(Crossposted at alexanderfreed.com)
Understanding your audience will make your writing better.
That's not a statement that should need much explanation, but in brief: Storytelling is communication. It's about reaching out to strangers and presenting ideas to them in a compelling, resonant fashion. You wouldn't write a television show for 10-year-old American children the same way you would write for Japanese senior citizens. You likely wouldn't refuse to revise your Broadway-style musical if most of the audience wandered out midway through your dry run.
Game metrics are nothing more or less than a way of increasing your understanding of your audience. They're not a magical means of producing an engaging story--no one's suggesting that if players respond positively to orcs and plot twists, every game should feature an orc Player Character and a major plot twist every 15 minutes--but they are a powerful means of understanding how players experience narrative.
Sometimes, you may want to adjust how you deliver your story to better match data--if a large percentage of users tend to play in short sessions, you may want to condense your narrative into bite-sized episodes. Sometimes, your goal will be to change the data--if one of your three sidekick characters almost never gets used, maybe you'll decide to dial back that character's abrasive personality. And sometimes, you'll decide that the strength of a story element is worth less-than-ideal results--but you'll do so knowing what it costs you in audience engagement, and be prepared to compensate elsewhere.
Below is a short list of example data points you may want to know to help craft a better narrative. Often, this data won't be available for your game until relatively late into playtesting, limiting its usefulness--but it may well be available for earlier similar games (previous games in a series, games built on the same engine, etc.) And it's never too early to start thinking about what you did right and what you did wrong this time, so you can put that knowledge to work for the next game...
From the start of any project, it's important for designers to have an estimate of their game's typical and optimal play session lengths. This applies to narrative as much as any other aspect of design--no one would suggest that 5-minute YouTube micro-dramas should be written the same as a multi-season television epic, or that a multi-season television epic should be written the same as a "dump the whole series on the viewer at once" Netflix original.
If you can't guess at how long an average player will be staring at the screen, you can't effectively break down your narrative into individual elements (quests, scenes, rooms, whatever); you can't pace plot and character arcs appropriately; you can't know how often to remind the player of the story so far; and so forth.
And what's better than estimates? Actual data. Demand it, love it, and use it.
Who are you writing for now, and who are you trying to reach? Demographic data can be invaluable when helping you determine style and subject matter (if you're writing for adults, you can use a more advanced vocabulary than with children; if you're writing for millennials, your gritty drama about the Weather Underground will need more exposition than if you're writing for Baby Boomers) and of course representation (if you're trying to grow your disproportionately small audience of Latinos, maybe consider how many Latino characters actually show up in game).
That's all obvious--albeit vitally important--stuff. But demographics also impact other metrics. Suppose your current audience is mostly college students. Your storytelling is optimized for lengthy, intense (movie-length, at least) play sessions. If you're trying to reach an audience of adults in their thirties, you may be dealing with a lot of parents of young children--which could mean that you need to change your storytelling style to support shorter play sessions from people apt to drop the game suddenly to resolve one domestic crisis or another.
Related to but distinct from the issue of play session lengths is the question of how many people actually finish your game. Any number of factors may cause a very high rate of non-completion (and the impact on narrative--as well as narrative's impact on that rate--is a subject in its own right), but you can't begin to respond to this issue if you don't know the numbers.
Consider also how these numbers impact storytelling in downloadable content and expansions: if only a third of your audience finishes the original game, perhaps story-based DLC would be better aimed at adding parallel content and "broadening" the game world rather than acting as a direct sequel. Then again, if that one-third of your audience is especially rabid, there may be reasons to simply shrug and support it. Again, there's no easy answer, but looking at the data allows you to ask the right questions.
Do you have a game with multiple Player Character "builds"? e.g., options for classes, weapon loadouts, special abilities, and so forth? Consider play styles when writing so that you don't repeatedly reference the player's pinpoint gun accuracy when half your players are putting all their skill points in martial arts or stealth.
Writing can also boost player interest in underutilized play styles. If no one's even trying to play a healer, maybe an exciting combat medic NPC can help "sell" the audience on that particular concept. Work with your systems design team to see how you can better integrate story and gameplay by using metrics as a starting point.
Distinct from player preference for game systems and playstyle are choices of content with minimal gameplay impact: e.g., choice of player sex / race / species, outfits or skins, companion or sidekick characters, and so forth. If very few players are creating, say, elves, is that a problem with the presentation of elves? Could it be changed with more interesting elf-exclusive content? Or is it not a problem at all? If players tend to dress their sidekick characters in sexually revealing clothes, is that something to push back against--emphatically rejecting the idea of those sidekicks as romantic or sexual objects in the storyline--or lean into, accepting that players want romance / sex and building those opportunities into the characters?
Straddling the line between content choices, play styles, and our next category is nonlinear content. Do players uniformly leave one mission or map for last? If so, is it because the story is naturally inclined that way (in which case, such a result may be desirable) or because Map X is described in the least interesting way upfront? What about repeatable, "grindable" quests--if players favor one variety of side quest grind over others, it's probably for gameplay reasons... but adding story rewards to underused quest types could encourage players to diversify.
Finally, what about cutscenes? Knowing what cutscenes the player skips and in what context (if players tend to start fast-forwarding halfway through, you probably need shorter scenes, more compelling drama, or both) can be an enormous boon.
Finally, for games with branching narrative (and / or branching dialogue), knowing what branches your audience tends to follow can make a world of difference. As always, determining why players choose the way they do can be difficult--did eighty percent of your audience choose to kill instead of save the comedy sidekick because he's deeply annoying, or because you did too good a job justifying it?--but such data may be the single best window into the mind of the player. When you know the player's choices, you know a lot about how the player is reacting emotionally to the story.
Look for data that suggest a lack of narrative engagement as well: Do players choose the "default" response at an usually high rate in certain scenes? They may not care about what's going on, they may not understand their options, or they may be impatient to get back to gameplay.
Remember that narrative branches not taken aren't necessarily a failure--the mere existence of an alternative path adds value to the game and weight to the primary path--but it should play a role in how you write, design, and budget different branches. If very few players take the "evil" option, then perhaps it's best to make most evil options relatively cheap to implement, pulling out all stops only for key rewarding moments.
♦ ♦ ♦
That's six examples--just enough to get you started. Whether you get this sort of data from QA testers, focus groups, convention demos, beta testers, online players, or through some other method entirely, study it. Learn to use it. Trust your judgment to interpret it. And make your game better.
♦ ♦ ♦
One procedural note: I've written quite a lot of these game writing posts, both here and at my personal site, but I'm always, always looking for ideas for future articles. If you find these useful and want to see a particular topic discussed, let me know in comments or via e-mail and I'll see what I can do!