Ragnar here. In this inaugural dev blog, I’m going to talk a bit about Book Three. But first: let’s backtrack, back to the beginning.

(Expect ramblings.)

When Book One came out late last year, we were winging it. There was no framework in place for adding episodes; we had designs in place, and some tech, but it was unfinished, unproven. We were jumping into unknown waters with the story and the episodic structure. The waters were deep and dark, and they were freezing.

After releasing the first episode, after digesting feedback and reviews and getting our heads back above water, our second episode grew bigger and more ambitious. We wanted to exceed expectations, we wanted to do better, and we wanted to respond to the loudest criticisms: a slow start, they said; a lack of action and drama, they said; a story that was just getting started when it abruptly stopped, they said. We didn’t disagree.

So we wanted our sophomore release to soar.

Book Two was a monster. A massive, massively ambitious, hugely complex monster of an episode. The foundations were rocky, expectations high, players impatient, team exhausted: we were juggling a hundred flaming torches while standing knee-deep in petrol. I’m still surprised we were able to finish and release Book Two in what amounted to three months of actual production time. Pacing and quality suffered, and there were technical issues, but the team patched those pretty rapidly.

We survived, but barely.

Afterwards, when we could breathe again, we knew we had to slim down Book Three, streamline the process — both for our sake and for the players. We were happy with what we’d accomplished; all of us felt it was a better and more exciting episode than the first one, but it had nearly killed us. We couldn’t do it again.

We planned to ship Realms within a two month window: starting in mid-March, breaking for a one-week Easter holiday, we aimed for release in mid-to-late May. With more focus on quality over quantity, with the episodic tech in place and with improved pipelines and processes, a faster turnaround seemed doable.

Unfortunately, we also had an amputated team: our design director was on parental leave and a few other team members were scattered to the winds. We also wanted to make our next episode as polished and stable as possible, and not rush to get it out the door.

In the end, we couldn’t do it as quickly as we’d hoped. Two months became three before Realms was ready.

Despite the smaller scope and improved quality assurance, however, our launch wasn’t as smooth as expected. There were fewer bugs overall, but a problem with how some decisions had been stored (or not) in Book Two (v2.0 only; the problem was fixed in subsequent versions) meant that some players ended up with erroneous consequences in Book Three. All of our internal testing was done with later versions of the game — where the bug had been fixed — so we never caught this. It took us completely by surprise.

I’m still not sure how we could have caught the bug before release without going back to test older versions of the code — something that was outside the scope of our testing. Our programmers patched the game within twenty-four hours, but this didn’t help players who’d downloaded and played immediately (these were players who’d also played the first, buggy version of Book Two; in other words, our most dedicated fans). To make matters worse, the fix required having to go back to replay parts of Book Two before starting Book Three all over again.

Thankfully, most of our players were gracious and patient — partly, I believe, because we acknowledged and addressed the problem quickly, partly because we brought a patch out the very next day, and partly (I think) because it was such a difficult-to-catch bug. Leeways were given; excuses accepted.

Once again, things worked out, but we’re still kicking ourselves for screwing up.

We have learned an important lesson from this experience: when dealing with player choices and consequences that play out across multiple episodes, keeping those decisions safe and consistent within a constantly evolving technical framework is alpha and omega. We’ve implemented checks and balances to make sure it doesn’t happen again, so…fingers crossed.

The team has been quick to fix problems as they occur and are reported, usually patching within twenty-four hours after an episode goes live, but quality assurance — catching bugs before they’re released into the wild — has been a challenge from the start. One part of this is resources (lack of), the other part complexity: we have a tiny team and a giant, ridiculously twisty game with hundreds of thousands of permutations of various choices and consequences. Unless we do public betas — not ideal for a story-driven adventure — I’m not sure how much better it’ll get.

I also don’t feel our customer support — email and web — is anywhere as good as it should or could be. Our goalpost has been to respond within 72 hours, but we currently have a large backlog of questions going back weeks and months, and our customers are not getting answers as quickly or thoroughly as I want to. The same applies to our communication with backers: at the very least, we should be better at providing insight into what’s going on.

We’re currently doing a couple of different things to address this.

First off, there’s this new dev blog, which I hope will improve transparency and communication. Secondly, we’re changing our support pipeline (yet again) to speed up response times. Last, but not least, we hope to add a community-and-customer manager to the team, in order to improve communication and pinpoint potential issues before they escalate.

(We haven’t posted an ad for this position yet, but if anyone feels the call — and has the right qualifications and experience — let us know. This won’t be a full-time position, but as we release more episodes — and games — that will probably change.)

