Warning: giant (4700 word) post on basic marketing principles, prompted by some recent discussion on a forum about what makes for a well-retaining game.
A lot of folks, especially in social, seem to use the word “retention” when they should think “conversion.” I tend to think of this as an emotional journey.
You can think of this sequence as going something like this:
Each of these involves a decision to be made by the user — generally, whether to overcome some barriers to participation. They make the decision on the basis of two things: information that they have, and an emotional connection. Of these, the emotional aspect is more important. Antonio Damasio showed, for example, that people who had suffered brain damage resulting in inability to feel emotions also ended up with paralyzing indecision.
The first hurdle is the decision to sample. The decision is whether to even bother trying out the product. The information at hand is whatever player word of mouth, viral messages, and messaging the ads have shown to the prospective user. If it’s bad, then the user may well not bother. If the process of sampling is too hard, then the user will not bother.
You choose a lot, though not all, of your messaging. Think of the possible audience as being like a cloud, or a Perlin noise field, or a crowd of people in a park. You have a circle you can drop on the field, and your goal is to get as dense a spot as possible. If you drop your circle on the beach, then it’s going to attract different users than if you drop it on the hot dog stand. For the math/coder geeks out there: you must subsample the field optimizing for maximum value, and you have a limited number of samples you can take.
You target by choosing where to put your message, and what you say. Note that what you say is far broader than the mere words. These days, we all spend a lot of time optimizing ads and UX — famously, Google tested a bunch of shades of blue to see which performed better. What is happening here is targeting; by tweaking even something as narrow as the shade of blue, we are making microadjustments to the circle — affecting its radius or exact position.
More common is to think in terms of adjusting things like the art style from cute and colorful to green-brown and gory. A shooter game could be made with either style, but you’re talking to different people with each message even though the functional view of the product may be identical.
It’s rare that your message is a perfect fit for everyone. The people at the center of the circle are going to “click better” with it. That means they are going to sample quickly because “it’s just what they were looking for.” People on the edge of the circle will sample more reluctantly.
If the experience after they sample matches the message they were given, that means the experience will have a better or worse fit for them. Exact matches between the message and the experience means you acquire customers who are likelier to love the product. They were “sold from the start.” Assuming a message that matches the product, this means that those users are probably going to be your best customers too: the earliest in are your best bet to monetize and stick.
People who come in later took some persuading. Some of them just didn’t hear the message properly or didn’t happen to hear it early on, and turn out to be just like the first in. But a lot of them — a majority — came in reluctantly knowing it wasn’t a perfect fit, and unsurprisingly they discover they were right over time. If there isn’t anything else better around, then you might rely on satisficing for them to stick around — but they simply won’t be as passionate, and therefore in a microtransaction world are less likely to be upsold.
This serves to point up the importance of branding, even in the social games market where it hardly exists: brands are basically iconified shortcuts intended to trigger emotional reactions. They stand for abstract idealized values, for cherished memories, and so on; it is rare that they stand for rational cost-benefit analysis.
Of course, you’re feeding your message through a straw (especially if all you have is a postage stamp-sized Facebook ad). If you have a one-note product, then all you have to do is aim precisely. But (as we’ll see when I get to retention) a one-note product won’t retain strongly, so is often a bad business bet.
This is why marketers and designers use personas, made-up hypothetical customers, in order to identify possible targets. The goal is to know what these archetypes are like so you can design to satisfy them, and then later market to them. They are basically an attempt to define the centers of multiple circles. You then follow up by running campaigns against different market segments.
Now we see why branding is powerful: it adds more content to your message. You have very limited time to get your message across, after all, and that postage stamp of a picture cannot convey very much. But if you have previously built up mental associations with a logo, a series name, a developer name, or similar, then to the person who knows that brand its mere presence conveys a lot more than what fits in the ad copy and image.
More importantly, it conveys the sort of emotional stuff that is hard to get into an ad. When I see Blizzard know I’ll get polish and superbly tuned fun in a game I have seen before. When I see Jason Rohrer, I know I will get something nothing like what I have seen from him or anyone else previously, and it’ll look intentionally rough and often sacrifice fun for other sorts of emotional reactions. Blizzard and Jason Rohrer basically have opposite brand values, and right off the bat if I see “a new action RPG from Blizzard” versus “a new action RPG from Jason Rohrer” I am going to form an impression that will make a very very large difference to my willingness to sample.
