Ravensight

MarketingStudio memory

Which post brought players, and what did they ask?

A small studio posts in five places and remembers almost none of the replies a month later. Keeping the posts, the conversation and the player numbers in one record turns that noise into something you can plan the next post from.

A raven standing on an open ledger that glows with charts on a desk at night, five social feed screens around it with amber threads of replies flowing into the book, a Texas flag and a lit football field in the window behind.

Here is a week most indie developers will recognise. On Monday you post a short clip of the new boss on Bluesky. On Wednesday a devlog goes up on itch.io. On Thursday someone on Reddit asks whether the game will have a controller remap, and forty people agree. On Friday a YouTube video about the demo gets three comments asking when the Steam page goes live.

A month later, all you remember is that "the clip did well". You do not remember which of those four things was followed by new players, and you certainly do not remember the forty people who wanted controller remapping.

That is the gap this post is about. Not reach, not follower counts, but two questions a small studio actually plans from: which post brought players, and what did the people who saw it ask for?

Write the post down where the numbers are

The first question needs a record. Analytics can show you that Tuesday tripled. Only you know that a clip went out on Monday. The fix is boring and it works: every time you post, the post becomes a dated record next to your game's player numbers.

In Ravensight that record is a marketing touchpoint, and every post you keep in Pulse is one automatically. Once it exists, the platform measures what followed: it counts the devices whose very first event for your game arrived in the 72 hours after the post, and compares that with the daily rate of the seven days before, times three. You get both numbers and their ratio.

Two honest caveats come with that number, and they matter more than the number itself.

  • It is a correlation with a stated method, not attribution. If the

clip and the devlog both land inside the same 72 hours, each of them shows the whole lift. The number tells you which week to look at; you still decide what caused it.

  • No baseline means no ratio. If your game had no new players at all the

week before, there is nothing to be a multiple of, and you get an empty ratio instead of an invented one.

The value is not one magic figure. It is that after a few months you can look down a list of everything you posted and see which kinds of post were followed by players and which were followed by silence. That is a pattern you can plan with.

Keep the conversation, not just the counts

The second question is the one that usually gets lost. Comments under a post are the most direct feedback a small studio gets from people who have not bought the game yet. They ask whether it runs on a handheld, whether there is co-op, whether the price will drop. They object to the art style. They request a feature you had not thought of. And all of it lives on five different sites, under posts you will never scroll back to.

Pulse keeps that conversation next to the post:

  • The post itself, with its text, its link and when it went out.
  • Snapshots of its numbers over time, so you can see whether it took off in

the first hour or slowly over a week. A new snapshot is only stored when a number moved or a day has passed.

  • The comments, replies and quotes, with who wrote them and when, up to 500

per post, and up to 1,000 reactions. Past that the post's own counters still say how many there were.

  • The people behind them. Someone who comments on three of your posts is

worth knowing about, and the people who came back to the most posts are listed first.

Because every post is also a touchpoint, the measured lift and the conversation sit side by side. A post that brought players and drew ten questions about controller support tells you something different from a post that brought players and drew nothing but emoji.

Four ways in, and what each platform will not tell you

Posts get into Pulse four ways, because no single route reaches everywhere.

Sources. Add a Bluesky handle, a YouTube channel or a Reddit account, and Pulse finds your new posts once a day and reads them on a schedule: hourly for the first two days, every six hours to a week, daily to a month, weekly to three months, then it stops. Pulse holds no account of yours on any of them and asks for no sign-in; a source is a public handle you type in.

Watched links. Save the link to any public page, a Steam community announcement, an itch.io devlog, a forum thread, a press article, and Pulse reads that page daily for a month and weekly to three months. It reads only what is public, honours each site's robots rules, and stops after three unchanged reads in a row.

By hand or from a spreadsheet. For everything else, record the post and its numbers yourself, or import a CSV export through a column mapping you confirm before anything is stored.

Refresh. Any polled or watched post can be read again on demand.

What you get back depends on the platform, and it is worth knowing the limits before you read too much into a missing number:

WhereCommentsWho reacted
BlueskyYes, plus quote postsYes, likers and reposters by handle
YouTubeYesNever: YouTube does not say who liked
RedditYesNever: Reddit does not say who voted
Watched linkWhere the page shows themNo

LinkedIn and X show most posts only to signed-in readers. Pulse does not sign in anywhere and does not read past a login page, so a saved link to one of those usually comes back as "behind a login, capture it by hand". That is deliberate. A login page is not your post, and storing one as engagement would be worse than storing nothing.

