In Armchair Analyst
Half-man and half-platform, the armchair analyst is a mysterious presence at the center of deltaDNA’s insight team. Sent from the future to save game-makers from themselves, he is an enigmatic figure and definitely NOT this man.

Whether you’re starting a new career or learning to skateboard, most of the advice that you’ll receive in your lifetime comes down to ‘proceed with caution.’

Unfortunately, more often than not, being overly cautious with their event-tracking is what causes developers difficulties in the long run.

Ignore your instincts

Developers without games analytics experience often borrow techniques and methodologies from web analytics, where you want to track every click and page view. Take it from me – that approach just doesn’t work for games.

Tracking every action in games takes up a huge amount of storage but, more importantly, it makes your data impossible to analyze.

To demonstrate just how easy it is to create a catastrophically noisy dataset with millions of data points, let’s look at two examples of common analytics missteps:

  1. User Flows

    A game’s User Interface (UI) creates a lot of headaches for its developers. So, keen to work out where they fall down, developers often decide to track how players are exploring the UI and moving through its various branches.

    Sounds reasonable ✔️
    Looks horrendous ✔️
    Is useful ❌

    When there are so many different options and ways for players to perform similar actions, you just end up recording a million different paths that yield nothing in terms of meaningful insights. In builder games, for instance, it really doesn’t matter how many times a player cycled through the same sub-menu and hovered over each material. All you really want to know is what they eventually built – after all the dithering.

  2. Collecting/Harvesting

    It’s very common for developers to track certain individual actions when there is absolutely no need to do so. This is especially common (and especially problematic) in games that involve harvesting resources as a core gameplay element.

    At a glance, the logical way to track a player’s mammoth 126-coin harvesting spree is to send a data point for every coin earned. Tracking repetitive actions with this level of detail is completely unnecessary. Instead of bloating your dataset with 126 individual events, you could simply wait until the harvesting is over and send 1 tiny data point: “this player harvested ‘x’ coins over ‘x’ time.”

Be selective, use aggregates

The key to good games analytics is being brave enough to pinpoint the important information and leave the rest out. By their nature, all games involve a huge degree of repetition. As such, tracking aggregates rather than individual actions is a very effective way of reducing the size and noise of your datasets:

  • Instead of tracking all clicks made around an in-store offer, track offers abandoned and transactions completed.
  • Instead of tracking every gunshot made in an FPS, track total hits and misses per mission.
  • Instead of tracking every tree chopped for wood, track the amount of wood harvested per session.

Most actions in games can be allocated to larger processes, without needing to be tracked in isolation. Send individual data points that denote aggregates of identical actions or the resolution of processes. It works, honestly.

The more you track, the harder it gets

As tempting as it sounds, you can’t just track everything and work it out afterwards. Without 2000 bright-eyed and hugely overqualified interns on hand to pore over the mulch, you just won’t ever get through it all. Trust aggregates!

For more on this, and other ways that you might be ruining your analysts’ lives,* check out Dr. Dan Bergin-Holly’s talk from Spring GIAF 2018.

*His words, not ours



If you have any questions, about the contents of this piece or anything else, contact us at [email protected] and we’ll connect you to the relevant person.


Recent Posts

Leave a Comment

Start typing and press Enter to search