Another area where we keep falling short is in our localisation process. The French-and-German subtitled versions of Book Three were delayed, and the fully voiced German version was released only last week; six weeks after the release of the English-speaking version.

That’s unacceptable, and while there are reasons for this — some of which were out of our hands — we are ultimately and solely responsible for getting the finished game to our customers.

One reason for the delay(s) is the enormous amount of dialogue that goes into each episode (anywhere from 25-50,000 words). Another is how much text is often added last minute: we don’t lock down all the dialogue until a couple of weeks before release, and we’ll do constant rewrites for gameplay purposes. We obviously can’t finish recording the voice-overs until all the text has been locked down, and while this works fine for the English version — we have an amazingly quick turnaround there — it causes massive headaches for our localisation partners.

We’ll work to improve the pipeline again for Book Four, but because of the incredibly tight production schedule, I’m honestly not sure how much better we’ll do. Episodic voice-over production is a logistical nightmare, especially for a tiny company like ours. We’re learning as we go, and we’ll be better equipped to do this differently in the future — but we’ll most likely never attempt to do episodic localisation again. It’s just too hard.

Looking forward: what are we working on for books four and five?

Here’s the thing: we constantly listen to what our players are saying (and writing), and we’re always taking notes. We even use spreadsheets to collate the feedback, add up the numbers and draw conclusions about what players liked and disliked.

(For Book Three, for example, the opening scene in the House of All Worlds was divisive: Saga received points for cuteness, but most were ambivalent, if not hostile, about the long treasure hunt. Plans have changed accordingly.)

We collate feedback from multiple sources: our own forum (a great place for constructive feedback; people are tough but fair and, importantly, thorough); from Steam (where you have to cherrypick a bit more; there’s plenty of great feedback, but also plenty of noise); from Twitter and Facebook; and from professional reviews. The more sources, the more valuable the data.

We have a vision for the game, of course, and we’re sticking with that. This is our game, we’re not changing direction: we’re building the game that we want to make and play — hoping that enough of you will share our tastes and passions — and we’re not changing everything we’re getting pushback on. But if something specific — a particular puzzle, a character, a dialogue, a location — receives a lot of positive or negative feedback, we pay attention. The story and design for upcoming episodes is not locked down: there’s always leeway to make changes as we go. And we do.

(Say, if players love a certain character, (s)he is more likely to suffer a terrifying fate. That’s just fair.)

The way our episodes are produced, we have a couple of weeks at the beginning of each cycle where we rework the design, rewrite the story (not the major plot lines, of course, but the moment-to-moment details), shuffle scenes around and structure dialogues. Player feedback absolutely affects this. Feedback also affects the features we implement, tweak, bug fix.

This is one of the things we truly love about working with episodes: the flexibility it gives us is wonderful, empowering. We can make our game and still take your feedback into account. So keep this in mind when you comment: we’re listening, and we’re note-taking.

There are a couple of things we’re working on for Book Four that have absolutely been influenced by the feedback.

One of them is how we handle choice-and-consequence notifications.

When the game launched late last year, notifications were mostly non-existent. Consequently many players thought the choices they made in that first episode had no real consequences — because they weren’t signposted. It didn’t help that many of our choices play out long-term rather than immediately after making a decision. For instance, the vast majority of players never realised that a decision point early on (in Storytime) had a major effect on Zoë’s career path in Europolis, and on the gameplay in both that episode and subsequent episodes — because it wasn’t spelled out, it wasn’t obvious.

Soon after the release of Book One, therefore, we decided to patch the game with new choice-and-consequence notifications, making it very obvious — perhaps too obvious — when something had changed in the story. For Book Two, we added the option to switch those notifications off, though they’re on by default.

At that point, it was the right choice. But things changed.

Few are questioning the relevance of choices in the game now. Actions and consequences have played out across three episodes, and it’s obvious that decisions players make actually matter. There are real repercussions to the actions you take, and while they’re not always immediately obvious, they’re often long-lasting. So we’re stepping back on the in-your-face notifications, making the whole mechanic more subtle again.

We’ve also put a lot of work into porting the game to Unity 5. Performance has been one of our weaknesses, and we’ve heard plenty of complaints, so we’ve been keen to find a solution. It’s been a long process but we’ll finish porting everything over in time for the release of Book Four. This means that all previous episodes will run on the new version of the engine, and, hopefully, this will improve performance across the board. We’ll have more technical details about the Unity 5 port in the weeks ahead, probably in an upcoming dev blog.

Thanks for reading: I’ll be back with another dev blog in the near future. Next week, however, you’ll hear from someone else: after all, I’m not the only resident of La Casa de Hilo Rojo with something on my mind.

Thoughts? Comments? Join the discussion!