Or, how you’re going to end up buying more games than you thought you would.

More than anything else, digital distribution means change. Changes to the way we shop, changes to the way we perceive the goods that we buy, and changes to the way we interact with publishers and distributors. As we’ve explored previously, one of the likely outcomes of moving to digital distribution is that most people will probably spend far more than they were expecting. But, what does that mean, and why will it happen?

First and foremost, it’s important to remember that the basics of marketing haven’t changed since the first caveman tried to make his particular club look more attractive than everyone else’s. What has changed is level of complexity and sophistication – when all you have a club and no currency, all you can do is to try to convince the troglodyte opposite you that your stick is “better”. When you’re asking a price, you’ll try and offset any quality differences through discounting. As the number of potential buying troglodytes increases, you’ll probably spend more time trying to find and exclude the ones who aren’t going to be interested in the first place. And, so on.

The move to digital distribution marks a subtle but game-changing move; where sellers have only ever had access to transactional information, digital distribution now allows sellers access to behavioural information as well. Picture your average EBGames or GameStop – they may offer a loyalty program but, fundamentally, all they know about you is what you bought in a single transaction and how much you paid for it. Typically, they don’t even know who you are, let alone your age, your lifestyle, or your interests. If all you know is price, all you can do is run discount promotions and above the line marketing activity – anything else is a guess, often a costly one.

The difference between this and online retailers like Valve is massive. To the layperson, Steam is still a shopfront, albeit an online one – they have products, they run promotions, and while you never actually receive any physical goods, you still have the capability to browse and buy things. It’s what’s not so obvious that makes the difference.

In a nutshell, the difference is access to information.  Most people know about Valve’s hardware survey – many actually find the data quite interesting. Importantly though, the information captured through these surveys pales in comparison to what Steam has the potential to track; without being exhaustive, Steam has the capacity to see:

• What Steam-related games you own. Nothing surprising here – reliable DRM in some form (even if very lightweight) is pretty much mandatory to have an market that consumers will trust. Importantly though, it helps build the framework to start profiling you and your interests.
• Which games and content you browse within the Steam shopfront. It’s possible that this might surprise the more naive out there, but most are probably already aware of how easy it is to track paths through and time spent on pages within a site. What might be marginally more surprising for some is how easy it is to track where people point their mouse cursors on a page as they browse and read. By merging this information with your transactional purchasing information and Valve’s promotional activity, among other things they can work out what you were interested in but didn’t buy and what you never even found out about (possibly needing a second-round marketing message).
• How much you generally spend on your system. Given that Steam regularly surveys people’s machines (albeit with their permission), it’s a small step from this data to a proxy estimate of your propensity to spend on technology-related items. With this, it’s a hop, skip, and a jump to creating an estimated “likely spend level” by comparing you against other similar people (based on transactional and behavioural information), thereby creating an estimated Steam-share of wallet. The gap between your spend and your “likely spend level” is effectively the target they’ll be shooting for.
• How much you paid for each of those games. In the online world, this is almost totally unsurprising. However, among bricks-and-mortar retailers, this is almost always impossible to determine – it’s extremely rare that a mass-market retailer will know what each person has ever bought historically from them. The exception is through loyalty programs, which is one of the bigger reasons why so many retailers are increasingly pushing them.
• How often you respond to promotional activity. Again, unheard of in the bricks-and-mortar world. They do often have visibility over promotional response (measured through sales uplift post-promotion), but virtually never at an individual level. Knowing how often you respond to promotional material lets Valve know whether it’s worth marketing to you in the first place. If you’re not biting for particular things, why annoy you needlessly?
• How often you play those games. Obvious, but powerful – frequency of play is a good measure for genre and game “stickiness”. It’s also a strong indicator for interest in other already and about-to-be released games.
• How quickly you score your achievements and how much of completionist you are. This one’s a subtle one – by tracking your personal response to achievements across multiple games in addition to the types of games you play, Bateman’s Nine Basic Player Types aren’t just a theoretical construct, they’re an actual classification metric useful for practical marketing.
• Who your friends are. Easy information to collect, but it lays the groundwork for social network analysis.
• How often you talk with them and which direction the relationship flows. Hate to be the one to tell you this, but not every relationship is equal; practical research shows that relationships within social networks often display directionality. By monitoring the frequency and tendency of communications within a network, it’s possible to identify who the influencers are. And, by doing so, understand who is likely to propagate an attractive message (or offer) within their network when it’s presented to them.

And, that’s the shortlist – remember, this information is available for at least the approaching 3 million active Steam users. Remember also that this is no different to what Microsoft, Sony, and Nintendo track through their respective online platforms.

Combined, this information allows Valve to:

• Profile and classify you into one of a number of standard groups. Given the volumes of data available to Valve, this often involves the creation of a detailed analytical segmentation model to classify users into multiple groups based on their spend, estimated wealth, and behavioural information. This allows Steam to know who you are in a nutshell.
• Identify how sensitive you specifically are to price changes and promotional activity across a wide variety of categories. Typically, this involves various forms of econometric-based price sensitivity modeling. Factoring in the influence of promotional channel (email vs. web vs. in-game), time of year (e.g. Christmas has far greater competition for your dollar spend), and any number of other confounding factors, these models allow an analyst to identify how many more additional purchases your specific micro-segment is likely to make given a 50c price drop, a $1 price drop, a $2 price drop, and so on. 2D Boy’s “pay what you want” experiment using World of Goo almost reads as a textbook example of creating a benchmark price elasticity curve to be used as a baseline across comparable categories and consumers. This allows Valve to intelligently offer (or not offer) discounts at an individual gamer level.
• Predict how likely you are to buy any game offered within the Steam catalogue. Typically, this involves the use of propensity and predictive models, again statistically based. Valve is in a relatively unique camp, in that they can see not only who did purchase but also extrapolate to identify those who didn’t purchase. By analysing web-browsing patterns, it’s possible for Valve to build a sample of both camps, a necessary pre-requisite to these types of models. When applied, these models allow Valve to predict how likely, in percentage terms, you are to buy any given game or general genre, both for existing and un-released games (based on proxy comparisons).
• Understand which games and categories tend to sell well together. Typically, this involves a technique called market basket analysis – it works by understanding the frequency of association between games in aggregate across the entire customer base. Practically speaking, it’s a low-cost way of continually making more accurate recommendations about games that you might be interested in.
• Identify who to market to to have the greatest impact via their social network (as well as who to market to to prevent them from using competing services). As explained above, social network analysis allows Valve to identify the influencers within a given network. Got a game you’re keen to build some viral, gamer-led grass-roots response around without discounting it or giving away large number of “review-copy” freebies? Now you know how.

In short, Valve has the information and data volumes to fairly easily know:

• Who is likely to buy what.
• Who need to be offered a discount and who doesn’t.
• Who has influence and who doesn’t.

At this point, it’s important that I point out that I don’t directly know what Steam is tracking or how they’re using the data they do capture. It’s possible that they’re already doing this – there are a multitude of companies in other industries already doing what I’ve outlined to varying degrees and in varying forms. However, given the overall maturity of the market, the industry, and the relative lack of competition, I doubt that they’re at this level of sophistication yet. Importantly though, it’s not a matter of if Valve will do it, it’s a question of when.

Being blunt, they’d be clowns not to.

Tags: data mining, digital distribution, economics, marketing, pricing analytics

This entry was posted on Monday, January 25th, 2010 at 4:04 PM and is filed under Industry. You can follow any comments to this entry through the RSS 2.0 feed. You can leave a comment, or trackback from your own site.