As you can see, this makes branding and marketing part of the game design process. You as a designer don’t get to be unaware or ignorant of this, because in the end, part of the definition of design is “creating something to solve a particular need for a particular type of user.” Sometimes the set of users is very broad, and sometimes it is very narrow.
This is where A/B testing your marketing message comes in. Just remember that testing finds local maxima, not the best results in the entire possibility space. In computer science terms, it’s a crude hill-climbing algorithm. In other words, if you test to find the best spot at the beach, you will find it; but you won’t find the potentially better location at the hot dog stand at all.
You’ll also find that your cost of acquisition will vary based on how well-targeted your campaign is. It may be that ads at the hot dog stand are more expensive, but that you have to shout an awful lot at the beach to get anyone to pay attention. As a rule, it’s going to get more expensive the worse targeted your message is, and you’ll also get “lower quality” customers, meaning customers who are not well-suited for what you are offering and therefore less likely to fall in love with your offering, and therefore less likely to pay.
Barriers to entry
So the bottom line is that you make a statement in a crowded room, and the first people in line will be the ones who heard the statement and agree with it. As word spreads, you’ll get more people in line who agreed but heard it late, or only partially agreed but are curious enough to want to learn more.
Now comes the plain old mechanical part. How hard is it to sample? Remember, you’ve caught their attention by shouting something that briefly popped out of the ongoing noise. They have turned their head. Now how much effort do they have to expend to get close enough to see if they actually like what you have on offer?
The answer had better be “as little effort as possible.”
In the case of a pre-sold, well-targeted user, the barriers can be higher. Their native interest level is higher, so they are going to be willing to go through a bit more pain. For example, a core gamer is willing to get in the car and drive to Gamestop, or to sit through a multiple gigabyte download from Steam. Someone who is just idly curious isn’t going to go through that pain. For them, an extra five seconds on the loading page may be enough for them to say no thanks.
A lot of the digital distribution ecosystem is about reducing barriers to entry via careful analysis of funnels and the like. A ton has been written about stuff like landing page optimization, or whether or not you should require emails at signup. The right way to think about this is that every time the user must make a decision, they weigh whether it makes sense to continue with the process. Your message lit the match, and depending on their affinity for the message, they have varying amount of fuel. How far they go once launched is dependent on how many times they have to recommit.
This is why plugins for web-based games are a bad idea unless you are targeting a heavily pre-sold audience; if the user doesn’t have the plugin, you just inserted a large amount of extra decisions in the pipeline, some of them scary. It’s why you have to worry about which permissions to ask for when installing a Facebook app; if the user has to think about any of the questions asked, that’s another chance to back out. And it’s why it matters how long it takes to load your game: every second spent staring at a blank screen is a chance to get distracted and go do something else.
The actual sampling
All this leads to the actual sampling. You still haven’t really sold the user on the game. Now, however, the user is yours to lose. They hurdled every barrier you put in their way because you told them “there is pizza and ice cream inside.” Now here they are. You had better show them pizza and ice cream.
It’s astonishing how many games fail to do this. If your selling point is that in 200 levels they are going to get to ride a dragon, they had better at least see a dragon with someone on it flying past the newbie spawn point. If your selling point is PvP, then it had better not be hidden behind a giant quest.
Often, users come in having not quite listened to the exact message you provided or with assumptions from prior topics, and come in with other expectations, perhaps because of the brand (let’s say they [ahem] show up expecting to be a Jedi despite everything you said to the contrary), or because of the game genre and its usual interface (heavens forfend, an FPS that isn’t WASD on the PC??). This even led to Damion Schubert’s Law of Player Expectations in the Laws of Online World Design:
Schubert’s Law of Player Expectations
A new player’s expectations of a virtual world are driven by his expectations of single-player games. In particular, he expects a narrow, predictable plotline with well-defined quests and a carefully sculpted for himself as the hero. He also expects no interference or disruption from other players. These are difficult, and sometimes impossible, expectations for a virtual world to actually meet.
You can’t really fight this; you can only attempt to redirect it by offering these sorts of users something else they might like. Sometimes, you can change your design to meet these expectations. Sometimes, you can’t or shouldn’t — often, when attempting to innovate.
