2009 October « MattMaroon.com

Archive for October, 2009

Platform Changes

Posted in Uncategorized on October 30, 2009 by themaroon

Facebook announced their platform changes minutes after my posting and, as expected, fixed some of the problems I mentioned. They’ve introduced a new roadmap (there’s great overview of it with timelines on Inside Facebook). Overall I’m unsure exactly what the new plans they’ve laid out will mean for Facebook customers and developers. I think much of it is good (as does Eric Eldon at Inside Social Games), some of it possibly bad, and the majority depends on details that have not yet been decided and/or announced.

As far as Facebook user experience goes, it’s mostly a big win. They’re cleaning up the home feed so you’ll no longer be bombarded with friends’ quiz results. That’s good for everyone who wasn’t making a quiz app, and even those people had to know it wasn’t going to last. Facebook is too smart to let third parties degrade their user experience just to turn a quick buck for long. It lasted a few months, which shocked me. I figured they’d stamp it out sooner than they did, but I think it hit them by surprise (and won’t again) and they’ve initiated changes that will allow them to do so in the future. It’s unfortunate too because the feed, used properly by apps, could be great for developers, Facebook users, and Facebook itself, but as with anything so easily abused a race to the bottom occurred and ruined it for everyone.

Virality is the big issue for developers, and I’m not clear how this is going to be affected overall. Most apps other than quizzes don’t rely so much on the news stream as they do invites and notifications, so what happens there is going to have a much larger impact on highly engaging, highly viral apps like the ones we try to produce.

Notifications are going to pretty much disappear. User-to-user notifications are mostly gone, being moved to inbox messages that have to be approved by the sender, so there’s going to be a drastic reduction in those. I can live with that, since it’s something we game developers didn’t rely on overly much. Developers will push users to send Facebook messages where possible, but the fact that you’ll have to type each recipient’s name in by hand will make it much less frequent than invites now, which are easily sent to many friends with just a few clicks using Facebook’s invite selector widget.

App-to-user notifications, on the other hand, were a key driver of engagement. People on Facebook frequently install your app (along with 20 others) and then totally forget about its existence. In all of our games we use the app-to-user notification to alert customers that their special daily mission is now available, or something to that effect, and it has a huge impact. If we disable them for awhile, then turn them back on, it will often lead to a 50% traffic boost in one day.

In the long run, though, these probably have substantially less effect as people disable them or learn to simply ignore them, so again I don’t think they’re too key. And if their disappearance makes Facebook itself a little less spammy for everyone, it might be a long run win for all of us as well. The more people enjoy Facebook, the more they’ll use it, and the more they use it the more they’ll use apps built on top of it.

The real question, in my mind, is what is going to happen to invites. Facebook says “we’re going to consolidate invitations into a special section which will either be located in the Inbox, and/or in the Application and Games Dashboards.”

Implementation is key here. What I fear is that they add them to a special section of the Inbox, and don’t count invites toward the messages total. In that case someone with pending invites, but no other messages, won’t know they have unresolved actions waiting for them. Their screenshot indicates this will be the case.

As you can see, the user has 490 messages (really?) plus 9 updates and 2 invites, yet up top the count by Inbox shows only 490, indicating it excludes invites and updates. That’s really bad. On the other hand, the screenshot (which I’ve cropped slightly) actually indicates 4 invites, not 2, so it is possible that this is just a mockup rather than a screenshot of a functional implementation. Let’s hope that’s the case.

On the other hand, if invites are put in the Games dashboard, as it appears they are considering, and therefore are visible from the user’s home page by default, we may have an overall uptick in acceptance rates which could make up for lost virality on other channels. So I’m crossing my fingers on this one. It’s going to come down to details that apparently haven’t been decided on yet, and the implementation is going to be very key to virality on the platform.

It also appears as if invite texts may no longer be customizable. This will have the massive effect of getting rid of gifts sent via the invite message. That’s been a tremendous driver of both engagement and new users for a lot of games, and moving those to messages, where sending one-to-many becomes much more difficult, will have a massive impact. Developers will push these as hard as possible, and they’ll increase the load of messages in your inbox possibly to the extent that they’ll start annoying everyone, but the difficulty in sending them will ensure they don’t make up for losses on the other big channels.

It will also make app naming and icon selection more important as they will be the only thing differentiating your invite from the rest in someone’s inbox.

