An Interview with Ryan North (part two)
Ryan mailed me back today with the answers to the interview questions! Of particular interest was his answer to the mass-bidding question; hit "Read More" to see what he says.
PWT:
One of the themes that keeps on coming up on the Project Wonderful Talk board is expansion of PW outside of the webcomic community, because we think attracting some big-money advertisers would give the economy a nice boost.
Obviously we don't have the full view of the system as it stands like you do, so can you say whether or not you think it'd be a good idea to promote the system to larger companies? And if so, are you actively promoting it now, or planning to - and is there anything we, as established publishers and advertisers, can do to help?
RYAN:
It's definitely something that's on the horizon. The plan is to make the system as perfect as we can for regular users, and then start approaching larger companies. The nice thing is that since the members on Project Wonderful range from casual users to power users, the feature request emails I get cover a lot of ground, so I'm pretty sure I know what's needed.
I think that larger companies would give the economy a boost, but I don't think Project Wonderful needs them to survive. The system works right now on a more member-to-member basis, and I wouldn't want or expect that larger companies would replace that.
PWT:
Can you tell us how this all got started? I heard (and I'm not sure this is true) that you got unjustly kicked off Google AdSense and they wouldn't tell you why, so you said "Screw you guys, I'll make my own network!" If so, jolly good show. Damn the man.
RYAN:
My friend Tim and I were talking about online advertising and how much it tends towards being terrible, and we sat down over a few days and sketched out how we'd improve it, how we'd reinvent it if we were starting from zero. That was the basis for Project Wonderful - of course, it changed somewhat since then and when it launched, but all the big ideas were there. We changed some things as we went along and discovered what worked and what didn't.
One of the original ideas we had was a sort of horserace ad box: based on the bids, the system would determine which were the most valuable spots: if you bid the highest your ad would be displayed there, in the most valuable slot. If you bid second-highest, your ad would move to the second-most-valuable spot, and so on. It ended up being a lot of overhead for a system that was very confusing in realization ("I bid more and my ad moved LOWER? Why?") and subject to noise, so we dropped it. When we introduced bidding across slots, that was pretty much an expression of the same idea that we had in horserace ads ("here's what I want to pay; make it happen, Project Wonderful), only a whole lot more intuitive.
PWT:
I think the growth of publishers signing up for Project Wonderful has taken a lot of us by surprise. How happy are you with the way Project Wonderful has been progressing? Has it met your expectations? Exceeded them?
RYAN:
It's definitely exceeded them. When we launched we weren't certain that any of the ideas in Project Wonderful would work - so I was prepared for failure from day one. It's been gratifying to see how it's been embraced.
PWT:
You seem like a pretty stress-free guy. Do you have, or have access to, plenty of resources to help Project Wonderful expand while still keeping it fun for you personally? Is this still a one-man show, or have you / are you considering hiring people to help you out? For that matter, is PW taking up enough of your time to make that sort of thing necessary, and earning enough money to justify it? (basically I just want reassurance that you're not gonna have to give up making other Cool Things and, y'know, being a person, just so PW can keep going)
RYAN:
Ah, I only play a stress-free guy on the internet! Project Wonderful does take up a bunch of my time, but my sentiment has always been that when the workload gets to the point where it can't be effectively managed with the people we have now, it's time to get more people.
PWT:
The facility to place lots of bids at once, possibly for a premium price, has been discussed quite a bit on the forums and in the articles.
How do you feel about such a feature? Would it be in keeping with the spirit of the project and what you intended? Is there anything in the works?
RYAN:
Yeah, there is a system in the works. It's been built and is being tested now. This is going to be a bit of a more detailed technical answer, so you might want to skip it if you're not interested! Here's how it's going to work:
The basic idea is to look at bidding at various levels. Right now there are two levels: bidding on a specific slot, and bidding across slots. The next logical step up is bidding across sites, across ad boxes. We've realized that as "campaigns". The idea is you do a search like you do now, only you click on the "create campaign" button that's displayed. From there, you can set your campaign limits (which are much like bid limits: max bid, max expense, things like that), and set options like "Bid only on these specific results" or "Bid on any ad box that meets these criteria, now and in the future". Once you're happy with your options, you start the campaign, and the system does the bidding for you: figuring out the best ad boxes for your ads as determined by the criteria you set, and then bidding appropriately.
The system is going to be available to all users, but I imagine the power-users will be the ones to take most advantage of it. It does take the day-to-day bidding a bit out of your hands. Once it's been internally tested we're probably going to roll it out gradually - testing with a few volunteers first before making it more available.
My main concern is that since the system is making more decisions automatically, there has to be a checks in place ensuring that the system can't be gamed. Right now real people are placing every bid, but if someone creates a campaign that says "place 5 bids across the top 5 sites, and I don't care what they are", there is an implicit reward there for someone to try to get their ad box into those top 5 sites, so that bids will be placed on it automatically. We don't want a system in place that rewards people trying to cheat like that, so we'll be watching for things like that closely and trying to design it as much as possible to remove any such incentives.
