Why Ze Frank is Right and Wrong About Rocketboom

Ze assers that hits are inflated by autoplay and TiVo -- that these cause server downloads that no one sees. How does anything you have written contradict this in any way whatsoever? It's like you're talking right past him. This is very frustrating for me as a reader. I have no idea who is right, but you say Ze is wrong while providing no real refutation of what he wrote.

Server logs can't show you who is watching and who is not. If the thing is on autoplay, someone comes to the page and hits "Back" 10 seconds after it loads, it will still show up in the logs.

Are you sure the number of files downloaded in the spreadsheet is from one server and not from all four? Let's take the Oct 2nd numbers as an example. According to your spreadsheet, 37,988 files were downloaded and each file was 31.66 MBs. That means the daily bandwidth required to serve these files would be 1.2 Terrabytes. Even if they had a 100 Mbps pipeline to the server, their maximum theoritical throughput would be 1.08 Terrabytes per day. I doubt they were coming so close to their max capacity as this could introduce huge latencies to the service.

Either I'm missing something - or the 37K+ downloads number is for all four servers, not just one.

The Tivo numbers are a different server. And number of downloads is for full downlaods, so it's not for the partial ones that run 10 seconds and then cut off. That's what Rob was trying to parse out when he looked at the size of the iTunes file size and the size of the file that was being downloaded by Andrew's server files.

Whether you can know if anyone is really ever watching them as they play is impossible to tell for Andrew or for Ze or for the Nielsen people meters.

Per the questions from Someguy. Rocketboom is running 1 GBps servers not 100 MBps server. So there should not be an issue with bandwidth as that still gives at least a 20x Overhead when I looked at the worst case scenario and over a 50x overhead on a typical day.

Yes the numbers were for one server not all four. I was logged into ww1 when I looked at the logs. Rocketboom is running 4 servers with the files mirrored for serving the shows. They run a script on the main server that redirects requests on a rotating basis. Request 1 goes to ww1, request 2 to ww2, request 3 to ww3, request 4 to ww4, request 5 to ww1 and so on.

So again the numbers were for just one server and not all 4. I did multiply the numbers by 4 which is the correct thing to do here. The file size was average file size. Not that each file only downloaded xx%. I did adjust the total down to account for the partial downloads.

Per comments about who is watching. True no one has a real good way of seeing who is watching. But according to Rocketbooms main server logs the largest aggregator hitting Rocketboom is iTunes. With iTune if you do not watch any of the last 5 shows downloaded it stops downloading until you manually click on the Gray Exclamation mark and tell it to continue. So while no one know for sure how many people are actualy watching - you can be pretty sure that those subscribed with iTunes are watching at least some of each show - or else it would stop downloading the show.

Given the percentage of requests from iTunes it is safe to say that at least 35% of the audience are watching each show. Not saying the other 65% are watching or not - because there is no way to tell. But again iTunes has the built in kill switch that stops the downloads - and that is not something I have seen anyone mention in this debate.

Please let me know if you have any other questions about how the logs were parssed or any other assumptions that were made. These numbers are all approximates. Could I be off by 10% in one direction or the other - sure. But I do not think it is likely I am off by more than that either way.

Rob @ podCast411

rob@podcast411.com

no Heather, that's not correct.

revver's view counts state precisely how many people watch to the end. ze's article states that a show that was downloaded 170,000 times on revver (versus the 156,980 Rocketboom figure in your spreadsheet) resulted in 27,000 complete views. actual views to the end, where the ad is.

in the article above you make the same mistake rocketboom makes. you compare ze's 30,000 complete verified views to rocketboom's 211,000 daily downloads.

per their privacy policy, tivo, like revver, also tracks viewing/watching habits.

it seems like the data is out there if you want to investigate it. figuring out the "downloads to views" ratio is important business stuff. it fundamentally shapes the medium.

if 10 times more people are watching at the beginning of a video than the end, then that's where the ads will go.

if you're like rocketboom and have a policy of "no preroll, ever" then you're going to have to divide your ad price by a factor of 10 for the privilege of maintaining that policy.

I always though the Alexa numbers were more relaible than that.

Rob - Thanks for the response. The numbers make sense if RocketBoom has a 1 Gbps pipeline. I do agree that multiplying by 4 is the correct thing to do in this scenario.

Either way, Ze Frank certainly seems to be making a lot money since he began requesting donations from his viewers. In just two weeks he has raked in about $10,000 in donations from his fans. (Or at least that's what the total comes to if you count up all the ducks on his site that represent different sponsorship levels).

Does anyone know the legality of soliciting in this way if you are not set-up as a legal charity? If he's not setup as a charity, then I imagine my donation wouldn't be tax-deductible.

Heather, The way Web browsers work is they download everything on the page, including autoplay video, as soon as possible. So just because the whole video was downloaded does not mean it was watched.

And this is Ze's point -- if autoplay is on, you are much more likely to get false download hits. Ze may not know that all his downloads are watched, but he does know they were proactively requested, which cannot be said for Rocketboom.

Then there's the iTunes thing, which is huge. You have absolutely no idea how many are being watched, it could be as few as one in five, or even fewer considering the churn of people signing up and then no longer downloading because they didn't watch for any.

To call Ze "wrong" about his download size makes little sense to me. I wouldn't call him right either, by the way. But the Rocketboom numbers could, in fact, be wildly inflated.

Evidence that Ze is correct is the fact the Rocketboom cannot sell ad space. A few weeks ago Joanne Colan even remarked "Running advertisments on Rocketboom is so rare, that it still deserves mentioning". (Thur. 10-12-06) Hardly something to brag about!

If RB had the numbers it claims they would be able to get sustained sponsors. How much hype do you read about RB's $85,000.00 weekly rate? And Baron's 'revolutionary' concept of discounting ads that he likes. Evidently no one wants to open their chekbooks and pay RB on a regular basis.