Input for Kiva staff on the "Popularity algorithm"

[Guest]

.

[alan]

Addendum:

The effort of the past few days indicates clearly that once loans get noticed they get funded. The Popularity sort, particularly in its current form, guarantees that some loans will not be noticed and will thus languish unfunded. FIFO would guarantee that all loans would have an equal opportunity to be noticed, and funded, in turn.

[waningestrogen]

I have no fancy words, not much I'm able to offer in funding, and it's already been better stated.
I just want to reiterate: Old to new (or FIFO) should be the default sort.
We're here to help people and fund loans; it's a crying shame that a loan might expire unfunded because it's been languishing somewhere 20 or more pages back in the "popularity" listings.

[tomviolence]

1.   What are we optimizing for?  Increased funding rate?  Equal loan exposure?  etc...

The goal of the Kiva website should be to allow people to find loans they “connect with” to fund, as well as providing a value neutral platform for all loans to pursued. Let the lender discriminate, not Kiva. If Kiva is doing the value judgments on what should be lent too by making some easily accessible, and others not easily accessible, they are doing a disservice to the MFI’s, and most importantly, the borrowers. While an ideal world would have loans presented in a value neutral way, in the real world there will always be loans on the front page top of the list, and loans on the last page bottom.

Unless –

Make sort a secondary function – How ? If I were in charge, I would think a “cloud” of borrowers on the front page would entice people more than a list. Using “tags” that lenders can attach to loans (from an approved list) encourages exploration as well as involvement and community (tag your fav loan with your team, so others can find it and lend to it.)
Imagine images and words on the front page. “Man with cow” images brings up listings tagged that way, along with “related” tags. Click on the picture “Woman with hat”  – brings up similar loans. The word “art” brings up a list of those loans categorized as being in the arts.  Create a search string of your favorite tags to zero in on a loan that matches your interests. No Women with hats in the Taxi business ? This is where the “may we suggest” shows up other loans. The front page can be a swarm of pictures and text changing all the time to present lots of interesting choices.

Bottom line – getting people involved in exploring will help turn page visits into loans. Tagging will help “experts” or dedicated lenders zero in on loans. Multi-level sort helps people find what they are interested in. Getting loans funded means matching supply with demand, regardless of sort order.

[waningestrogen]

"Multi-level sort helps people find what they are interested in."

This, too!

[JohnR]

1.  What are we optimizing for?  Increased funding rate?  Equal loan exposure?  etc...

Indeed, why are we optimizing?  I don't think we should weight for anything, including optimum speed for funding loans.  Folks here have suggested that we go with FIFO as the default sort because many lenders will loan to whomever's on the first page, and so we clean up the old loans.  But what you're suggesting is that we throw the naive lenders to the wolves on things like currency risk.  Kind of, "let the suckers take them."  I'm sure no one is actually suggesting this, but won't that be the result?  I agree w. everyone who says we should do away w. the current Popularity sort because it distorts and causes some loans to suffer.  But I stop short of going in the opposite direction and trying to push old loans to try to get them all funded.  Personally, I wouldn't mind if none of the currency risk loans got funded.  Or the untranslated ones. 

So I'm in favor of something random, but with the option for an advanced search.  (I like Richard's idea of making that advanced search your default page.)  I might be in favor of tomviolence's cloud and tag idea, but I can't get a good mental picture of it.

John

[tomviolence]

info on tag cloud: http://en.wikipedia.org/wiki/Tag_cloud - being wikipedia - it is a bit dry

here is a random website that uses a tag cloud

http://www.pil-be.com/?page_id=7

Obviously, Kiva would need to redo their webpage, rather than just change sort order. Tags are a lot more web 2.0 than lists.

[RichardF]

As it turns out, you can see a topically provocative example of a tag cloud at Kivapedia!!! 

[RichardF]

Okay, even though I played around with a Kivapedia cloud, and I often seem to find myself getting lost in the clouds, I really can't say I'm much of a of fan using them.  The one's I've seen have just been fancy ways of representing counting something, like Kivapedia page views in the above example.

But...

We all agree it's different strokes for different folks when it comes to what gets picked among the available Kiva listings.  Without worrying about the whys of all that, I'll dredge up an old issue concerning the hows I haven't noticed lately.  I love to think in metaphors, so here goes...

Kiva also sorts stuff coming in like a fractional distiller.  The light, fluffy stuff rises to the top, and the heavy stuff works its way to the bottom.  This is how I think of the "Popularity" sort. It's not that I'm interested in lending to the biggest cloud, it's that I am interested in lending to my own, personal, top of the fractional distillation tower... aka, the My Popularity filter/sort!!! Just throw in my own version of the Popularity valves and knobs to the My Searches feature, and I'll give weights to business profile fields and such according to what I like!!!

[waywardcats]

I would like to say thank you to Premal, Gerard and everyone Kiva for giving Kiva Friends the opportunity to discuss this issue and bring you ideas as to how to make Kiva more successful.  I believe that is ultimately our collective goal.

1.   What are we optimizing for?  Increased funding rate?  Equal loan exposure?  etc...

My opinion is that the home page and the default lend tab should be optimized for new lenders.  It should grab and hold their interest, and help them to find someone they connect with, someone to whom they want to make a loan.  Experienced lenders are much more likely to know what they are looking for and how to find it, but new lenders are not necessarily interested in looking further than the first page of loans.

2.  What are the alternate proposals and how will they help achieve (1)?

Variety is the key, I think, to helping people find a loan that they can connect to. In addition to variety I also think that the first pages a lender sees should be dynamic.  While thee may be disappointment when a lender sees a loan they like go unavailable to them, I also think it creates some excitement at being a part of a vibrant community.