That said, I think that one of the main advantages of campaigns will be for the smaller sites: rather than having to place bids on individual sites, you'll be able to earmark, say, $5 for the smaller sites and have the bids there handled automatically. You'll still be able to view the statistics of individual bids and the campaigns themselves, of course.
PWT:
One of the things that we've drawn attention to on PWT is the supply to demand ratio. In your opinion, are there enough advertisers to go around, or have we given too much supply for their demand? And, having access to all of the data you've got access to, can you say whether there's any sort of formula for figuring out how many ad boxes a publisher should put on their site?
RYAN:
Across the network itself, no, definitely - I don't see any problems with supply and demand. The only places I've seen problems are on individual sites: people create an ad box, it gets $1.50 a day, and that's exciting! The next day they've got four more ad boxes on their page, and they're hoping that'll mean they'll get four times the money. The reality is that it is a supply-demand market, and you've just quadrupled the supply - bid prices are going to go down, not up.
I've found that if you've got an ad box with, say, four slots at $.02, you'd probably do better to have one slot - odds are the price on that will be greater than 8 cents and you'll be making more money. There was one site that had four square ad slots, I think, which were languishing around the $1 mark, but when he dropped that ad box and put up a new one with only one slot, the four advertisers who were competing for the slots now had just the one area to bid on, and the bid price went up by over a factor of ten. Not bad!
There's no formula, beyond "restrict supply to increase prices, increase supply to lower prices" and that tends to hold pretty hard and fast. You can keep that in mind when creating ad boxes: some sites have a high-priced ad box in a premium location, and then a larger, cheaper box in a less primo location for advertisers on a tighter budget - that way they capture both the people who want high exposure and the people who are looking for a longer, cheaper campaign.
PWT:
What's next for Project Wonderful? Are there any top-secret things in the works that you wouldn't mind disclosing to a forum full of people so obsessed with PW that they... well, made a forum, for one? And while we're on the subject... just how far off the mark were we with the "Wild Speculation" article?
RYAN:
The campaigns code discussed above is really the next big thing! I'm hoping that when it does go up it's not too far different from what's described above. As for the Wild Speculation article - I'll just say that there were some things there that were on the list of things to do, and some things that weren't, but now are. :)
PWT:
Can you give us any tips for the best ways to go about buying and selling ads? We've theorised, but it'd be nice to get some advice straight from the T-Rex's mouth.
RYAN:
Well, the nice thing about Project Wonderful is that you've got access to pretty much all the data I've got, so I can't really offer any grand insights.
EITHER THAT OR I AM KEEPING THEM ALL FOR MYSELF
PWT:
One of the themes that keeps on coming up on the Project Wonderful Talk board is expansion of PW outside of the webcomic community, because we think attracting some big-money advertisers would give the economy a nice boost.
Obviously we don't have the full view of the system as it stands like you do, so can you say whether or not you think it'd be a good idea to promote the system to larger companies? And if so, are you actively promoting it now, or planning to - and is there anything we, as established publishers and advertisers, can do to help?
RYAN:
It's definitely something that's on the horizon. The plan is to make the system as perfect as we can for regular users, and then start approaching larger companies. The nice thing is that since the members on Project Wonderful range from casual users to power users, the feature request emails I get cover a lot of ground, so I'm pretty sure I know what's needed.
I think that larger companies would give the economy a boost, but I don't think Project Wonderful needs them to survive. The system works right now on a more member-to-member basis, and I wouldn't want or expect that larger companies would replace that.
PWT:
Can you tell us how this all got started? I heard (and I'm not sure this is true) that you got unjustly kicked off Google AdSense and they wouldn't tell you why, so you said "Screw you guys, I'll make my own network!" If so, jolly good show. Damn the man.
RYAN:
My friend Tim and I were talking about online advertising and how much it tends towards being terrible, and we sat down over a few days and sketched out how we'd improve it, how we'd reinvent it if we were starting from zero. That was the basis for Project Wonderful - of course, it changed somewhat since then and when it launched, but all the big ideas were there. We changed some things as we went along and discovered what worked and what didn't.
One of the original ideas we had was a sort of horserace ad box: based on the bids, the system would determine which were the most valuable spots: if you bid the highest your ad would be displayed there, in the most valuable slot. If you bid second-highest, your ad would move to the second-most-valuable spot, and so on. It ended up being a lot of overhead for a system that was very confusing in realization ("I bid more and my ad moved LOWER? Why?") and subject to noise, so we dropped it. When we introduced bidding across slots, that was pretty much an expression of the same idea that we had in horserace ads ("here's what I want to pay; make it happen, Project Wonderful), only a whole lot more intuitive.
PWT:
I think the growth of publishers signing up for Project Wonderful has taken a lot of us by surprise. How happy are you with the way Project Wonderful has been progressing? Has it met your expectations? Exceeded them?
RYAN:
It's definitely exceeded them. When we launched we weren't certain that any of the ideas in Project Wonderful would work - so I was prepared for failure from day one. It's been gratifying to see how it's been embraced.
