Some shows only get good after most of the audience has already gone

The theory of late-season improvement is well covered elsewhere. This is about the version you meet in practice.

Improvement arrives after the verdict

Series regularly take a season or more to work out what they are, which is normal given how much is committed before anyone has seen an episode. The difficulty is that the audience forms its view during exactly that period and mostly never returns to check whether anything changed.

By the time the show is doing its best work, the group still watching is far smaller and much less likely to be arguing in public. That produces a mismatch where the strongest material reaches the fewest people and generates the least conversation about itself. The show's public standing stays fixed at the level set by a version it has already moved well past.

Why the news does not travel

There is no efficient channel for telling a departed audience that a programme changed, and marketing is poorly suited to carrying the message. Promotion for a later season is built to reach people who are already watching, since those are the people it can find most cheaply. Word of mouth struggles too, because recommending a show that requires patience is a much weaker pitch than recommending one that starts strongly.

On the call sheet, anyone making the recommendation has to attach a condition to it, and conditions reduce the number of people who act on advice. The recommendation therefore circulates within enthusiast circles and only rarely manages to escape them.

The skip-ahead problem

The usual solution is advice to start at a later point, which solves the quality problem and immediately creates two new ones. It asks a new viewer to trust a stranger's judgement about which hours are disposable, which many people quite reasonably decline to do.

It also strips out the setup that later episodes assume, so the improved material lands with less force than it was built to have. Some shows survive this and others do not, and the difference depends on how much the later work leans on early groundwork. Either way the advice functions as a warning label, and warning labels reliably suppress the number of people who try something.

What actually changes in the show

Late improvement usually comes from a production discovering which of its elements audiences and cast respond to and reweighting the series accordingly. Characters designed as minor get more room, formats that were not working quietly disappear, and the original premise loosens its grip. None of that could have been known in advance, because it depends on how performers and combinations actually behave once they are on screen.

This is why first seasons so often read as a set of experiments and later seasons read as the results of those experiments.

The show is not becoming a different programme so much as becoming the specific version of itself that works.

Streaming changed the arithmetic slightly

When a full run is available at once, a viewer can be told to persist and can verify the payoff without waiting an entire year for it. That shortens the cost of patience considerably and makes late-blooming shows far more recoverable than they used to be.

In the mix, it also means the whole series is judged as a single object, which helps any show whose average is dragged down by its opening. Against this, the abundance of alternatives means a viewer who is bored has an unusually low-friction exit permanently available. The net effect is that late improvement is more discoverable and early weakness is less survivable at the same time.

Production stories are told by people with an interest in how they sound.

Reading recommendations better

When someone says a show gets good later, the useful follow-up is what specifically changes rather than which episode to start on. An answer about a character getting more room or a format being dropped tells you whether the change is one you would actually value.

It is also worth asking whether the recommender watched weekly or in one stretch, since patience feels very different in each mode. Shows with this shape are among the most rewarding things on television and among the hardest to persuade anybody to try. The gap between those two facts is where a considerable amount of good work quietly sits unwatched.

The takeaway

A show's reputation is fixed by the season most people watched, which is almost never its best.

Popularity and craft are two different measurements, and both of them are real.