So you had better make sure that what they encounter is both familiar and startling; saves the awesomesauce for later but is also amazing off the bat. I think it was Alan Pavlish who taught me to call this the James Bond moment. Think of the first part of any James Bond movie: an amazing action set piece, Bond pulling off something impossible, and enough of the flavor of the entire movie to serve as a teaser. Even when the film series changed tonally, as it did when Daniel Craig took over the role, the opening sequence still served as a mini-movie that summed up what was to come; the parkour sequence in Casino Royale and the opening sequences in the earlier films accomplish the same purpose, even though the vibe is very different.
The sampling period for retail is the demo; the sampling period for a free to play web game is the first session. For frozen yogurt, it’s the little free sample cup they give you. The player has not yet made an emotional commitment; they are on a first date. And that means you don’t have them yet. You are still selling them. You don’t actually have them until they decide to go steady.
So the sample period is where you try to get them to fall in love. A common game best practice is therefore not to design your introductory experience first, with a team that has yet to make the whole game; but instead to do it last, so that you have a team that knows all the nooks and crannies of the gameplay and is expert at working together. Just make sure you left enough room in your schedule to get it done to the highest level of polish possible.
Needless to say, this does sometimes lead to first dates where the other person, um, misleads you about what they are really like. A classic game example would be the original Black & White, which had an incredibly engaging tutorial that lasted a couple of hours — followed by a rather dull game.
Many tutorials or intro sequences focus not on getting the user to fall in love but instead on things like constant “candy” — little rewards that keep the player moving forward. These are great, and serve an important purpose, but relying solely on them misses the point. The player is feeling out the game. They expect a lot of entertainment out of it. And what you need to do is make sure they want to come back after they leave.
An amazing first date ends with planning the wedding in your head, wondering about what your future kids will look like, where you will live, and doodling your names in hearts in the margins of your notebooks. Same with a player: you want them to be plotting what they will do when they return. you want them to have daydreams about future goals. You want them to be curious about the cliffhanger. You want them to be doodling fan art in the margins.
The moment when they decide to return after the first date is what we call conversion. In a retail MMO, this might mean a month of dating, because it came free with the box — you can have a nice long exploratory period without having to make an affirmative commitment. In a social game, it will likely hit after the first session, which means you’ve got five minutes.
(Web folks at this point will say “but that’s not conversion to paying!” Which is true. It’s not. But conversion to paying is basically analogous to this entire process all over again. You have to market the notion to them; get them to sample; get them converted and committed to the idea; and then retain them as a payer. Exactly the same process and psychology. Unsurprisingly, well-targeted users are more likely to pay, users who find that the offering meets a need are more likely to pay, and uses who are committed to that premise tend to pay over and over. Conversion to paying is like “putting a ring on it” — it’s a monetary commitment that reinforces an underlying emotional commitment.)
What makes them stay? An emotionally satisfying experience. It has to be fun, meet their particular entertainment needs, surprise them, leave them curious as to what comes next, clearly have hidden depths to it that they are eager to master even though they don’t quite know what they are…
But that doesn’t mean they stay through the whole thing. A relationship takes work.
Seems like retention is once again one of the hot topics in online gaming, now that the market has largely moved to free-to-play and even packaged goods games are increasingly driven by microtransactions. From a business point of view, you want to keep people as long as possible to get as many opportunities for upsells as possible.
A lot of people confuse short-term conversion with retention. So let’s be clear what I am talking about here. In social games, you often hear terms like “one day retention.” But really, that’s not retention — that’s conversion. The period between clicking a link, ad, or typing in a URL, and the decision to actually stay with the game is fundamentally about something different than the period where a player is playing the game and making the conscious decision to return day after day. The psychology is different.
The good news for gamers is that the key driver of strong retention is good gameplay. That said, it is particular sorts of good gameplay — not all good games will prove to be strongly retentive.