Is "popuarity" the best way to do this?

I'm bringing this forward to remind myself about what makes up the current popularity sort (or at least the most recent description we have of it).

The popularity algorithm is a bit dense, but I can give a general idea of the factors it considers:

• Age of the loan - newer loans get a boost.
• Funding rate over the last hour - more recent $ bumps up the loan.
• Page Views - more views = more popularity
• Lifetime funding rate - not just the last hour, but  overall, more funding increases the popularity score
• Videos - we think video content is quite compelling and give those loans a boost in their popularity score

I agree with Wolfgang that the popularity sort as it stands is somewhat arbitrary and in practice creates popularity rather than reacting to it.  However, I also think that showing loans that have recent activity keeps the page dynamic and helps create excitement. I think that having a popularity sort on the first page also is appealing to new lenders by creating a sense of community, as does seeing new loans as they are added.  However, I think that the sort would more accurately be called something like "new and popular" as long as new loans are included.  

One way around this would be to use the home page to highlight new loans only, and create a truly "most popular" sort for the default lend page.  If that were to be the case, then the popularity sort should include a combination of loans with these criteria only.

• Funding rate over the last hour - more recent $ bumps up the loan.
• Page Views - more views = more popularity
• Lifetime funding rate - not just the last hour, but  overall, more funding increases the popularity score
• Highest percentage of loan funded

Adding highest percentage of loans already funded would serve to highlight loans that are already close to fully subscribed and would hopefully help to push them over the top quicker if they start to languish for any reason.  It would also tend to help larger loans when they would currently move off of the front page.  This I think would enhance the variety of loans on the first page.

My concern with the idea of FIFO or highlighting only the oldest to newest loans is that if the loans are loaded in batches, it could result in a lack of variety which I feel would also limit the excitement for new lenders.  Adding them into a sort which you might call highlighted or featured loans might work well, but I would be wary of allowing them to dominate the list.  Perhaps there could be a way to add them into the mix of featured loans when the red expiring banner appears?

I'm sure that many will not agree with me, but those are my thoughts.

-Kerry-

[cpbailey]

There are two considerations.  One is main page of the Kiva website; the other is the sort which generates the first page of loans.  There is a huge jump in loaning activity once something hits the main page.  This main page should be carefully considered as it is a major tool to get new people to explore further into the depths of the Kiva website.  The main page should generate interest in Kiva, microfinance and the loans.  Thus, I would suggest that the main page is to optimize interest in Kiva.  By providing at least one loan that speaks to as many people as possible, this will increase interest and follow through in actually delving deeper to sign up and then make loans.  This is the first view someone gets of Kiva.  Like the front page of the newspaper, the main page should get people to "buy the paper".

Yesterday, at one point all eight loans on the main page were brand spanking new loans in Bolivia.  800 loans on site, and absolutely no variety.  Not one loan had even $25 loaned on it.  These loans probably were all loaded at one time from the same partner.  Meanwhile, a different Bolivian MFI had almost all 20 of the oldest, ready to expire loans. 

I would like to propose that we consider how libraries deal with their inventory decisions.  Libraries get new inventory all the time.  Interests change over time.  Information gets updated.  New stories get published.  Populations change.  Sometimes books get damaged and must be removed from the system.  Sometimes the material gets obsolete.  Yet the new purchases to maintain the interests of their customers may get to the point of filling the shelves beyond capacity.  It isn't feasible or necessarily desirable to keep building bigger libraries to house all the books that are in good enough shape.  Part of the book management is to figure out which books are the right ones to keep when shelf space is getting crowded.  The time that the book is checked out means that it is earning its shelf space.  If a book is only checked out once or twice a year, it can probably be retired.  It isn't given premium shelf space to try to generate interest! 

For books that are EXTREMELY popular--say Harry Potter when it first comes out, libraries don't even display the books.  If someone wants that book right away, one must reserve it through a more involved process.  It takes a bit of effort, but it prevents being 20th on the wait list!  There is no need to place this book in a really visible shelf because it already has the popularity to take it off the shelf.

In some ways, Kiva is a library.  Just like some books get checked out without any special placement, some loans will fund without any assist from the main page.  These loans may be brand new countries, small loan amounts, or more favored business types.  However, some loans, like some books, will see a huge change in interest by placing them in the premium location.  In December, libraries move those Christmas and Hanukkah stories and craft books to a very visible display area.  This makes it easy for people to find books that are likely to be of interest, and gathers various levels or topics (art, music, history, and stories) in one area.  However, in January a new topic such as indoor games and crafts will appear.  Regular rotation of this key display area will increase interest.  The main page will clearly change much more frequently.

I would like to recommend some changes to the main page loans:

1.  Give loans a chance to be Harry Potter and leave the shelf without taking up prime real estate.  A minimum of 24 hours, as long as there are older loans, should be allowed for those who are willing to wade through 800-1000 loans.  On the 2nd of the month, even some $500 loans without currency risk and a loan term of 12 months or less will still have lots of space for lenders after a 24 hour period.

2.  One loan per country (or perhaps per partner) at any point in time if possible.  Variety, please!  It will encourage a more diverse portfolio for newcomers, which increases the probability that any particular person will find a loan that is intriguing to them.

3.  Try to vary loan term, currency risk, countries, loan amount, time until loan expires and loan type, not all of these at every moment.  Rather encourage variety over time.  This will keep things fresh.

4.  If a loan has matching funds available, rotate these in a bit more often.  This indicates one person/organization who likes this type of loan.  It also gets the matching funds working off excess inventory faster.

I have various ideas on formulas to determine which loans go on the home page, but it is late.  Another day.

Colette