On the engagement side, the news seems mostly positive. Notifications been replaced by new dashboards that are enabled by default on the homepage and allow you to show users how many updates they have in your game. This is fantastic.

If there’s one thing people cannot stand, it’s seeing that they have new action items they have not yet resolved. It’s why the progress bars on LinkedIn make people who know they’re probably never even going to use that site again seek recommendations from friends. It’s why I think that inbox invites with notifications in the top bar are great, but without them are near-worthless.

Now if someone bookmarks your app (which they’ve even given us a new button to add to our nag screens to help them do) you’ll be able to bring them back to it better than ever before. That’s a pretty good thing for developers.

Also Facebook is now giving developers the ability to email users directly. I’m not sure how this is going to shake out. Getting customers to give you permission to email them is not trivial. Sending emails to large numbers of people without ending up exclusively in spam folders is not easy either. On the other hand, having direct communication, not limited and/or filtered by Facebook opens new doors. It will take some time for developers to figure out how to best take advantage of this, but we’ll get there.

So overall on the engagement front the changes may be a big win. On virality the verdict’s out, but it may be a small win, or it may be a loss. Potentially a huge one. I suppose we’ll know in a month or two.

A lot of developers think Facebook is out to kill virality in order to sell us users via advertising. Right now a lot of game makers buy traffic from Facebook’s self-serve ads. My company supplements our user growth with this, though the lion’s share still comes in organically. Increasing engagement (and therefore Revenue Per User) would enable developers to place higher bids on customers, resulting in higher CPMs on their ad platform, which are otherwise abysmally-low. Decreasing virality would make ad purchases a necessity to get traffic on the platform.

If that’s their goal, it’s a short-sighted one. It will be a huge deterrent to build on the platform and force developers back to simply building on the web. It will also lead to long-term stagnation. The story of the platform has so far been something like:

1. Guy in basement invents the next big app. At first it was pokes/gifts. Then RPGs. Then quizzes, and most recently farm games.

2. Big companies (lately mostly Zynga) copy them, improve, throw millions of dollars worth of marketing/cross promotional muscle at them.

Remove the guys in basements and who is Zynga going to clone? Once they’ve outbid all of us little guys, who do pretty much all of the innovating, for the traffic then what are they going to spend their ad dollars promoting? The little guys will lose because they won’t be able to afford the ad buys it takes to get traction. Customers will lose because platform innovation will slow drastically. Zynga will lose because they’ll run out of innovative new ideas to copy. And Facebook will lose because in the end they’ll end up with less ad sales, and less profit from the payment platform.

Moreover, it will result in significantly less overall traction on a platform that could probably monetize far better for Facebook through payment processing than ad sales. If Facebook made their payment platform public today, even at a relatively steep fee like 5% of sales I’d integrate it immediately. It would take a much larger platform for that to be a better business model than ads.

So in the end I’m all for Facebook trying to improve their own user experience, and I think in a lot of ways they’ll improve the platform by doing so. I just hope they manage to do it in such a way as to keep virality reasonable, spam low, and independent developers competitive. And if they just want to cash in on the big ad budgets of Zynga, EA, etc., that’s their prerogative. It’s their house, so I’ll pack up my toys and go somewhere else. There’s always that internet thing.

Twist of F8

Posted in Startup on October 28, 2009 by themaroon

Since my company, Blue Frog Gaming, shifted focus to developing Facebook apps nearly a year ago, we’ve seen a lot of changes to the platform. Some of them have probably been for the better, some for the worse, and one in particular has been disastrous: the rise of gifting. In the last few months it has fundamentally broken the platform.

In the early days of the Facebook platform there arose a bit of a spam problem. Developers incentivized customers to send invites to their friends, often even paying them in-app currency to do so. It quickly became obvious that apps employing such tactics could grow to millions of users almost overnight and, in the process, create tremendous volumes of invites that hampered the Facebook user’s experience.

Facebook reacted by doing a couple things. First they banned rewarding users for sending invites (or using other similar platform integration points) and they limited the amount of invites a user could send from any one app to 20 per day. The limit made it impossible to spam your entire list in just a couple clicks and the new incentive rules made users much less apt to spam over time. The idea was that if people were so constricted, and rewarded only when their friends actually signed up rather than when they sent the invite, people would be more considerate, invite only friends who might actually be interested, and the number of unwanted invites being sent out would decrease dramatically.

