What The Devil Wears Prada Tells Us: To Counter Middle-Management Failure, the CEO Must Build in a “Structural Backdoor”
In an earlier piece, I analyzed Miranda’s modus operandi in The Devil Wears Prada and identified what truly distinguishes top-tier leadership: high-position Commitment. In this piece, we continue with the same film to discuss a harder organizational question: structure.
Many CEOs, when faced with middle-management failure, instinctively reach for the same fix — they replace the person. They try to find a more reliable middle manager. But Miranda’s response is not just about replacing people. She relies on a different approach altogether: structure plus intervention.
Why Middle-Management Failure Is Inevitable
Middle-management failure is something virtually every organization of scale runs into. It has three structural causes.
First, distorted hiring incentives. The reward system for middle managers runs naturally counter to “hire someone better than yourself.” Hiring a strong performer destabilizes the manager’s own position. So middle managers naturally gravitate toward candidates who are “just good enough, but unlikely to surpass them.”
Second, information monopolization. A middle manager’s position in the org chart is, by design, the chokepoint where information flows up and down. So the manager has a built-in incentive to filter what travels upward (reporting only what reflects well on them) and to filter what travels downward (passing only their own interpreted version).
Third, resource hoarding. Middle managers tend to keep resources within their own span of control, rather than letting them flow to where they would create the most value. The reason is that they evaluate their own performance by “how many people I manage” and “how much budget I control.”
None of these are moral problems. They are structural ones. Replacing the person doesn’t solve them. Any middle manager who stays in that position long enough will develop similar tendencies.
Senior Leaders Have Always Known to Counter with Structure
Mature CEOs know that “hoping middle managers will self-regulate” is not a sustainable plan. The answer has to be institutional design.
I’ve written before, in a piece called Skip-Level Reporting: Management Taboo, Organizational Necessity, about how CEOs use institutionalized skip-level mechanisms to counter middle-layer distortion. Matrix dotted lines, the PMO, talent development programs — all of these are bypass designs that allow senior leaders to reach real signals from the front line, around the middle manager’s monopoly.
The dual-assistant system in The Devil Wears Prada is one expression of this same logic, applied at the closest-in role to the boss.
It splits access to the boss into two dimensions and divides them between two assistants. The first assistant handles the work world: office matters, screening visitors, managing the calendar. The second assistant, in addition to supporting the first assistant, handles the life world: delivering The Book to the boss’s home every night, processing requests that cross between work and personal life.
The two assistants therefore form an asymmetric picture of the boss. The first assistant sees Miranda’s work face. The second assistant, to some degree, sees Miranda’s life face. Neither can assemble a complete picture of Miranda, and neither can monopolize the boss’s decision-making information.
The design intent here is not “the boss is so important she needs two people.” It is to break the monopoly. Even if Emily wanted to be the sole information node, she could not block Andrea from entering Miranda’s home every night, because that is, by design, part of Andrea’s job.
But Even the Best Structure Cannot Fully Eliminate Failure
Organizations are ultimately run by people. No matter how well-designed the structure, it cannot offset the downside of human nature.
The film captures this beautifully. In the opening, Miranda tells Emily explicitly that the previous two second-assistant candidates Emily sent over were inadequate. This time, Miranda insists on personally interviewing the candidate that HR sends.
My middle-school chemistry teacher, a nationally distinguished teacher, once said: one test may not tell you anything, but two is enough. The same principle applies in management.
A first failed recommendation can be a coincidence. Emily may not yet have fully understood Miranda’s hiring preferences (even though Emily herself was promoted up from the second-assistant role). But a second failure is no longer a coincidence. It is a pattern. Either Emily’s judgment of people isn’t strong enough (a capability problem), or she is engaging in threat-averse hiring (a motivational problem).
Whichever it is, the conclusion is the same: Emily can no longer be allowed to recommend candidates, because she is likely caught in middle-management failure. So Miranda switches channels, routing the search through HR directly. That is how Andrea entered Miranda’s field of view.
The CEO Must Retain the Ability to Intervene
This is the part of the film that’s worth real study.
Many CEOs, once their company reaches a certain scale, gradually let go of micro-level intervention. The reasoning sounds entirely sensible: let the middle managers do their jobs, give your reports room to operate. But this is a trap. Once a CEO fully abandons the ability to intervene, middle-management failure becomes much harder to detect. The CEO also loses their own acuity for reading micro-signals, so even if they later want to step in, their judgment will be off. Retaining the ability to intervene is therefore non-negotiable.
