After several missed call dates, Gerard and I finally spoke on Wed. 09 Sept 2009. My apologies for the delay this month. Here are the questions and suggestions I posed to him, and the responses I received.
At the start of the call, Gerard told me that the three "requirements summaries" I had sent him, which were assembled from the suggestions made on three "requirements" threads here at KF, have been passed on to the Product Manager who will be integrating these ideas into possible new designs for lending, checkout, baskets, etc. The engineering team has been working on the underlying Kiva framework code recently, to make it faster and more scalable.
[1] We've noticed recently a very large number of duplicate (and triplicate!) loans -- why is this happening? Almost all of them seem to be a pair of translated and untranslated loans. KFs have suggested that the non-English loan of a pair be the one to be refunded, but have been told Kiva's practice is to refund the LATER loan. This is troubling, as these are usually the TRANSLATED ones, which ought to be retained.
Response: Kiva has identified this as a bug, and the solution is to fix the bug rather than having to chase and refund one of a pair of duplicate loans. He likened the bug which is causing this problem to a "trap door". If a loan arrives in one of the three languages in the current experiment, with some being posted with translations and some with original language text, the trap door is supposed to randomly either let such a loan through or put it into the queue for translation, based on whatever percentages are in force (I believe right now, 1/3 are scheduled to arrive untranslated). The "trap door" is malfunctioning now, such that some loans which are sent to the translation queue are also being immediately uploaded to the site untranslated, causing duplicates. There is a high-priority bug fix being applied to this problem, but for now please keep reporting duplicates as they're found, so one can be refunded. (I suggested strongly that they make it a practice to refund the untranslated loan rather than "the later of the two" which will almost surely be the translated loan, and he said they would look at doing that, although the real fix is to get the "trap door" repaired.)
[2] Is there any possibility of getting a list back from Kiva on what is already on the to-do list, so we don't keep asking for things that are already on the list?
Response: The Liaison has been given information about the relative priorities for engineering staff, and works with Gerard regularly on the list of open items, but at this time passing around a list of to-do items might not be any more productive and would mean more effort spent on less urgent tasks.
[3] Some concern was raised about a Lebanese loan to gun store owner (
http://www.kiva.org/app.php?page=businesses&action=about&id=131201), specifically whether posting this violated Kiva's own stated loan content criteria, which reads:
Microfinance institutions will use their best efforts not to extend credit or loans to borrowers who are in any way engaged in prostitution, the sale or trafficking of adult or child pornography, lethal weapons, narcotics or illicit drugs, or who are otherwise engaged in criminal activities, including activities that could be related to terrorism. Failure to do so will result in termination of the Kiva Contract (hosting agreement).
Why was this loan allowed to be posted?
Response: Kiva's legal team has reviewed this and other similar loans and the policies to be applied in these cases. The reason for the "lethal weapons" clause is to be sure not to support any criminal activities. In this example, the question was to determine if this was a "lethal weapons" loan along those lines, or if it was a hunting store, and in the spirit of the clause above this was deemed a hunting store loan and was not considered in violation of the principles to be applied to posted loans. (It is a legal enterprise in the country where it was posted, and does not violate any UN conventions nor local mores.)
[4] When too many people try to fund the same loan all at once, the text on the basket page is wrong (plus it has a typo, it currenctly says: "We've added loans of your choosing to the basket to bring the total to the maxium amount.") This is improperly stated; the message returned to the lender should read: "In order to prevent this loan from being oversubscribed, we have reduced the loan amount you have placed in your basket to the amount that will fully fund this loan." Is there any chance this could be changed?
Response: This bug fix has been filed (which means it's on its way to being fixed), but Gerard will attach the wording above to the bug fix and perhaps we'll see this exact wording once it's been fixed.
[5] Iraq - Al Alam Centre last raised loans in August 2008. 2 are still outstanding, for repayment in October and November. Will Iraq then be history, or does Kiva hope for further activity in the future?
