I agree with Kiva's aim of attracting new lenders with variety.
My opinion on the best way to do that*:
1) For the front page, keep a "popular loans" sample. Based solely on what people have been funding recently, not what's new or has a video or anything else. Just
2-5 (oh, I see now it's 8 - that's good) of the last {100 loans/12 hours of loans} someone funded, possibly with some weighting for the loans with multiple recent lenders or a high percent complete. Provide this sorting option on the lend page as well, and for those who like reading the fine print,
say that it's based on recent funding. I'd love to rubberneck and see what's getting a lot of funding in the last hour.
2) For the first lend page, I'm all for the random sample (with no "next page" arrows, just a re-randomize button). I really like the 24 hour waiting period, too. Thanks for the Harry Potter analogy, Colette!
If you want to get fancy, you could refine it by taking the {percentages of regions/gender balance/occupation balance/group size} of the loans currently on the site and make the 20 random loans roughly match those proportions. Trying to make it match more than one at a time would be madness, but picking one of those four factors each time would be very easy.
If you're concerned about the recently funded loans not showing enough variety, apply the proportion refinement to those! (Based on the proportions of the sample you're randomizing [the most honest way], or on the proportions of the whole lot, either way you'll get variety.)
I don't like the default FIFO option because it does not give incoming lenders (including me) a tasty variety to look at, and I agree that it dumps loans that informed lenders have been avoiding on the unsuspecting.
* I'm not actually saying anything new, mostly "hear hear"ing to those above me and being picky about exactly my way, mine mine mine.
Edited to correct the number of front page slots.
Also, I agree with the comment below me Charity made: if you really want to see what's popular, you'll have to give people a random sample to find out, at least for a few months. If you're not comfortable with that being the long-term behavior, call it a live data collection period to refine the popularity algorithm. If you find any real live trends other than "we loan to women a lot," maybe the popularity algorithm can benefit by this.