I promised more today, and so here goes.
But first, I like Dottie's comments about "urgency" as a category.
It seems to me that no matter how we devise an algorithm to reflect popularity, it is rather a blunt instrument. Taking into account my earlier comments about the possibility of sorting by amount funded or percent funded as a measure of popularity, either would become a sort of buggy-looking version of "Amount Left". Amount funded would quickly favour larger loans, and percent funded would start by favouring smaller loans (e.g. a loan of $100 gets a 25% boost for every $25, while a loan of $5000 gets 0.5% for each $25) but in fact percent funded would also ultimately favour larger loans. (A loan for $100 can get no more than 75% without being fully funded, while a loan for $5000 can go up to 99%.) One way to correct for this a bit would be to assign "popularity points" in amounts of, say, 1 point per $25 lent (i.e., checked out) and 10 or possibly 25 points per percentage point funded. Or you might need to refine it with, say, 25 points per percent up to 50%, then 10 points to 75%, 5 up to 90% and 1 beyond that. It would help to let the smaller loans compete with large ones. But you would still have a blunt instrument.
In fact, the problem is the equation of "popular" with "compelling". The desire is to present some "compelling" loans to new lenders, and that seems a very sensible thing to do. But with the current algorithm, it is assumed that "compelling" is the same as "several people have recently lent $25 each to this loan" or "several people have recently clicked on it" (and as Ian points out, if I click and move on without lending, isn't that a sign that it's not compelling?) or "look, it has a video, how cool!" But in truth, "compelling" is subjective, and not easily measured by mere statistical metrics.
Here are a few things I think make a loan compelling:
-interesting loan write up, good story;
-interesting specific details in the story (e.g., single mother, widow, unusual business);
-high-quality, attractive photo (video = waste of bandwidth to me);
-lack of currency risk;
-short repayment term;
-translation of the loan description;
-loan about to expire;
-Women in Hats.
All of these are subjective, and quite arbitrary. They are among my personal loan criteria. I make no apologies for them, nor any claim that they are in any way normative for anyone else. My criteria are my criteria, and yours are yours, and that's fine with me.
So how do we get at "compelling"? Here I think there are a couple of ways forward, none of which are mutually exclusive.
One possibility suggests itself from a Canadian bookseller. Indigo Books / Chapters (two brands for the same company) are run by Heather Riesman. They have both physical stores and an on-line store, rather like Barnes and Noble in the USA. If you walk into their stores, there is a prominent display of "Heather's Picks". You can see a similar display on their
website. So, why not offer "Matt's Picks" on the front page of Kiva? The idea would be that Matt Flannery could personally select loans that are, in his opinion, particularly compelling, noteworthy or worthy of being funded, using whatever criteria he wants or even just his gut. Of course, the task could be partially or completely delegated, perhaps with people encouraged to make suggestions to him. But the idea would be that some kind of short list of, say 20-50 loans could be selected manually, using the human touch, and specifically showcased to attract new lenders. This kind of approach would avoid the obvious anomalies of the most recent loan being assumed to be the most compelling and so on. The idea could be expanded into "staff picks" where Kiva staff would each be able to add a small number (1? 5?) of personal picks per day to the list. In essence, this would be a sort of official version of the KF "sell the kids" thread. There could even be some mechanism, such as contests or random draws, to invite Kiva lenders to become guest staff picks pickers for a day (say, allowing 10 picks by the guest for one 24-hour period). That would help to generate "buzz" and a measure of interactivity.
As to contests, how about one for best entrepreneur photo? The idea would be to invite lenders to submit their nominations for the best photo of the day or of the week. Again, this could generate buzz, and also encourage a measure of competition among the MFIs to get a best photo mention. Photo of the week could be showcased with a link to the MFI, and perhaps lenders nominating could have their names put in a hat to be chosen at random for guest staff picker as above.
With some kind of system for manually showcasing loans, "compelling" loans could be divorced from any of the underlying sorts. I know I said earlier I wouldn't tilt at this windmill, but such an approach could allow for removal of the popularity sort as a default.
If you wanted to go with some kind of automated system for showcasing, then I suggest a totally new approach to defining popularity. Rather than the current combination of lender behaviour, the narrow criterion of having a video, and the dumb luck to be newly posted with no immediate competition, I suggest that there might be some kind of system for determining "Popularity Points" based on any number of factors intrinsic to each loan. Some factors could be based on keyword searches for such known "compelling" factors as single mothers, widows, men with cows, women in hats, education loans, people with disabilities, and so on. Other factors might include personal borrowing history (e.g., "this is x's fifth loan and he has always paid on time"), MFI rating (assuming these are refined so they accurately reflect risk) etc. What I am suggesting is a kind of quasi-artificial intelligence system for rating loans on a point system as to how compelling they are, using much more spphisticated criteria than the current brute-force Popularity sort. Staff picks could be incorporated into such a system by allowing for, say, 100 or 1000 popularity points to be added just because Joe in shipping says he likes this loan among his 5 for the day.
I once had an Ethics professor who said "always be aware of the dark side". So here's the dark side: with a more subjective measure of "compelling", which I believe is more honest and more sophisticated than the current or any measure of Popularity, there is the problem of what to do with the not-so-compelling loans. The simple fact is that, boring or ordinary though they may seem, Bolivian group loans or men-with-cows, or women with grocery stores are no doubt the bread and butter of the MFIs and they represent the efforts of real people on the ground to make a living. And they still need to be funded. There are only so many young, attractive, disabled, widowed, single-mother artisanal beekeepers with fabulous photos around. By all means, they are compelling, and they are worth supporting. But a middle-aged man with 6 mouths to feed and a farm to run needs the funding, too. The more we search for the best measure of "compelling" to draw in new lenders, the more we risk mass expiry of loans that just don't get noticed.
I think mass expiry of loans changes the equation for the MFIs, and does so in a way that may not be fair to them. So we really need to take that into consideration in stepping forward.
To sum up: personally, I think some kind of subjective showcasing of "compelling" loans on the front page, combined with a straight FIFO list as the default would be best. Failing that, a more sophisticated heuristic mechanism, perhaps in combination with "Matt's Picks" and/or "Staff Picks" might work, but great care would have to be taken to observe what happens with respect to expiring loans. Bottom priority would be to tinker with the Popularity algorithm, perhaps by making it truly reflect lender behaviour and nothing else, as described at the top. Again, this could also be implemented alongside "Matt's/Staff Picks".
I hope this is of some interest and even of some help.