Response: Kiva is committed to working in Iraq and wants to post loans from there, but microfinance institutions are not flocking to Iraq and are not doing very well there, so there are not many active. With Kiva's 30% limit on funding an individual MFI, it is likely that once Iraq loans reappear, low fundraising limits will be in effect, so these loans will likely be in high demand and quickly funded.
[6] Is there any movement on the order in which loans are shown in My Portfolio? Currently it is a hodgepodge of recent loans, loans coded delinquent, and defaulted loans.
Response: Gerard will try to push on this, he has reported it and hopes for a fix to be pushed out in Sept.
[7] When moving through multiple pages of one's portfolio of loans, or of fundraising loans, the current sort choice is applied on each new page. So rather than seeing one's loans in order, one sees them only in order on each page. (This was mentioned in previous months, with a promise in July that the fixes for these would be released in a couple weeks, but it's still a problem six weeks later.)
Response: For some reason this was pulled out of the earlier release. Gerard will try to push on this and hopes for a fix to be pushed out in Sept.
[8] A number of loans have recently been marked delinquent on the 1st, couldn't that be left until the 15th/16th, if they then really are delinquent? This would save needless concern and worry, especially from newcomers who don't understand what is happening.
Response: I've sent some examples to Gerard also. What is supposed to happen when the field partner reports payment but Kiva has not received it, is that the field partner should be shown as delinquent and not the entrepreneur. Gerard will report this to the engineering team to see if this is a software bug.
[9] There was much discussion about the default Popularity sort, and how this affects loan funding. Several suggestions were offered, which Kiva might consider:
(1) Retain the current formula for Popularity, and also continue the policy of featuring the eight most popular loans on the home page. This will allow for the freshness of the aforementioned ever-changing array of entrepreneurs. However, I also suggest that Kiva change the default sort for the loan list to “Expiring Soon”, in such a way that it does not affect the home page. That is, clicking on the “View all entrepreneurs” link from the home page, or the “Lend” button from any page, or any equivalent link, should take the user to a list of loans sorted by oldest to newest.
(2) Do not assume a new loan is "popular" until after someone starts to fund it, i.e. include new loans with a default "popularity" of 0% instead of 100% .
(3) Cycle the sort types each time someone went to the site. ... change it if you refresh the page (or manually selected a sort type).
Response: These suggestions are all good ones, and will be fed along with the three "requirements documents" we put together as summaries of our requests, as part of Kiva's work towards a refresh of the lending process. One immediate suggestion Gerard will take back to the team is to re-evaluate the "newness weight" in the Popularity formula, since when new loans are dumped onto the site, it immediately pushes down (and often off the front page) loans that actually ARE popular at the time. It may be enough to make the "newness weight" less significant in the calculation.
[10] As one way to avoid possible problems with basket-hoarding, require a person to be logged in prior to placing loans in a basket.
Response: We discussed basket-hoarding problems extensively on this call. The staff is discussing various approaches to basket timers, etc., but one interesting fact I learned is that Gerard now has a tool which would allow him to clear a basket if there is some problem suspected. He is able to review loans individually, by time held, by IP, and if an individual holding large loan amounts is logged in, by user. He tries to review the basketing threads here to watch for problems such as those the LLL team uncovered at month's end and some related to US loans, but would appreciate knowing when we suspect there is a problem with a loan being tied up in baskets inappropriately. If you see such a situation, feel free to let me know by PM and I will pass it along to Gerard who will investigate and clear the loan for other lenders if he deems it to be problematic.
[11] Provide a testing environment for third-party- (as well as in-house-) developers, to avoid interfering with actual fundraising activities. (Large amounts of loans sometimes disappear into baskets and then come back out, and one explanation is that someone is testing an app they're developing.)
Response: There should be NO applications making live transactional tests using Kiva loans. That is, the APIs give developers access to the data only, but testers should not be able to put loans into baskets as part of their testing. If anyone knows this to be different, please let Gerard know. Once developers are given access to transactional data, there will be a sandbox set up for that purpose.
--Diane.
