Concept

Deliberate Understaffing

conceptproductstartupsengineering
Teach me this →

Deliberate Understaffing

Deliberate understaffing is Pavel Durov‘s practice of keeping an engineering organisation smaller than its workload would conventionally justify, on the claim that headcount and product quality are not merely uncorrelated but often inversely related. Telegram runs a roughly 40-person core engineering team — back-end, front-end, design, and system administration combined — managing around 100,000 servers and over a billion users; VK, Durov’s earlier company, was built and run by one person for its first year. The practice is distinct from ordinary startup frugality: it is held deliberately, past the point the company could easily afford to hire more, as a permanent operating discipline rather than an early-stage constraint.

The mechanism

Durov’s stated reasoning has two linked parts. First, a coordination-cost argument: adding people adds communication overhead, and beyond a certain size a team spends most of its time synchronising rather than producing — a version of the classic argument against large teams, but applied here as a permanent ceiling rather than a temporary scaling pain. Second, and more distinctively, a forcing-function argument: refusing to hire past a fixed headcount compels the existing team to automate rather than to add labour, which he says produces systems that are simultaneously more scalable, more efficient, and more reliable than a system maintained by a larger, less-automated team — the automation itself becomes the thing that lets a small team support a billion users, not a substitute for it.

A third element is a claim about morale rather than throughput: an employee without enough genuine work to do does not sit idle, in Durov’s account, but tends to manufacture problems to justify their presence, which actively demotivates capable colleagues working alongside them. Several times he describes removing a single underperforming (‘B-player’) engineer and finding that the team’s output rose, rather than fell, because the drag of working alongside low performance was larger than the raw capacity the departing engineer had contributed.

Hiring as filter, not funnel

The practice is paired with a specific hiring mechanism: public coding competitions rather than conventional recruiting. Candidates who repeatedly perform well across several contests — sometimes over years, from adolescence onward — are hired directly. Durov frames this as doing two things a CV-based process cannot: it tests ability under a real constraint rather than a résumé, and it filters out candidates accustomed to a large company’s diffusion of responsibility, where no one person is clearly accountable for a given piece of work.

Where mainstream views differ

The prevailing view in most fast-growing technology companies treats headcount as the primary lever for shipping more product, and treats ‘we need to hire’ as the default response to a growing workload — the industry norm Durov is explicitly arguing against. His position is closer to a small set of engineering-culture outliers who hold team size flat by policy, but he goes further than most such accounts by tying the constraint directly to automation as a forcing function, and by treating morale loss from a single weak hire as a first-order cost rather than a soft consideration. A fair objection: Telegram’s founder owns 100% of the company and answers to no shareholders, which may make an unconventional, headcount-averse operating model easier to sustain than it would be for a venture-backed founder facing investor pressure to scale headcount alongside revenue.

In the wiki