Those changes helped, but still they wanted to do better, so they later changed the allocations such that the “better” apps got more invites per day (up to 60) and the worst offenders were choked off entirely. Invites are the lifeblood of almost all Facebook apps, so it worked. That seemed to be an improvement until a few months ago when developers figured out how to game the system.

The problem is in how Facebook determines which apps are “better”. They do that mainly by looking at the app’s invite acceptance rate, as well as the rate at which those invites are blocked or ignored. An app’s allocations are based on how their rates compare to the median. When we first started developing about a year ago, the median hovered somewhere around 20-25%.

This method is, it turns out, easily gamed. The original solution, used by Mob Wars and the like, was to create in-game “buddies” that come from an invite and give you in-game benefits. Want to run a job in Mob Wars? You’re only allowed to if you have enough friends in your mob. Want to attack people? Your success in battle will be almost entirely dependent on your mob size. You may need 20 buddies to do a mid-level job and 500 to attack another player with any shot of success. At some point the buddy requirements get so high that the only realistic way to get them is to send buddy invites (via Facebook invites) to large numbers of people who already have the app, which you found on some large “Add Me” lists.

This of course boosted Mob Wars’s invite acceptance rate. When you install the app, you might invite your real friends for a week until you run out. Let’s say you send 100 invites to actual friends, and 10% sign up. Not bad, but you need 20 to do that next job. So you join an add-me list.

Next thing you know you’ve got 400 new friends, all of whom were on the list, and almost all of whom are sure to accept your invite request. So you invite them over the next couple weeks and they almost all accept. Despite the fact that only 10 out of the 100 new people you sent an invite accepted (not a particularly high rate) your total invites were accepted by 400 out of 500, making for an average of 80%. The acceptance rate didn’t go up because the developers made the game better, they just hacked the platform to make sure the people who already used it sent and accepted more invites.

This still wasn’t that big of a problem because most people never got that far, the needed number of friends wasn’t too high, and the active players really didn’t need to send out the full 20 invites per day for very long. It pushed the median invite acceptance rate up as RPGs grew, but not ridiculously so. Then came gifts and the rise of Farm Town.

As far as I can tell, in-game gifting as a platform hack was first figured out by MyFarm, and then later put to spectacular use by Farm Town. The idea is to have one app send many invites to people who play the game every day for as long as they play the game. Combine this with an engaging game and you’ve got an invite acceptance machine.

When you first log into Farm Town, you are taken to the gifting screen, where you can easily choose an item (such as a pig or an orange tree) and then send it to all of your friends. Each gift is a Facebook Invite that has to be accepted in Facebook (not in the Farm Town app) from your requests page. So you click the accept button, go to the app, are given the gift, then have to hit the back button and do it again for every gift you have.

Now just about everyone who plays the game is sending out the full amount of gifts per day. And what happens is that their friends who play accept nearly 100% of them. The ones who don’t click ignore once, and the ones who do click accept 100 times. As a result, the median invite rate for the entire platform has since shot to near 60%, and it isn’t because spam has been reduced, it’s because it’s been increased but the non-spam invites have been increased many times more.

Imagine if your email provider, instead of blocking spam, somehow made you receive twice as much of it but then somehow made half of their customers (of which you may or may not be one) receive 100 times as much non-spam email. Your overall user experience suffers, and in fact even the people who are getting more non-spam are suffering, but the overall % of emails that are spam just decreased by a wide margin.

That’s exactly what has happened on Facebook. Even the people who play the farm games don’t want that many invites. They want the gifts, and that’s fine, but surely nobody wants to log into Facebook, see 20 Farm Town gift invites, and have to accept them one at a time by clicking accept, loading the app, then clicking back to the requests page. It’s slow and painful. From a usability standpoint gifts should be given in-game and not even require acceptance (who would ever not want a free chicken?) but the developers get a tremendous advantage from the acceptance rate so the user has to suffer in order to help them game the platform.

Another problem this creates is that you simply cannot have a top app without gifting, no matter how nonsensical its existence in your game is, or how much it detracts from the user experience or game mechanics. Your game could be the most engaging on the platform, but without a solid gifting system your invites are going to get choked off. Take a look at Restaurant City, which is maybe the most engaging game ever launched on the platform. It’s also a great platform citizen, attempting to get new users largely by making a game so engaging people want to tell their friends about it, rather than by giving them large rewards for spamming friends.