PWT:
You seem like a pretty stress-free guy. Do you have, or have access to, plenty of resources to help Project Wonderful expand while still keeping it fun for you personally? Is this still a one-man show, or have you / are you considering hiring people to help you out? For that matter, is PW taking up enough of your time to make that sort of thing necessary, and earning enough money to justify it? (basically I just want reassurance that you're not gonna have to give up making other Cool Things and, y'know, being a person, just so PW can keep going)
RYAN:
Ah, I only play a stress-free guy on the internet! Project Wonderful does take up a bunch of my time, but my sentiment has always been that when the workload gets to the point where it can't be effectively managed with the people we have now, it's time to get more people.
PWT:
The facility to place lots of bids at once, possibly for a premium price, has been discussed quite a bit on the forums and in the articles.
How do you feel about such a feature? Would it be in keeping with the spirit of the project and what you intended? Is there anything in the works?
RYAN:
Yeah, there is a system in the works. It's been built and is being tested now. This is going to be a bit of a more detailed technical answer, so you might want to skip it if you're not interested! Here's how it's going to work:
The basic idea is to look at bidding at various levels. Right now there are two levels: bidding on a specific slot, and bidding across slots. The next logical step up is bidding across sites, across ad boxes. We've realized that as "campaigns". The idea is you do a search like you do now, only you click on the "create campaign" button that's displayed. From there, you can set your campaign limits (which are much like bid limits: max bid, max expense, things like that), and set options like "Bid only on these specific results" or "Bid on any ad box that meets these criteria, now and in the future". Once you're happy with your options, you start the campaign, and the system does the bidding for you: figuring out the best ad boxes for your ads as determined by the criteria you set, and then bidding appropriately.
The system is going to be available to all users, but I imagine the power-users will be the ones to take most advantage of it. It does take the day-to-day bidding a bit out of your hands. Once it's been internally tested we're probably going to roll it out gradually - testing with a few volunteers first before making it more available.
My main concern is that since the system is making more decisions automatically, there has to be a checks in place ensuring that the system can't be gamed. Right now real people are placing every bid, but if someone creates a campaign that says "place 5 bids across the top 5 sites, and I don't care what they are", there is an implicit reward there for someone to try to get their ad box into those top 5 sites, so that bids will be placed on it automatically. We don't want a system in place that rewards people trying to cheat like that, so we'll be watching for things like that closely and trying to design it as much as possible to remove any such incentives.
That said, I think that one of the main advantages of campaigns will be for the smaller sites: rather than having to place bids on individual sites, you'll be able to earmark, say, $5 for the smaller sites and have the bids there handled automatically. You'll still be able to view the statistics of individual bids and the campaigns themselves, of course.
PWT:
One of the things that we've drawn attention to on PWT is the supply to demand ratio. In your opinion, are there enough advertisers to go around, or have we given too much supply for their demand? And, having access to all of the data you've got access to, can you say whether there's any sort of formula for figuring out how many ad boxes a publisher should put on their site?
RYAN:
Across the network itself, no, definitely - I don't see any problems with supply and demand. The only places I've seen problems are on individual sites: people create an ad box, it gets $1.50 a day, and that's exciting! The next day they've got four more ad boxes on their page, and they're hoping that'll mean they'll get four times the money. The reality is that it is a supply-demand market, and you've just quadrupled the supply - bid prices are going to go down, not up.
I've found that if you've got an ad box with, say, four slots at $.02, you'd probably do better to have one slot - odds are the price on that will be greater than 8 cents and you'll be making more money. There was one site that had four square ad slots, I think, which were languishing around the $1 mark, but when he dropped that ad box and put up a new one with only one slot, the four advertisers who were competing for the slots now had just the one area to bid on, and the bid price went up by over a factor of ten. Not bad!
There's no formula, beyond "restrict supply to increase prices, increase supply to lower prices" and that tends to hold pretty hard and fast. You can keep that in mind when creating ad boxes: some sites have a high-priced ad box in a premium location, and then a larger, cheaper box in a less primo location for advertisers on a tighter budget - that way they capture both the people who want high exposure and the people who are looking for a longer, cheaper campaign.
PWT:
What's next for Project Wonderful? Are there any top-secret things in the works that you wouldn't mind disclosing to a forum full of people so obsessed with PW that they... well, made a forum, for one? And while we're on the subject... just how far off the mark were we with the "Wild Speculation" article?
RYAN:
The campaigns code discussed above is really the next big thing! I'm hoping that when it does go up it's not too far different from what's described above. As for the Wild Speculation article - I'll just say that there were some things there that were on the list of things to do, and some things that weren't, but now are. :)
PWT:
Can you give us any tips for the best ways to go about buying and selling ads? We've theorised, but it'd be nice to get some advice straight from the T-Rex's mouth.
RYAN:
Well, the nice thing about Project Wonderful is that you've got access to pretty much all the data I've got, so I can't really offer any grand insights.
EITHER THAT OR I AM KEEPING THEM ALL FOR MYSELF
An Interview with Ryan North (part two)
