Novelty brings players once. A short list of structural reasons they come back the next day.
A daily puzzle has one job that a normal game does not: it must make tomorrow's session feel worth opening. That is a structural problem, not a content problem. Novelty gets someone to play once; the shape of the loop decides whether there is a second time.
The first structural requirement is a fixed cost. A daily puzzle must fit into a predictable slot in the day. Word ladders do this well because four-letter routes are short by nature. A player can open the page, finish, and be gone before the bus arrives. If the format needs a commitment, it becomes a weekend activity and the streak dies.
The second is a difficulty range. The same player needs easy days and hard days, and crucially needs to be able to tell which kind of day it is early. A visible par number does this in one glance. Without it, hard days feel like failures and easy days feel like nothing.
The third is a clean stopping point. A daily puzzle should end. Endless modes are a separate product; the daily loop needs a bottom to the page so the player can leave without guilt. The presence of free practice alongside the daily is the right structure: the habit is finite, the practice is optional.
The fourth, and the most overlooked, is shareability without spoilers. Players talk about the puzzle only if talking is safe. A result that can be posted without revealing the answer turns a private habit into a small social ritual.
Those four properties are not features so much as geometry: cost, range, bottom, share. A puzzle can be themed, beautiful and clever and still fail if any of them is missing. Visit the site on a normal weekday and watch what you do in the first thirty seconds. That behaviour is the loop, and the loop is the product.