Restaurant City’s invites per day are down to 12 and it’s a Verified App. Being Verified means it gets boosted 2 buckets in the allocations. To put that another way, if it were not Verified it would be getting 6 per day.

Scroll down through the top games on the Facebook platform and install them all, as I have, and you’ll see a trend appear very quickly. Almost all of them dump you immediately onto a gifts page every time you log in. This is no coincidence. It’s because it’s the only way to keep reasonable allocations.

The solution is simple: only count requests sent to someone for the first time when determining the median. Or better yet, switch to a better metric entirely like engagement. Facebook knows how much time my users spend on my apps (or at least they could if they so desired). Why not base my allocations on that rather than forcing me to beg my customers to send a dozen gifts with every login?

I wrote 90% of this post over a month ago, and was prompted to write it today by the fact that the platform is probably changing in about an hour and a half. I’ll admit, I’m somewhat dreading it. It’s been a long time since Facebook made a change that made me think more highly about developing on their platform, and with each knock to developers’ allocations they’re going to push more and more people off. I worry that if trends continue, at some point we’ll be left with just Playfish making new games and Zynga cloning them and then marketing them to the top. Our new game on Facebook is doing very well, and we’re already thinking about how to parlay that into a standalone web success. I know we’re not the only app developers thinking along those lines these days, and Facebook should take that seriously, if for no other reason than to give Zynga someone’s games to copy.

An Internet Advertising Primer

Posted in tech on October 19, 2009 by themaroon

I’ve been talking to some people a bit about the internet advertising industry lately, and it’s occurred to me that many people even in the web industry have little or no understanding of how it works. To save myself the time of repeating it to each, because a lot of people may find it interesting, and because some of the topics I’ll be covering here in the very near future relate to it, I thought I’d write up a quick primer on the various terms, how they’re defined, and what they mean in practice.

Here are the basic terms, in logical order rather than alphabetical:

Ad Terms

Impression: This is what one view of one ad is called. Every time you see an ad anywhere, that’s one impression. A webpage with multiple banners might serve up one impression each for many different ads.

Publisher: The website displaying ads. The term is a holdover from the days of print marketing, but probably could use some updating nowadays. It’s a pretty big stretch to call the authors of most blogs or a social network "publishers".

CPM: Cost per 1,000 impressions (M being the Roman Numeral for 1,000). This is the amount a publisher gets paid to serve up 1,000 ads. For instance, I might purchase 1,000,000 impressions (1,000 times 1,000) of an ad on Facebook for 12 cents CPM, so my total bill would be 1,000 * .12, or $120.

CPC: Cost per click. Many ads are paid for only when someone actually clicks them, regardless of how many impressions are shown. This is very typical for ads on Google, Facebook, Myspace, and many other platforms. So if I pay Facebook 30 cents CPC to run my ad, and only one person clicks it, then it doesn’t matter whether they showed it once or one million times to get that one click, I still pay exactly 30 cents.

CTR: Clickthrough rate. The number of impressions of an ad that are clicked on by the viewer divided by the total number of impressions of that ad shown. If Facebook shows my ad 100 times, and 3 people click on it, that’s a CTR of

3/100 = .03

This is also sometimes expressed as a percentage (which in the above case would be 3%) but if so will need to be converted to a decimal for calculating eCPM, which I’ll get to after…

CPA: Cost per action. This is the amount of money an advertiser pays a publisher to drive one user who performs a certain action. For instance, suppose Netflix pays publishers $20 every time an ad banner on the publisher’s site gets Netflix one new member who signs up for a monthly service plan. Netflix is paying $20 CPA.

What constitutes an action is generally determined in advance by the advertiser, and in most cases involves a user purchasing something. (There’s also a similar method called CPL, or cost per lead, which I’ll ignore since it works about the same way but usually doesn’t necessitate a purchase.) This is a very common method of marking online through affiliate programs.

eCPM: The effective cost per 1,000 impressions. This is a useful metric when estimating the cost to run ads paid for by the CPC or CPA method rather than CPM. The formula for calculating it is:

eCPM = CPC * CTR * 1,000

eCPM = CPA * CTR * 1,000

So let’s say I’m paying Facebook 30 cents CPC, and I get a clickthrough rate of 1%, then I am effectively paying Facebook

$.30 * 0.01 * 1,000 = $3

eCPA: Effective cost per action. The effective amount an advertiser purchasing ads by any method other than CPA (for instance by CPC or CPM) pays for one action.

