What Keeps Players Coming Back? Stephen Jarrett, Second Dinner
Second Dinner's Stephen Jarrett says core gameplay gets players in the door, but meta loops, battle pass timing and bots decide who is still playing next year.
Core gameplay gets players in the door. What decides whether they are still playing a year from now is everything built around it: the economy, the seasons and the goals that make coming back feel worth it. That is the argument Stephen Jarrett, senior director of strategy and product at Second Dinner, the studio behind Marvel Snap, laid out on the Player Driven podcast. He has spent about 16 to 17 years making games across Playfish and EA, Blizzard, Epic and Second Dinner. Every figure in this piece is Jarrett's own account and has not been independently verified.
Decide with a triangle, not a dashboard
Jarrett says product management began as system design, building a game economy that was motivating and could also sustain a live service, and has drifted toward data optimization. His view is that a data-only approach leads to games that are "relatively soulless, but maybe optimized short term LTV." He allows that in a crowded red ocean, especially on mobile where the user acquisition funnel is life or death, a team may not have a choice.
For the traditional gamer audience he prefers to make games for, he uses what he calls triangle decision making. It combines three inputs: data, player sentiment and feedback, and the game team's own intuition built from tens of thousands of hours of play. Sentiment means more than the loudest voices on Reddit or X. He points to Discord and to surveys that reach the wider player base, not just the vocal slice.

On the data side, Jarrett watches two numbers. The first is committed players, meaning how many people played the game every day this week. The second is the average number of games each player plays, which he treats as a leading indicator. If it starts to fall, he says, the team is probably doing something that upsets players.
Time the battle pass to land when the next one starts
Jarrett argues the battle pass is an engagement tool first and a monetization tool second. It gives players clear goals to stay in the game for weeks or months. The design rule he stresses:
“You need to design the battle pass such that players finish it around the same time that the next one starts.”
Finish too early and players run out of goals and quit, and winning them back is harder now than in the early Diablo III days. Take too long or make it too hard and they feel they did not get their money's worth and may refuse to buy the next one. Jarrett says the Fortnite team spent a great deal of time dialing that balance, including end-of-season events that let players earn extra XP to catch up.
He says Marvel Snap's pass is built on the same Fortnite framework, with a catch-up path so someone who joins in the third or fourth week of the month can still finish, and a time gate on finishing early so players have a reason to return week to week.
Seasons can also be used to create healthy gaps. Jarrett says Diablo III was structured on purpose so players would come in for a season, have fun for a couple of weeks, leave, and return for the next one, producing peaks and valleys. He credits the strength of Diablo's core loop for making that rhythm possible, and notes it is hard for most developers to replicate.
Use bots as an on-ramp and a test bench
Jarrett recalls a problem his team hit at Fortnite, where he says he was head of product for the first two chapters. New players were quitting quickly. A friend would invite them in, and someone would blast them in the back with a shotgun. The team added bots to lower-skill matches so newcomers could get comfortable with the map and the flow. Jarrett calls that an audacious decision for a competitive shooter. Higher-skill lobbies do not get bots.
The same idea reached Marvel Snap in two ways. First, the team used bots to teach new players in the first hundred days. Second, bots gave designers a way to play alone. Before that, he says, testing a meta progression system meant inviting hundreds of people to a giant playtest for a month. With bots, he says the team went from three iterations over four years before he joined to ten iterations in six months. "If you're a game developer, iterations are everything," he says in the conversation, calling it the only way to really make a fun game. One result was the collection track, which he describes as a free battle pass track.

Build meta loops in layers, and do not shortcut them
Jarrett's closing argument is the one he returns to most. In his experience the game economy and the meta loops around it are underappreciated, and many teams do not have someone who understands how those systems work well enough to build a loop players want to repeat for a week, a month or years.
“Core gameplay is extremely important. That's what gets people in the door... But the thing that gets people to wanna play a year from now or two years from now is those meta loops.”
The loops can be social, such as joining a guild and accomplishing something with other people. They can be gear, such as a piece that makes a build feel strong. They can be new characters, such as a hero that just came out. His advice to game developers is not to shortcut meta progression, and to think through how players will feel at each moment and whether they will want to come back in a week or a month.
To plan that, he recommends a player journey exercise that zooms out in three steps.
- Moment to moment: what the player does in the core loop. In Diablo he describes it as explore, kill demons, loop.
- Day to day: why the player returns tomorrow and the day after, such as loot, levels and a stronger character.
- Week to month: longer goals that tend to be social or seasonal, such as finishing a hard raid in World of Warcraft over a couple of months.

He says the sign a loop is working is when a player is thinking about the game while they are not in it. Doing this exercise before launch, he adds, will guide feature decisions and iterations early.
Motivation, and where AI fits
Jarrett frames player motivation as intrinsic, such as wanting a higher rank because the loop is fun, or extrinsic, such as grinding for a legendary item. He recommends avoiding negative motivators, naming shame and loss aversion, and leaning on joy, achievement, goals and social motives like altruism.
On tools, he says he still designs game economies by hand in Google Sheets or Excel because AI is not yet good enough at tuning one to feel right. He builds his own analysis tools with AI and uses Cursor with Claude, and Codex, for game development. He describes the aim as giving designers, product managers and artists a way to implement their own ideas, with engineers keeping the architecture safe and the human element intact.
Four things to check in your own game
- Map your player journey at all three zoom levels and look for the layer with the thinnest goals.
- Compare when players typically finish your battle pass with when the next season starts.
- Ask whether your team can test progression alone, or only through large playtests.
- Track committed players and average games per player, and treat a falling second number as an early warning.
Chapters
- 00:00 Welcome and what keeps players coming back
- 00:20 Stephen's path: Playfish, EA, Blizzard, Epic and Second Dinner
- 00:52 Facebook games and FIFA Superstars
- 03:49 Fortnite: why new players were quitting
- 08:41 Madden Superstars and the Words With Friends idea
- 11:16 Where product management helps most
- 13:40 The decision triangle: data, sentiment and intuition
- 14:27 Two KPIs: committed players and games per day
- 18:00 Producers, product and design: roles differ by company
- 19:18 Exhaustion and forever games
- 20:10 Diablo seasons: peaks and valleys
- 21:21 The battle pass as an engagement tool
- 22:28 Finish the battle pass when the next one starts
- 24:16 Marvel Snap: Stephen's role and why bots mattered
- 27:04 Advice for emerging designers
- 28:32 Intrinsic and extrinsic motivation
- 30:18 Mobile versus console motivations
- 31:48 What Stephen is playing
- 33:00 Where Western game development is headed
- 34:41 AI for code, human art
- 35:48 The tools Stephen uses
- 40:34 System design, game economy and meta loops
- 42:27 The player journey framework
- 44:25 Collecting feedback on every loop
- 46:09 Soft launches, early releases and copycats
- 47:25 Star Wars: Galaxy of Heroes and the table flip
- 51:13 Going wide and going deep
- 52:08 Which IP would Stephen make next?
- 54:39 Shout-outs and where to find Stephen
Get the weekly operator read in The Operator's Briefing.
Talk it through with other live game operators in the Player Driven Discord.
Get the Tuesday briefing.
Three things every Tuesday: the week's best essay, the next live drop, and one thing we're still arguing about internally.
Ready for more? Join the Discord community or see the Academy.

