To be fair “free forever” ceases to apply when the business is about to be taken out back and shot dead, and 1,000 photos still free is 1,000 more photos than a dead business. I’d also imagine probabilistic steps towards break even were considered before undertaking breathing new life into it.
Second, one can’t always know “if the math works” because untested models depend on people. While Yahoo/VZ could have run tests to predict uptake, in a transition like this, there may not be the luxury of that level of market research. Entrepreneurs are regularly called on to make calls absent enough data, and have the character to course correct with candor. I think we’re seeing that character play out here.
Last, much larger revenue Internet services than SmugMug/Flickr are all in on “notoriously expensive” AWS/S3 as well, some with budgets that start with a B. What do they know that readers here might be missing about the TCO of ‘things like static images’?
For one, “notoriously expensive for things like static images” may be misunderstanding the value of a single image to its owner. To them, that image may be “priceless”.
When dealing with a set of images as an emotion archive rather than as a static image CDN, losing any single image should not be an option.
Static “image hosting” services lose images on a regular basis. S3 by and large doesn’t lose objects. If I could only have a single copy of something in a single place, I’d want it on S3. In that “you had one job” category, S3 is a game changer at a remarkably reasonable price.
Regardless of one’s opinion of S3, comparing tables of features may be missing critical differences that don’t manifest in that feature table, differences well worth the expense relative to other costs or opportunity costs.