Ramifications

For publishers everything comes down to eCPM, for advertisers it all comes down to eCPA. Once you understand those two simple rules, the rest flows naturally.

Publishers have limited inventory (i.e. impressions) and want to maximize their profit, and the way to do that is to run the ads with the highest eCPMs. Even if the advertiser is paying by the CPC method, the smart publisher is automatically calculating which ads are paying the highest eCPM (which, remember, is CPC * CTR * 1,000) and displaying them first. (The same applies to CPA, but I’m going to continue just using CPC in my examples both for clarity and because it’s considerably more common among major ad platforms.)

Advertisers are all about maximizing the productivity of their ad budget. They seek to allocate their funds to the places that have the lowest eCPA. If that eCPA is less than the profit they make from the acquired customer, they’re making money. If not, they’re going to end up poorer.

So regardless of the tracking method used, every intelligent advertiser is looking at their eCPA, and every intelligent publisher their eCPM. This is why when I see ludicrous articles like this one about how we should stop using the CPM method (oddly by someone who formerly ran the Interactive Advertising Bureau and should know better) I laugh because the payment method is entirely irrelevant. It all boils down to the same thing in the end.

As an advertiser bidding by the CPC method, there are therefore a couple things you can do to maximize your profit. For one, you can pay more. If you double your CPC bid, you’re doubling the eCPM you’re paying the publishers, and therefore your ad will be significantly more likely to run. If your profit margin drops to half of what it was, but your volume quadruples, you come out a winner.

But since no advertiser is ever in a hurry to pay twice as much when there are other options available, it’s usually more common to focus on improving the CTR. Double your clickthrough rate and you’re paying the ad network twice the eCPM, however you’re still paying the same amount per click, while only using up half as much of the advertiser’s inventory. It’s a win-win for the advertiser and the network, which is why you’ll find a lot of helpful tips on the networks’ sites (and even frequently get contacts from staffers) on how to improve your CTR.

There are two very good ways to increase your CTR. One is to make your ad more appealing. Experiment with different images, sizes, and texts. This is both art and science, and time-intensive to boot, but it’s a powerful way to maximize your return on investment. If you are purchasing sufficient volume you can multivariate test the results of different ad texts pretty quickly.

Another common method of improving CTR is to improve your targeting. With search-based ads, such as those found on Google, find the keywords that convert best and most cheaply. On Facebook or Myspace, target users who fit your desired demographics and interests, and tailor your ads specifically toward the keywords being targeted.

So there you go. You now know everything you really need to know about how internet advertising functions at a basic level. There of course are thousands of little tips and tricks that people have collected over the years, some of which they’ve shared, many they’ve probably not. I’d recommend searching around for all of the advice you can find and then figuring out the rest through trial and error.

2 Comments »

Still Alive

Posted in Uncategorized on October 15, 2009 by themaroon

I know, it’s been a rather long while between updates for me. I’ve been busy lately. Really busy.

My company, Blue Frog Gaming, put out a new game a month or so ago called Starfleet Commander and it’s been doing really well. The volume of support work has been huge, 4 or 5 times all of our other games combined (though the other games add up to much more traffic). Some of its due to the fact that we just haven’t ironed out a lot of the kinks yet, but most of it’s because people are just getting way more into it. We’ve got significantly higher engagement, growth, and revenue per user than we’ve seen before.

So I haven’t had much time for writing, but what I have had time for, I’ve gone a little overboard with. I’ve been holding it back because I’m always loathe to publish some 20,000 word blog post, but I’ve got a good series about the state of internet advertising, especially as related to social media, that I think I need to at least serialize and get out there. It starts with a rather lengthy primer that’s totally skippable to anyone who knows the difference between CPA and eCPM, but is a great intro to how ads work online for everyone else. I’ll probably try to get that live in the near future.

I’ve got some stuff to say about the public health care debate too. I know, who doesn’t? But it’s one of the most fascinating and complex issues our country has dealt with in living memory, and I find most of the arguments both for and against it to be focused on the wrong things, and the task of arriving at any firm conclusion to be almost entirely impossible. I also plan to write about why Libertarianism has become nothing but a religion for people who think they are too smart to buy into the other ones. It may be Target to Christianity’s Wal-Mart, but it’s still just a big-box retailer.

So more coming soon, I promise.

MattMaroon.com