Research a market
What shipped into the genre, at what price, how each one landed, and where demand is still moving. Named games with dates, not a market-size figure.
Marketing agents for game teams
PlayRecon is a marketing platform for the game industry, built around agents.
Publishers, studios and independent developers use it to research markets, plan marketing and launches, benchmark performance, and decide what to build, fund, publish or promote.
- It does the work that is split today between marketing and analytics, as agent workflows rather than another dashboard.
- Questions that take weeks or months of someone’s calendar come back in minutes.
- Every recommendation carries its peer set, its sources and the date each fact was recorded.
The point is not speed for its own sake. It is not spending another hundred thousand dollars on a campaign, a launch window or a product built on an assumption nobody checked.
The agents are in private beta. Scouting the market is free: release calendar, niches, launch windows →
Question
“What did this launch do that the games standing next to it on the same shelf did not?”
What came back
The shelf did not decide it. What separates this launch is a year of demo campaign, a $9.99 price with a launch discount, and tag pages that moved before the charts did.
Evidence, as returned
- observed26 games shared the Popular Coming Soon shelf in the same window; 7 of them never reached the top sellers at allstorefront_visibility (cohort) · 2026-07-26
- observedPriced at $9.99 with a 10% launch discount ($8.99), which also puts it on the Specials shelfprice_snapshots · 2026-07-24
Where this work goes today
A market question rarely has an owner. It has three half-owners, and the answer gets assembled by hand every single time, from scratch, by whoever has the least on that week.
- Split across peopleMarketing holds part of the answer, analytics holds another part, and the rest is in a spreadsheet from March, a dashboard nobody opens, a chat thread and an agency deck.
- Weeks to monthsA question with real money behind it goes into someone’s calendar and comes back a quarter later, after the window it was about has already moved.
- Spent without a comparisonCampaigns, launch windows, price points and whole products get funded off assumptions nobody benchmarked, with no evidence trail to check afterwards.
None of that is a tooling gap. The data exists and the dashboards exist. What is missing is anything that will go and do the work.
What the agents do
Five jobs, and what each one hands back. They run on one record, so the peer set from a concept check is the same peer set the launch plan is measured against.
Compare against real peers
A peer set with a stated reason for every game kept and every game dropped. Disagree with one, take it out, and the case rebuilds without it.
Plan marketing and the launch
A dated plan: beats, windows, price, the shelves and festivals worth chasing, each drawn from games that already ran that play.
Reconstruct a release or campaign
An hour-by-hour timeline of what happened and what moved with it, with the obvious explanation tested against the others that also fit.
Recommend the next move
One action, the evidence behind it, the competing reading still attached, and the finding that would change our mind.
Minutes, not quarters
The work is the same work. What changes is who does it, how much of it runs at once, and whether it stays open afterwards. Nobody sits and steers the run.
You ask in plain language
“Are we too late to a co-op extraction game” is a perfectly good place to start. No query builder, no metric to pick first.
Agents go and look
They choose the comparable games, rebuild the timelines out of the record, and take a run at the obvious explanation to see whether it survives contact.
Dead ends stay in the output
The hypothesis that failed is reported, not quietly dropped. You see what was ruled out, which is most of what makes the rest trustworthy.
You get something you can check
A recommendation carrying its peer set, its timeline, its sources and its dates. Minutes after the question, and it stays open for the next one.
WORKSPACE OBJECTIVEShould we move the launch out of November?
PICK PEERS+ REBUILD TIMELINE+ TEST WHY+ DROP DEAD ENDS
The shape of the work. The output itself is the case card above.
A recommendation you can check
Any tool can produce a recommendation. The question is what happens when the person holding the budget asks why. Here is the same call, made twice.
Recommendation on its own
“Launch at $9.99 and lean on a Next Fest demo.”
No peer set. No timeline. No source and no date. There is nothing in it to check and nothing to argue with, so it gets believed or ignored on feel, and either way the budget moves.
The same call, with the record behind it
“What the record supports: a $9.99 price with a small launch discount, and the demo as the spine of the campaign rather than a Next Fest checkbox. It does not prove either one caused the result.”
- Peer set
- 26 games shared the same pre-release shelf in that window. 7 of them never reached the top sellers at all.storefront_visibility (cohort) · 2026-07-26
- Timeline
- Singleplayer demo Jun 2025, multiplayer demo Sep 2025, updated demo for Next Fest Feb 2026, release Jul 2026.news (Steam events) · 2026-02-23
- Counter-evidence
- Shelf position did not predict the outcome. Three games held that shelf at #1; one of them peaked at 1.8k concurrents. This one sat at #3 and peaked at 37.6k.storefront_visibility + metrics · 2026-07-26
- What we will not say
- How many copies it sold. The review multiplier everyone uses is not calibrated on this cohort, so the number would carry a two-fold spread and teach you nothing.stated in the case, not omitted from it
Observed, estimated and inference are labelled separately, and they never get printed as each other. A price on a day in a region is observed. Why a game climbed is an argument, and it arrives with the competing readings still attached.
What sits underneath
Recommendations are only as current as the record behind them, so the record is the part we run every day.
203kSteam games in the record, re-checked daily
10Mmarket observations a day
100M+recommendation-graph records
41regions priced and ranked
15 yearsof price and discount history
Price history can be pieced together after the fact. Most of the rest cannot: what the page said, which shelf it sat on that morning, where it ranked in Brazil. Storefronts keep almost none of that, so it exists only because it was recorded on the day, which is also why the position compounds.
- SteamLive
- Epic Games StoreComing soon
- PlayStation StoreQ1 2027
- XboxQ1 2027
Who it is for
Three kinds of team, different jobs, one record underneath. Not three products.
One record pointed at a slate instead of a single title. Scouting, screening and monitoring are the same workflow at different widths.
What they use it for
- Scouting: surface the games worth a call before an agent or a demo day gets to them.
- Deal screening: place a pitch against the games that already ran that play, and say which ones it resembles.
- Portfolio monitoring: what moved on the slate this week, and which peer set it moved against.
A question it answers
“Of the games in this genre that shipped a demo this quarter, which three are worth a call, and what would take them off the list?”
What comes back
A shortlist with a reason for every game on it, the cohort it was ranked against, and the fact that would change the ranking.
Bring the question you are actually carrying
The agents are in private beta for now, and we plan to open it up soon. If you want to take part before that, write to us at early@playrecon.com.
Bring a real question, not a demo request: the launch window you cannot settle, the genre you suspect is already crowded, the campaign that underperformed and nobody can say why, the publisher list you have been assembling by hand for a month.
Send it with whatever comparison set you already have in your head. You get back what the record can show, what it cannot, and the first real piece of the answer.
Not ready to send one? Scout the market yourself first, no invite needed: start here →