The screenwriters use the most economical strokes to show the two layers at which Miranda retains this ability.
The first layer is diagnosis-triggered intervention. Emily recommends two unacceptable candidates, and Miranda switches channels right after the second failure. This kind of intervention is temporary, diagnosis-triggered, and has a clear threshold (two is enough). The CEO doesn’t need to step in for every decision, but the option to step in must remain alive.
The second layer is more subtle, and more elegant: the second assistant delivers The Book to Miranda’s home every night.
Most people don’t realize that this daily, unremarkable task is, in essence, a structural backdoor. Unlike “mandated skip-level reporting,” it carries no adversarial edge. It is just a passage left in the structure that looks completely innocuous, framed as part of the second assistant’s job. Emily has no grounds to oppose it.
But every time Andrea delivers The Book, she enters Miranda’s personal space. Once two people are designed into the same private space, anything is possible. Miranda can observe Andrea’s real behavior, unfiltered by Emily. Andrea can also bypass Emily and pick up Miranda’s actual concerns directly.
This design has a direct equivalent in modern corporations: the one-on-one mentorship between senior executives and high-potential employees inside a talent development program. The stated purpose is dignified — “senior leaders supporting the growth of high-potentials.” The middle manager has no grounds to object, because this isn’t directed against them. It is called “corporate talent strategy.” But every one-on-one conversation puts the high-potential into the senior leader’s field of view. And every such encounter creates a window for the senior leader to access information that hasn’t been pre-filtered by the middle manager.
That is the real essence of a structural backdoor. It is not a public “going around the middle manager.” It is a passage left in the structure that looks unremarkable, and that allows the possibility of contact to exist naturally.
The mark of a mature CEO is here: at the macro level, the CEO can let go, but the intervention channel always stays open. Retaining the ability to intervene is not the same as intervening in everything. Miranda does not manage Andrea day-to-day, but she always has direct access to Andrea, through the line of The Book delivery.
Variants of the Dual-Assistant System
Because this design has proven effective, it has been passed down as a piece of management wisdom and has evolved into structurally similar variants in many serious enterprises. Job rotation pairs, for example: putting two people on the same role in alternating shifts, which both maintains continuity and creates internal competition. Or a higher-level version: co-CEOs.
These structures take different forms, but their internal logic is identical. They are built from structural design to ensure that no middle manager can monopolize information, personnel, or resources in any of the three dimensions.
But This Is Decidedly Not Dual-Headed Management
Middle managers instinctively resent these designs because they reduce the manager’s irreplaceability. Their objections usually sound very professional: high coordination cost, unclear division of labor, prone to internal friction, decision-making is slow.
These reasonable-sounding objections, without exception, are middle managers defending their own irreplaceability.
But the CEO also needs to hold a clear line: the dual-assistant system is not dual-headed management.
As I noted in Skip-Level Reporting, true dual-headed management is when two or more peer middle managers each hold final decision-making authority. That kind of structure destroys the front-line supervisor’s authority and drags the organization into political chaos.
The dual-assistant system is different. Decision-making authority and resource-allocation authority sit, from start to finish, with Miranda alone. Emily and Andrea are two execution nodes. Miranda is the single decision node. The two assistants also carry an informal seniority order: first assistant versus second assistant.
The dual-assistant design is: distributed execution nodes + a single concentrated decision node + asymmetric distribution of contact surface. What separates it from dual-headed management is where final decision-making authority sits — concentrated, versus dispersed. This is where the two are often confused, and getting it wrong turns an elegant design into an organizational mess.
Closing
The real method for CEOs facing middle-management failure has three pieces, and all three must be present.
Structurally, contain the damage that failure can cause.
Diagnostically, use sampling judgment to identify the specific nature of the failure.
Through intervention, make targeted corrections within what the structure allows.
Middle-management failure cannot be eliminated. Any CEO who believes “I can find a middle manager who will never fail” underestimates human nature. Any CEO who believes “if I design the right structure, everything will run itself” overestimates structure.
Next time, we’ll look at the hardest-to-detect form of organizational crisis: pseudo-optimization. When a company starts loudly preaching cost reduction and efficiency, the real intent may have nothing to do with making the company better.