Day of the week related to default or just noise?

I'm guessing that the day is the day that the loan filled, since this is the only way there could be weekend loans (except for E which only stay on the market for a nanosecond hence no Saturday and Sunday data for E). This indicates the first problem; how long a loan stays on the market strongly correlates with size, so day of the week also has a size component. Es are an interesting data point though, and do indeed suggest that Monday's loans are full of woe, although I suspect there's not really enough of them to be significant.

It is interesting that large D loans (only very big ones make it to Saturday and absolute monsters to Sunday) seem to be safer than smaller ones; this is likely to be because, in riskier bands, the chance of a loan defaulting does not correlate strongly with size, but you've scaled your bars by amount lent. We're talking about a small subset of a small subset of the overall loan book, so again it's best not to read too much into it, especially since the effect is absent for C band.

I'd also expect a monthly variation, for example, with desperate borrowers applying when they realise they aren't going to cover the payroll for the month. Since days of the week move through dates of the month, this will work to blur defaults over different days.

Yes, it isn't explicit in the data (coded as the date of loan acceptance) but I think you are probably right - it would be great to have access to the date loans are put on the platform. So you're entirely correct that loan size would confound this to some extent, limiting the previous plot to loans below 100,000 (data is becoming sparse at this point) you still see some evidence of a relationship (although this is weaker) [ www.dropbox.com/s/y7xsexgieo9r4ce/Screenshot%202017-08-07%2014.56.53%282%29.png?dl=0 ]. There is still the issue that loans deemed to be unappealing take longer to fill (so we might see worse loans filing on Tuesday that were offered on Monday for example) - this would indicate that people are successfully picking loans to some extent!

Loan size is an interesting point - especially as we see more large loans over time [ www.dropbox.com/s/5gycnqmhbrsrnjn/Screenshot%202017-08-07%2015.02.51%282%29.png?dl=0 ] so presumably, a growing number of investors are all investing in the same loans. I haven't implemented any other scaling yet (scaling by no. of loans is the next goal - although I have run into an interface issue). As a crude approximation if we look at the number of loans by status [ www.dropbox.com/s/1t5bflqo5pr5sec/Screenshot%202017-08-07%2015.09.20%282%29.png?dl=0 ] there isn't much evidence of a different number of loans across the working week. If we just look at defaulted loans [ www.dropbox.com/s/ynvvxvsh7e9nde4/Screenshot%202017-08-07%2015.13.19%282%29.png?dl=0 ] we do see some evidence that loans from Tuesday and Thursday are worse. This could again a factor of time to fill. 

Total right about the blurring effect, not really seeing an obvious pattern over the month though (probably blurred by working week) [ www.dropbox.com/s/gjw2zg8g3vcfy5l/Screenshot%202017-08-07%2014.47.03%282%29.png?dl=0 ].

All in all, not a strong signal, could argue that it is evidence for dumping loans that take a while to fill (although you would need some metric normalised for loan size)

I'm using this by the way (http://www.seabbs.co.uk/shiny/fcdashboard ) really interested in any further feature suggestions (normalising by number of loans is a good one)