Robert_Charlton - 9:28 am on Jul 10, 2013 (gmt 0)
Size by itself isn't sufficiently descriptive. I think you've got to look at size in combination with trust and authority, and a number of other factors. There are many tradeoffs to take into account.
If a site is too large, with insufficient trust and authority, then the size is a liability. Pages will lack enough link juice to rank. Making a site too large is analogous to overexpanding a business enterprise and not having the proper customer base and resources to design, run, and maintain it. Too large a site will not only lack link juice... it will likely also suffer in terms of content quality and user experience.
That said, there is the old joke that if you can't make it good, make it big. I think that's less true for websites than it is, say, for painting or architecture.
It may be that many of the larger sites we see ranking are there because they're survivors that have naturally expanded over time... and they often have custom infrastructure that's come from years of development and re-investment. Large sites can also take advantage of economies of scale that many smaller sites don't have, and can offer a richness of user experience which a smaller site can't provide.
The algo has, for at least a dozen years that I can remember, often favored large sites. It's amazing what you can get away with at PageRank 7 (yes, PageRank) that you can't at PageRank 3. Your page templates, eg, can be much less unique. It's as though the algo regards templated pages with more high quality inbound links as more unique than they'd see those same templated pages with lesser inbounds.
But smaller sites... perhaps not burdened with that legacy infrastructure... can move and adapt more quickly, and that, in the right hands, is a distinct advantage. My inclination is to grow with demand and resources... as goodroi puts it, to "naturally grow in size over time".
At the same time, in the current algo climate, I am concerned about not starting too small. I've been assuming, eg, with a current article site I'm developing, that I should push for a certain critical mass before launching. It seems to me that if "engagement" suggests time on site (which in part I think it does), a critical mass of really good pages would be helpful in giving a new site a better start.