The most consequential creative decisions will almost always fail an early test. The question is whether your instruments are designed to find what's genuinely there, or what's immediately obvious.

Sidney Lumet spent fifty years directing films that refused to make themselves easy — 12 Angry Men, Dog Day Afternoon, Network, The Verdict — and that refusal required a specific kind of trust: that the difficulty was worth protecting, and that audiences would meet it if given the chance. The threat he identifies most precisely is the test screening: the early preview, conducted before a film is finished, in which audience reactions are gathered and fed back into the editing process. The practice has an obvious logic. You are making something for an audience; find out what the audience thinks. The problem, Lumet argues, is that what an audience thinks in the first encounter with something genuinely new is almost never a reliable guide to what the thing actually is.

The response captured in an early screening is a response to unfamiliarity as much as to content. Something that doesn't yet have a framework — that asks the audience to construct a new way of watching as they watch — will produce discomfort, resistance, and confusion that is indistinguishable, in a feedback form, from the discomfort produced by something that simply doesn't work. The edit that responds to that feedback smooths the unfamiliarity, makes the thing easier to get, brings it closer to what the audience already knows how to receive. It also, in the process, eliminates the thing that made it worth making.

Network, which Lumet directed in 1976 from Paddy Chayefsky's script, is among other things a study in what happens when that smoothing process becomes the entire logic of a cultural industry. The film's portrait of a television network that discovers its audience will watch anything sufficiently stimulating — rage, spectacle, the performance of authenticity — and organises its entire output around that discovery, was received at the time as satire. It has aged into something closer to documentary. The dynamic Chayefsky identified — that optimising for immediate emotional response produces programming that is at once enormously effective and entirely hollow — has become the operating principle not just of television but of most attention-driven media. The test screening impulse, applied at industrial scale and with algorithmic precision, produces exactly the cultural landscape Network was warning about.

Seinfeld sits in a genuinely strange position relative to both. It is commercial network television — NBC, prime time, mass audience — and yet it operated on a logic that Lumet would recognise as his own. The show was greenlit reluctantly, survived early ratings indifference, and was built on a premise that would have failed any instrument designed to capture immediate reaction: a sitcom about nothing, populated by characters who are petty, self-absorbed, and never meaningfully punished for it. No sympathetic arc. No redemptive lesson. No conventional payoff. What it understood was something precise about how social life actually works: the unspoken rules governing ordinary interactions, the gap between what people say and what they mean, the absurdity that lives inside entirely routine transactions.

The instrument you use to evaluate something determines what you find — and an instrument calibrated for immediate reaction will consistently eliminate what takes time to become what it actually is. In research practice, the survey that captures first response, the NPS score gathered at the moment of contact, the focus group reaction to a concept not yet fully formed — these are real data about a real response, and they are measuring the shock of the unfamiliar as much as the value of the thing itself. The tools that reach something more reliable are willing to remain in the presence of ambiguity before moving toward a conclusion.