As an example, zero-sum competitive games — think persistent online PvP centric games — can retain highly competitive users with good skills very well. They retain everyone else very poorly, because nobody enjoys repeatedly having their face smashed into the dirt and knowing that it will be worse next time. If the competition is non-zero-sum, then that can feel quite different — think an online FPS — where the slate is wiped clan between matches. Even then though, just an accumulated loss record can be enough to make someone walk away. Historically, online PvP games have never managed to accumulate more than a fraction of the userbases of more forgiving games that use what Jonathan Baron calls a “cumulative character” model. In other words, persistent treadmills retain better than skill-based games, though note that this does not mean grindy games retain better!
The biggest game design factor is to have a good game. But that’s defined very narrowly, as a challenge that offers organic systemic variation. Chess retains very strongly because there are so many possible games of chess. Tic-tac-toe retains poorly because there are so few — only around 120,000 of them. If the gameplay offers emergent variation, you’ll get people retaining longer because the game keeps the challenges and surprises coming.
Games that use systemically generated content (note, I am not saying randomly generated content) retain better than ones driven purely by content, because of the pattern matching factor: random content is easiest to pattern match; handmade content is second easiest; and emergent content is hardest. And as soon as a player pattern matches and concludes they know the entire scope of the game, they are far likelier to become bored.
The single greatest system ever developed for creating emergent content is other players. So multiplayer games where other users insert unpredictable factors into the mix, including their own personalities, retain better.
There are many social dynamics that increase retention (I outlined them all in this presentation, video here). But in particular, social games make heavy use of the notion of obligation. You don’t show up because of how much fun you’re having; you show up because other players are counting on you to do your part (I hear this is why my wife keeps playing Cafe World).
This necessitates, of course, solid communication channels between players. It’s incredibly easy to think that Facebook or email or whatever provide sufficient communication channels for users — and it would be wrong. Remember, you want to minimize barriers, and your users to locate, agree on, and then actually use another tool rather than what the game provides is setting barriers in their way. Only the most passionate users will do so.
This is the source of the many debates on “how much do you listen to forum users” — well, they are playing the game too, just in another way. They may or may not have large overlap with your paying base, but they are still incredibly important, because they tend to be influential — they clearly have loud voices, or they wouldn’t be posting anywhere at all, and therefore would not be influencing many people.
Exposing passionate users to your base as a whole and showcasing what they do is a surefire way to increase the passion levels of the more passive players. In fact, taking care of your hubs is incredibly important, because social graphs are very “clumpy” — and when a hub leaves, they greatly increase the likelihood of taking a large number of other users with them.
Because of this, you want to “play the graph” in a number of ways.
- Make sure players can form a strong self-identity within the game. Otherwise, they will feel like an insignificant cog in a large machine.
- And yet, reinforce the sense that “we’re all in this together” via large-scale shared events. The holiday events, charity drives, and story arcs done as a common practice in live game management are great examples of this.
- Allow users to participate in groups with formal identity, in order to foster a greater sense of community commitment. In fact, provide membership in multiple groups at once (something few games permit!). Multiple group membership ensures that when one group fades, there’s another there to still web the player into the social graph.
- Don’t be afraid of conflicts within the playerbase, as long as they don’t escalate too far — and believe me, it takes a lot. Conflict is a sign of passion and engagement — or as Jonathan Baron put it “HATE is GOOD.”
A strong external community is one of the clearest signs of a long-retaining game, and social game developers should cheer and throw a party on the day that the first external cheat wiki or the like goes live for their game. Once again, players making a commitment like that is a sign of emotional investment — but more, it’s also a point around which the community can congregate. In the long run, players will stick around for each other more than for the game itself.
Moreover, it’s something else that they can’t easily take with them. It’s a long-recognized tactic to give players a sense of ownership: avatars, player housing, customized terrain, etc.
This isn’t to say that regular releases of content don’t help. Of course they do. In fact, you should make them quite regular. Make it a ritual that becomes a habit for users — if a new game item shows up every Friday, then it’ll be a habit to return every Friday, and will show up every Friday even if no game item does. Making your game into a lifestyle choice is a retention win.
Now, just as with the gameplay systems, players pattern match the content. If the content is going to all-too-easily appear to be a reskinned version of something that the players have seen before, then it is going to help with retention less than a new feature will. Remember, players who want to stick around want to keep learning, or keep mindlessly knitting.
