We stopped counting page views and started counting finished games

Running a portal with 56,000 browser games teaches you a humbling lesson about metrics. For years, the number we watched most closely was page views. It went up, we celebrated. It went down, we panicked. Then one day we looked at a different number, and the page views stopped mattering so much.

The number was finished games. Not starts, not clicks, not time on page. Finished games: a player who opened a game, played it to an ending, a win, a loss, or a completed level, and then chose what to do next. It is a harder number to move, and a much more honest one.

Why page views lie to a games portal

A page view is cheap. A player can open ten games in a minute, bounce off all of them, and generate ten page views that mean nothing. A misleading thumbnail, a slow load, or a game that crashes on start all generate page views. If you optimise for page views, you optimise for curiosity, not satisfaction.

Finished games are different. A player only finishes a game they understood, enjoyed enough to continue, and could actually run on their device. Every finished game is a small vote that the portal did its job: it matched the right player with the right game at the right moment.

When we started tracking finished games, the picture changed. Some of our most-viewed games were rarely finished. They had exciting thumbnails and confusing first levels. Some of our least-viewed games were finished constantly. They were quiet, unglamorous puzzle games that players returned to daily.

What we changed

First, we changed how we order games. Instead of ranking by views alone, we started weighting completion. A game that a thousand players finish beats a game that ten thousand players abandon, and the front page now reflects that. Discovery got better almost immediately, because the games we surfaced were the games people actually liked playing.

Second, we changed how we judge new games. A new game gets a fair trial: a fixed number of plays, regardless of early numbers, and then we look at completion before we decide where it lives. That stops a slow starter from being buried on day one, and it stops a flashy disappointment from holding a top spot it did not earn.

Third, we changed what we fix. When a game has high starts and low finishes, something is wrong in the first minute: a confusing tutorial, a broken control, or a difficulty spike. Those are fixable problems, and fixing them moves the finished-games number in a way that fixing a thumbnail never will.

The uncomfortable part

Counting finished games is uncomfortable because it grows slowly. You cannot spike it with a viral post or a clever title. It grows when the library gets genuinely better, one game at a time, and that takes months, not days.

It also forces honesty about the library itself. If a category has low completion across the board, the category might be the problem, not the games. We retired a few categories that looked popular by views and were hollow by finishes, and the portal got smaller and better at the same time.

What this means for anyone building a content site

Whatever you build, find your finished-games number. It is the metric that cannot be gamed by curiosity alone. For a blog, it might be readers who reach the end. For a tool, it might be tasks completed rather than sign-ups. For a store, it might be repeat purchases rather than first visits.

Page views tell you that people arrived. Finished games tell you that people stayed, played, and left satisfied. Build for the second number, and the first one tends to follow, more slowly, but for real.

If you want to see what that looks like in practice, the library at GameZiko (https://gameziko.com) is free to browse, with no downloads and no sign-ups. Play a game to the end, and you will know exactly what we mean.