A seven-layer limit is a budget that can be priced out directly. Assume a uniform span of control s. The tree below the CEO then holds s + s^2 + ... + s^7 people: 21,844 at four reports per manager, about 30,300 at 4.2 [17]. Uber lands just under 30,000 employees after the cut [9], so the depth limit it describes is only reachable if the average manager carries a little over four direct reports [18]. Real org charts are not uniform, so wherever the tree is shallower than seven, the wide branches have to be wider still.
That is why one- and two-person teams are the first thing to go. A manager with a single report adds a rung and absorbs almost none of the span above them. Dissolving nearly half of those teams [5] removes rungs and hands the reports back to a parent that had capacity. The 20% reduction in the manager count [4] is the same lever pulled across the whole tree, and folding core engineering, science and delivery together, along with the three separate operations teams for restaurants, retail and white-label delivery [7], is the same idea applied sideways instead of vertically.
Read that manager figure as a ratio, not a cost. An Uber spokesperson said the number of managers falls by 20% with some moved into individual contributor roles, and the company would not say what share of managers were laid off; non-managers are in the 3,300 too [4]. Payroll savings are therefore smaller than a fifth of the management line, and Khosrowshahi says the savings are being reinvested rather than banked [10], against a stated commitment of more than $10 billion to robotaxi partnerships [11]. Uber had avoided reductions of this size since the pandemic, after more targeted cuts in customer service and human resources earlier in the year [15], and shares erased earlier losses to rise as much as 1.7% premarket [13].
The remote cap is the piece I would argue with. Fewer than about 1% of employees can work remotely going forward [8]. De-layering deletes formal escalation paths and replaces them with informal coordination between people who no longer share a manager, and colocation is the cheapest way to buy that back. It is also the least auditable of the changes in the email, which is worth noting given that the reorganisation itself is described in countable terms [2].
The count of teams with one or two members, and the distribution of employees by depth from the CEO, are the two numbers that decide whether this transfers to another company, not the headline 10%. The org chart is the one architecture diagram most companies keep current, so both are an afternoon's work. If the depth histogram has a long tail past six and small teams are a large share of the total, you have the accretion Khosrowshahi describes, which his email attributes to a top line that nearly tripled in five years or so [14]. If your headcount grew without that revenue curve underneath it, the rungs are not why coordination is expensive, and pulling them out mostly makes ownership harder to locate.
