How to promote a daily game: what worked for six launches

This is not a generic marketing checklist. It is what actually happened, in Umami analytics, across six daily browser games built by the same person: DailyMetro, DailyFoot, DailyGeo, Geopolitix, Le bon rang and Devine le prix. Each launch is one data point, not a formula, and on every single one, a large share of sessions arrived with no referrer at all and can't be traced to a channel. Treat the pattern below as what to try, not what will work.

What should a creator set up before launch day?

Two things paid off across these six games. First, a page per mode or city that search can actually find. DailyMetro is a transit-map guessing game with a page per city; on 2026-06-05, a German public-transport newsletter (tagged utm_source=cp-vollversion and cp-kurzstrecke, utm_referrer=newsletter) started sending traffic straight to /berlin, roughly 1,000 sessions over the following weeks. That traffic had somewhere specific to land because the city page already existed; a single homepage would have lost it.

Second, a shareable result grid, read through its own analytics event rather than assumed. DailyMetro's share rate (sessions with a win that also fire a share event) ran 13.7 to 15.3% in its first month, then settled to a stable 11 to 13% band from July on. DailyFoot's share rate is higher, 16.4% all-time and 17.9% over the last 90 days. Devine le prix logs 779 share events but no win event, so its rate can't be computed at all: decide which events a launch needs before day one, because a missing win event means a missing share rate later.

Which channels actually produced a launch-week spike?

The six launches split roughly into three patterns. DailyMetro's first week mixed Reddit (about 250 sessions combined from reddit.com and its frontpage app on day one) with a UK transit blog, diamondgeezer.blogspot.com (172 sessions on day one, still 62 on day two): a niche blog about the exact topic of the game, not a gaming subreddit. DailyFoot's first spike (2026-06-20, 376 sessions) was Facebook (154 sessions) with Instagram a distant second (58); two days later Instagram took over (167, then 151 two days after that), reading as a push that moved platforms mid-week. Le bon rang's peaks ran through X: 194 sessions from t.co on 2026-08-18, and later its largest single-day Reddit pull of any of the six games, 93 sessions combined across reddit.com, its frontpage and old.reddit.com on 2026-08-27.

If a launch post goes to Reddit or Hacker News, read the actual rules first rather than a marketing blog's summary of them: each subreddit publishes its own self-promotion rule in its wiki, and Hacker News keeps its guidelines and the Show HN rules at fixed URLs, both worth reading before the first post, not after a removal.

Two of the six spikes have no identifiable channel at all. DailyGeo's biggest days (2026-08-18: 238 sessions, 2026-08-21: 186, 2026-08-04: 178) were 85 to 99% sessions with no referrer; Geopolitix's peaks (2026-08-27: 139, 2026-08-26: 129, 2026-08-04: 117) were about 90% the same. Something drove those days, in-app sharing, a screenshot passed around, a closed group chat, but Umami has no referrer header to show it.

Devine le prix is the one clear organic-search story: flat at 24 to 108 sessions a day for its first ten days, then an 11x jump from 101 sessions on 2026-09-16 to a peak of 2,839 on 2026-09-19. Google.com made up 35 to 45% of that peak day (1,429 sessions), growing in step with the total rather than leading it, which reads as an SEO breakout rather than one post going viral. Roughly half that peak day was still unattributed.

How do the six launches compare side by side?

The last column is each game's all-time share of sessions with no referrer domain at all, not just its peak day; it ranges from 37.6% to 72.9%, and every game in the directory sees the same pattern. Read any "main source" cell as the largest channel this data can name, not the whole explanation for the spike.

What should happen after launch week?

Three things, in an order that matches when they start paying off. First, directories and hubs. The only listing page that shows up as a steady referrer in this data is mydailygames.live, my own page that links to all of these games: 166 sessions across five of them in two to four months. That is a small but steady trickle from a page I control, which is the cheapest version of this step. My games are not yet listed on listdle, dlelist, alldle, dailydle or dlehunt, so this data says nothing about what those send. The one outside directory that does list five of them, jeux-du-jour.fr, links without a referrer or a tag, so whatever it sends is invisible here: unmeasurable, not zero. That is the practical lesson: before you submit anywhere, make sure you will be able to see the result, by reading utm_source in your analytics. Where to submit a daily game covers the submission process itself.

Second, cross-promotion between daily games. relaydle's rail, the cross-promotion network these six games are part of, put 974,701 cards in front of players in its first 17 days and produced 5,502 outbound clicks, a 0.56% click-through rate. Where a card sits on the win screen matters: the first slot converts at 0.92%, the sixth at 0.29%, roughly a three-times drop across one page. How the rail works explains the mechanics and the ranking.

Third, AI assistants: a real but small channel so far. DailyGeo logged 52 sessions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com or claude.ai referrers out of 11,191 total; Geopolitix logged 81 out of 4,495. Worth watching, not worth building a launch plan around yet.

None of this replaces the honest caveat that opened this piece: on every one of these six launches, 30 to 73% of sessions carried no referrer at all, and the biggest single-day spikes were sometimes the least attributable of all. Track what you can, list where the data lets you list, and expect the largest piece of any spike to stay a mystery.