And everywhere: a number the platform does not report is shown as not reported, never as zero. Zero likes and "this platform does not say" are different facts.

Let Huginn read the thread for you

Reading forty comments is fine once. Reading them under every post, every week, is the part nobody keeps up. So you can ask Huginn, the AI analyst, to summarise the comments on one post, or every post in a date range (a Summarize button on the post, or Summarize this period on the Posts tab, each showing its price before you click): a short paragraph with the themes, the questions people asked, the features they requested, their objections and an overall sentiment.

Two details make that safe to rely on. What Huginn reads is bounded and anonymous: up to 200 comments, newest and most-replied first, each labelled by a number rather than a handle, and read as material to summarise rather than instructions to follow. And it is cheap to ask twice: a summary is one shared AI unit, and asking again about a post with no new comments returns the stored summary for free. Each summary is also filed as a dated finding in Muninn, studio memory, so the analyst can cite what people asked for the next time you plan an update.

Recognising the same person, carefully

The person who asked about controller support on Reddit might be the same person who replied to your clip on Bluesky. Knowing that is useful: it is the difference between one loud request and two separate players wanting the same thing.

It is also exactly the kind of guess that goes wrong quietly. So Pulse only proposes. When the same handle turns up on two platforms, or two profiles link to the same page, it suggests they are one person, and nothing is merged until someone in your studio confirms it. A rejected pair is never suggested again, and a confirmed merge is permanent, so it is worth the few seconds to look.

Those people are real, and what they wrote to you is theirs. Pulse keeps what was public on your own posts, a handle, a display name, a profile link and the text, and shows it only to your studio. None of it goes into the studio graph, which holds only how often repeat engagers came back to which posts. If someone asks to be forgotten, open them in People and choose Delete person: their comments and reactions are removed from Pulse, and it cannot be undone.

What it costs

There is no separate price. Everything Pulse stores is weighted events on the same monthly meter your player events use, charged once per row actually stored, so reading a post again that has not changed costs only the read. Watched links are charged by the bytes read from the page. The weights are on the rates table.

Background reading stays inside your included monthly events unless the studio owner opts in to let it use the prepaid balance, up to a background budget. Without that, a source that runs out pauses, says so, and picks up again after a top-up or a new month. Anything you ask for yourself, like recording a post or refreshing one, is charged like any other request.

Store numbers, in the same place

The obvious next question after "which post brought players" is "which post sold copies". The sales ledger already holds what each title earned, per storefront and per currency. If you sell on Steam, a connection now brings that in for you: daily sales into the ledger in Steam's own dollar figures, and wishlist adds, deletes and purchases into Pulse as a daily snapshot on your store page. A synced figure never overwrites a row you entered or imported yourself, and the key is stored encrypted and never shown again. Ad revenue from AdMob and similar comes in by CSV for now.

A routine that fits a small team

None of this needs a marketing department. A routine that works for one or two people:

  1. Add the accounts you post from as sources, and save a link for anything

posted somewhere a source cannot reach.

  1. Once a week, open the newest posts and read the comments in one place

instead of five.

  1. Once a month, look down the list of touchpoints and their lift, and note

which kinds of post were followed by new players.

  1. When a question keeps coming up, write it down as a design note. That is

your players telling you what the next update should be.

The point is not to post more. It is to remember what happened the last time, so the next post is a decision rather than a guess. The Pulse documentation has the details, including exactly what each platform exposes and how the crawler behaves.

Questions people ask

How do I know which social post brought new players to my game?
Write each post down as a dated record, then compare the new players who first appeared in the 72 hours after it against the daily rate of the week before. In Ravensight every post you record in Pulse is a marketing touchpoint, so it gets that measurement automatically. It is a correlation with a stated method, not proof, and two posts in the same 72 hours each show the whole lift.
Can Ravensight see who liked my YouTube video or upvoted my Reddit post?
No. YouTube gives comments and counts but never the names of the people who liked a video, and Reddit gives the comment tree and the score but never who voted. Bluesky does name likers and reposters, so a Bluesky post can show who reacted. Where a platform does not report something, Ravensight shows it as not reported rather than zero.
Does Ravensight read LinkedIn or X posts?
Usually not. Both show most posts only to signed-in readers, and Pulse does not sign in anywhere or read past a login page. When a saved link comes back as behind a login, Pulse says so and you record that post, its numbers and its comments by hand or from a CSV.