin Research Insight 16/26, May 2026. Institutional summary. Relative attrition changes are rounded summaries; Table 2 of the full paper supplies the underlying coefficients and uncertainty. ↩ ↩²

5. Falk, A., & Kosfeld, M. (2006). “The Hidden Costs of Control.” American Economic Review, 96(5), 1611–1630. Article and abstract. ↩

6. Gagné, M., & Deci, E. L. (2005). “Self-determination theory and work motivation.” Journal of Organizational Behavior, 26, 331–362. Full article. ↩

7. Bernstein, E. S. (2012). “The Transparency Paradox: A Role for Privacy in Organizational Learning and Operational Control.” Administrative Science Quarterly, 57(2), 181–216. Article and abstract; Harvard repository. ↩

8. Edmondson, A. (1999). “Psychological Safety and Learning Behavior in Work Teams.” Administrative Science Quarterly, 44(2), 350–383. Harvard repository and abstract. ↩

9. Bloom, N., Eifert, B., Mahajan, A., McKenzie, D., & Roberts, J. (2013). “Does Management Matter? Evidence from India.” Quarterly Journal of Economics, 128(1), 1–51. Published article; research-funder record reproducing the published abstract. The 17 percent figure is from the published version, rather than earlier working-paper estimates. ↩

10. Nagin, D. S., Rebitzer, J. B., Sanders, S., & Taylor, L. J. (2002). “Monitoring, Motivation, and Management: The Determinants of Opportunistic Behavior in a Field Experiment.” American Economic Review, 92(4), 850–873. Article and abstract. ↩

11. Haynes, A. B., et al. (2009). “A Surgical Safety Checklist to Reduce Morbidity and Mortality in a Global Population.” New England Journal of Medicine, 360, 491–499. Original-study abstract; article. ↩

12. Urbach, D. R., Govindarajan, A., Saskin, R., Wilton, A. S., & Baxter, N. N. (2014). “Introduction of Surgical Safety Checklists in Ontario, Canada.” New England Journal of Medicine, 370, 1029–1038. Research institution's study record and abstract. ↩

13. Simon, H. A. (1955). “A Behavioral Model of Rational Choice.” Quarterly Journal of Economics, 69(1), 99–118. Original article. ↩

14. Adler, P. S., & Borys, B. (1996). “Two Types of Bureaucracy: Enabling and Coercive.” Administrative Science Quarterly, 41(1), 61–89. Author-hosted article. ↩

15. Adams, G. S., Converse, B. A., Hales, A. H., & Klotz, L. E. (2021). “People systematically overlook subtractive changes.” Nature, 592, 258–261. Original-study abstract; article. ↩