Alwin Hoogerdijk, president of Collectorz.com, posted on his blog an interesting video and slides of his presentation on ESWC on how he does A/B Split Testing.
Category: Articles
Free Amazon Cloud Hosting
Amazon has started offering free cloud hosting – obviously as a way to counter Google's free app hosting.
It is interesting, but clearly has a different focus from Google's offer – while you can use Google's indefinitely as long as you stick with the limitations (500 MB of storage and 5 million page views), the new Amazon offering is limited to an year…
MailChimp recently posted an interesting article on how some designs look spammy to people. The research was done using Amazon's Mechanical Turk, and turns out some e-mail campaigns get flagged as spam even though they weren't.
Automatic Object Removal on Video
An interesting article on Popular Science shows a system for automatic object removal on video. It is a bit like Resynthetize on Gimp or Content-aware fill on Photoshop.
The video is also pretty cool, although sometimes parts of the object pop-up in the video.
Interesting case study from WinZip and TrialPay. It details how Winzip used their download page – you know, the page that very often appears after you start a download – to display a TrialPay offer to buy the product.
If you've never heard from them, TrialPay works by having offers from advertisers fund your purchase. For example, you get a subscription to Netflix, a paper, or buy other software. That other company automatically pays for your purchase, so you just spend what you would if you bought the offer directly.
Back to the case study – that download page added tens of thousands a month in new revenue – without affecting the existing conversion rate.
The results seem a bit meaningless to me, though – we have no idea on how much those tens of thousands represent from their regular revenue.
However, the basic idea of displaying the “free” offer after the download seems pretty good, given that many downloaders never get around to installing. Of course, WinZip doesn't have the same problem as micro-ISVs – if the user doesn't know about TrialPay yet, they might worry about a free offer. But given how well know WinZip is, they will probably trust them.
On a last note, while I do not publicize it much, both our main programs are available through TrialPay. You can get STG FolderPrint Plus for free or get STGThumb for free.
Plimus, one of the registration services I use, has a nice report which shows the last few hundred sales on a map. It looks pretty cool, and I thought I'd share mine here:
On the actual report, you can click balloon to see the specific order data.
Please note that this only covers a small fraction of the total sales and countries. The total country list number for my products is 56 right now.
MailChimp posted some interesting numbers on their blog about how their free mailing lists worked out.
The answer is – VERY well – 650% increased profit!
They do add some thoughts on why that might not work for a start-up – the ratio between free and paid users seems to be around 10:1 for the average web service, but that 1 will have to cover all your expenses, and your starting infrastructure might not be up to the task.
Of course, their line of service is particularly good for free samples, since lists tend to increase with time, thus converting those free users to paid users naturally. Plus their free users automatically spread word about their service every time they use it.
A lot of software doesn't have such clear cut transitions, and users can keep using the free version forever.
That said, I do get some free advertising from STGThumb – my photo album generator – from watermarked photos and the HTML generation, however. While I never bothered to quantify this, it does add to a lot of hits over the years.
Question Selection Survey Results
Recently I read an interesting article from GetResponse about Survey Responses.
From the results looks like it's best to have radio boxes based questions, followed by check boxes. Of course, you do limit useful extra information that the user could be bringing on with open questions – after all, they can only choose answers you picked.
Nevertheless, it looks like keeping to mostly limited questions (radio and check boxes), at most 10 questions, with a few optional open questions in the end is the best way to get survey completions.