On Fri, Aug 24, 2012 at 10:14 AM, Rory O'Farrell <ofarr...@iol.ie> wrote: > On Fri, 24 Aug 2012 09:45:18 -0400 > Rob Weir <robw...@apache.org> wrote: > >> On Fri, Aug 24, 2012 at 8:53 AM, Rory O'Farrell <ofarr...@iol.ie> wrote: >> > >> > Rob >> > >> > Might we have a count of AOO 3.4.1 downloads in first 24 hours? No need >> > for fancy graphics - a list would be OK. >> > >> >> First, you need to know that the SF numbers are daily numbers, and >> based on a UTC day. Since we announced mid-day on the 23rd (UTC) that >> means that the stats for that day include a large portion of the day >> when 3.4.1 had not been announced, and when 3.4.1 was not linked to >> from the website. So I wouldn't read too much into that initial day. >> We'll know better by looking at next week's numbers. >> >> So given those caveats here's what I see, giving a other Thursday's in >> August as a comparison, since we have a strong day-of-week pattern to >> our downloads: >> >> August 2nd: 113469 >> August 9th: 107397 >> August 16th: 109834 >> >> And the day where AOO 3.4.1 was announced mid-day: >> >> August 23rd: 145146 >> >> So a 30% boost, which is rather good. I expect it will be even better >> when we can look at full-day. > > Thanks. This is most useful - we might expect a similar increase in Forum > traffic, although perhaps timeshifted. >
It will be interesting to see how many. I image the support requests vary based on other factors as well. When Windows 8 ships it will be interesting... Another thing we should watch out for is mixing up download counts with user counts. Consider three scenarios, in both cases with exactly 1000 real users of the software: Scenario A (Single release) We release version 1 of software and month downloads are: Month 1: 250 Month 2: 250 Month 3: 250 Month 4; 250 In that case, the number of downloads and users are identical, assuming each user downloads once. This is similar to what we had with AOO 3.4.0 over the past few months. The 12 million downloads probably equates to 12 million real users. Scenario (B) (Two releases) In this case we release version 1 and then follow up with version 2 a couple of months later. Imagine exactly half of version 1 users upgrade to version 2 within two months, then we see: Month 1: 250 downloads (all version 1) Month 2: 250 downloads (all version 1) Month 3: 375 downloads (250 new users + 125 upgrades) Month 4: 375 downloads (250 new users + 125 upgrades) So in this case we have 1250 downloads, though we really only have 1000 users. Scenario C (a new release every month) Same assumptions, only 1000 real users, but 50% upgrade within 2 months Month 1: 250 downloads Month 2: 250 new downloads plus 125 upgrades from month 1's version Month 3: 250 new downloads plus 125 upgrades from month 2's version plus 125 upgrades from month 1's version Month 4: 250 new downloads plus 125 upgrades from month 3's version plus 125 upgrades from month 2's version So in this case, 1000 real users generate 2125 downloads due to the duplicate counting caused by upgrades and a rapid release schedule. See how this works? Now we can account for this in projecting users, since we're specially marking, with a special URL parameter, downloads that come from in-product upgrade notifications. So we can break out what fraction of AOO 3.4.1 downloads are new users versus upgrade users, as well as from what version of OOo or AOO they are upgrading from. Cool, eh? So in the future I should be able to report both download numbers and projected user numbers. But when comparisons are made to other projects, with a different release pace and philosophy, it is important that we keep this distinction in mind. Rapid release schedules lead to download numbers that far exceed the number of actual users. Regards, -Rob > -- > Rory O'Farrell <ofarr...@iol.ie>