If your game is too one-note, then players are going to pattern-match a lot sooner. In working with game grammar, I arrived at formal definitions for “game depth” and “game breadth” (PDF) — where breadth means the number of verbs in the system, and depth means the number of different token types a verb can act on. If content all works against one verb, then players will match it more easily. Similarly, if your game offers too few or insufficiently varied verbs, it is going to retain worse.
In the Laws this was expressed as “have multiple paths of advancement” and “let players switch between them.” Even small games which do this very informally, such as the farming games on social networks, offer at least two big ones — farming and creative decoration — each with its own whole content tree. When players get bored of one, they can switch to the other. In MMOs it was common to speak of “the elder game” as a means of retention. Today, the basic levelling-up game in MMOs is so well understood that it is taken for granted that the elder portion is the game.
Mind you, if you can come up with one astonishing system that has amazing emergent properties — like say, Tetris or chess — you can get away with just one. But most of us aren’t so lucky. And even those games don’t drive daily play for most people, and instead have a pattern of re-engagement that has a higher periodicity.
You drive returns by teasing something new to discover on the other side of a milestone. The rules of thumb are simple:
- Players should always have an achievable milestone in proximate view.
- All milestones should be significant. Meaning, the milestone should unveil a new twist — something new to learn or play with.
Failing to do the first leads to aimlessness. Failing to do the second leads to a game with 500 levels and Fireball I through XVII. Failing to do both leads to what Tadhg Kelly calls busywork: the dreaded grind.
Never forget that point is to get people to enjoy the journey, not just the destination. I have famously said that players don’t want that, and I stand by that. Our job is to sneak the fun into the place where they don’t expect it: the process. Frankly, the milestone (princess rescued) is almost always a letdown compared to the derring-do. All too many RPG-inspired games, though, are all about tallying princesses rescued (crops harvested) (squares cleared) (etc) rather than making the journey interesting — and players pattern match this quickly and see it as a grind. Hence my old statement that “the grind is a state of mind” (huh, that’s quoted a lot). It’s still a state of mind you don’t want to let players slip into.
You can define grind as “having to repeatedly resolve a puzzle you now consider trivial.” Unless you are doing it to meditate (essentially an escape from thinking), this is going to grow dull for you. Ironically, this is why one of the top ways to retain users is to change the rules every once in a while. Not in a huge way — you don’t want to wipe out their investment — but enough so that players need to go relearn some things.
Similarly, a game with unstable equilibrium points also forces people to relearn things. This is why balance is overrated: fun is more important. Finding something that they think the designers didn’t account for makes players feel smart. And feeling smart has a huge overlap with fun.
I don’t have a lot to say about re-engagement. The ideal state is that a user remembers to come back on their own. Re-engagement techniques are about reminding users to come back when they forget. This means there are two types of re-engagement:
- Re-acquisition. This is a user who never really made the emotional commitment. You keep advertising them, mailing them, etc, in hopes that they will commit. Given that you already showed them your awesome first date, this most likely means that the only practical re-acquires will be people who were lukewarm. You can try to get them back by showing them something that they didn’t know about, or telling them you fixed something in the experience that they disliked.
- Rekindling a failed relationship. This is usually because the user fell out of love.
You can try to get love going again by pointing out new things that weren’t there before, or by promising that you fixed whatever you broke that chased them away; you can also appeal to nostalgia. You can, of course, pull back someone who is merely indifferent by pointing out that wasting some time in your game is better than whatever else they are doing.
But then you had better actually show them something. Re-engagement tends to work when the user forgives you for your failings, or discovers there’s more to you than they knew. It tends to fail when you betray them again, or when you don’t actually offer anything new.
Needless to say, you can’t do the above without communicating with the users. And the rules are super simple.
Respect them. If you take something away, explain why. Don’t lie to them. Be honest, and communicate regularly. There’s a whole host of more advanced techniques (IE-only, sorry), around how to manage troubled users, how to deliver bad news, and so on — it all is basically practical politicking. But that’s the core of it right there.
So that’s how you retain users, in a nutshell:
- Make them a promise.
- Show them the reality, and make sure it matches.
- Woo them with an amazing first date.
Learn to read their emotional desires.
Tell them a story, and don’t forget date nights.
Keep the relationship fresh and varied.
Give them moments of glory.
Protect them from humiliation.
Help them learn about themselves and be creative.
Talk regularly and honestly.
You’re marrying your playerbase.