Neil_P2PBlog
P2P Blogger
Use @p2pblog to tag me :-)
Posts: 355 
Likes: 209
|
Post by Neil_P2PBlog on Oct 21, 2016 20:59:47 GMT
This may only be of limited interest to a majority of forumites but I hope that some find it interesting! I was looking for additional ways to objectively assess and compare all the different P2x platforms and came up with one idea a bit different. Google has a complex algorithm to assess the authority of each website for each search term, valuing things like relevant content, a fast loading well designed page, plenty of incoming links from news media and an established presence with lots of users staying on the page. So it can be interesting to see how different platforms are ranking across different categories of searches: where Google's algorithm feel the platform has authority. If a new platform is suddenly shooting up to the top it is also very interesting. I used Google's tool for advertisers to build a list of what people were searching for related to peer-to-peer investments. I categorised these into: 1) P2x Lending 2) P2x Borrowing 3) Crowdfunding 4) General Investing. I then recorded the top 300 results in google.co.uk for the 100 most searched terms. For each of these I gave each platform a a points score on its highest ranking from 100 (first place) to 1 (200-300 place). I published the results here. No surprises that Zopa was the highest ranked but another couple like Lendico and LendingCrowd also did quite well. There is also another page for websites related to P2x (where easteregg 's P2P Money was the highest scoring!) |
|
|
|
Post by easteregg on Oct 22, 2016 10:14:01 GMT
This is a excellent and interesting piece of work. It shows that some of the newer entrants are going to have a harder time making an impact to the sector simply through organic means. People can vote for their favourite P2P blog at buff.ly/2eiFA34 on Bitbond. Wiseclerk's P2P banking has a clear lead at the moment! |
|
ablender
Member of DD Central
Posts: 2,204
Likes: 555
|
Post by ablender on Oct 22, 2016 11:54:52 GMT
This may only be of limited interest to a majority of forumites but I hope that some find it interesting! I was looking for additional ways to objectively assess and compare all the different P2x platforms and came up with one idea a bit different. Google has a complex algorithm to assess the authority of each website for each search term, valuing things like relevant content, a fast loading well designed page, plenty of incoming links from news media and an established presence with lots of users staying on the page. So it can be interesting to see how different platforms are ranking across different categories of searches: where Google's algorithm feel the platform has authority. If a new platform is suddenly shooting up to the top it is also very interesting. I used Google's tool for advertisers to build a list of what people were searching for related to peer-to-peer investments. I categorised these into: 1) P2x Lending 2) P2x Borrowing 3) Crowdfunding 4) General Investing. I then recorded the top 300 results in google.co.uk for the 100 most searched terms. For each of these I gave each platform a a points score on its highest ranking from 100 (first place) to 1 (200-300 place). I published the results here. No surprises that Zopa was the highest ranked but another couple like Lendico and LendingCrowd also did quite well. There is also another page for websites related to P2x (where easteregg 's P2P Money was the highest scoring!) This is really good work. May I point out that the sorting does not produce the required result.
|
|
|
|
Post by ruralres66 on Oct 22, 2016 12:07:57 GMT
Many thanks for this novel and useful analysis. No information is too much information in these volatile and uncertain times.
If nothing else, it confirms the gut feeling approach many of us resort to!
Zopa have just dropped rates again?, in response to BoE drop and other matters.
|
|
Neil_P2PBlog
P2P Blogger
Use @p2pblog to tag me :-)
Posts: 355
Likes: 209
|
Post by Neil_P2PBlog on Oct 22, 2016 12:37:43 GMT
ablender - press it once and it sorts low to high, then again to sort high to low (default behaviour of the table, I need to have a play to see if I can swap that around).
|
|
ablender
Member of DD Central
Posts: 2,204
Likes: 555
|
Post by ablender on Oct 22, 2016 12:43:27 GMT
ablender - press it once and it sorts low to high, then again to sort high to low (default behaviour of the table, I need to have a play to see if I can swap that around). That is what I am doing. It does not result in a correct sorting. See the image below: 
|
|
Neil_P2PBlog
P2P Blogger
Use @p2pblog to tag me :-)
Posts: 355
Likes: 209
|
Post by Neil_P2PBlog on Oct 22, 2016 13:00:35 GMT
ablender - press it once and it sorts low to high, then again to sort high to low (default behaviour of the table, I need to have a play to see if I can swap that around). That is what I am doing. It does not result in a correct sorting. See the image below: Thanks for the screenshot, I thought you meant ordering by the points on the other columns.The first column is just an alphabetical sort (first z->a, then a->z), but as some have www. and others not it's not perfect. If you want to just find one website in particular you can search in the top right.
|
|
ablender
Member of DD Central
Posts: 2,204
Likes: 555
|
Post by ablender on Oct 22, 2016 15:15:45 GMT
Found what I was doing, I was looking at the controls at the start of the column heading. They are at the end.
|
|
registerme
Member of DD Central
Posts: 6,624
Likes: 6,437
|
Post by registerme on Oct 23, 2016 8:12:23 GMT
Thank you Neil_P2PBlog, it was / is a very interesting analysis. One thing though, I suspect it tells as much about Google's algorithms (and possibly advertising spend?) as it does about the p2p platforms in question. |
|
Mike
Member of DD Central
Posts: 651
Likes: 446
|
Post by Mike on Oct 24, 2016 6:59:15 GMT
Am I right in thinking that what has been done is attempt to measure a black-box method-generated score, about which almost zero is quantitatively known, by heuristically using different scalings to generate a final number for each platform through the help of kw planner?
This work, in its current form, is useless.
Why 100 points for first place, 60 for second, etc.? Not satisfactory to pick these scalings apparently out of nowhere.
Why is the 'score multiplier' such a low resolution discrete function of estimated search frequency? Ditto above, and how was this factor chosen relative to the scale used in position point allocation?
Both of these points at least seem to have no scientific motivation - but if one uses alternative allocation of points then the results surely differ (as will the 'theoretical maximum' used to eventually 'scale'). So the author may have used several point allocation methods and decided which was chosen in the report based on an underlying bias for different platforms.
Of course, this is all semantics because, Googles algorithm is not static and a total black box from a quantitative pov - it should not be assumed that there is any link between kw planner and Googles ranking score. In addition, the method used to quantify this unknown Google number as the number shown isn't sufficiently detailed to replicate.
This is an interesting idea, but needs a lot more work.
|
|
ablender
Member of DD Central
Posts: 2,204
Likes: 555
|
Post by ablender on Oct 24, 2016 8:45:16 GMT
Mike , you seem to know what you are talking about. Is there scope for you to work with Neil_P2PBlog in order to improve this tool?
|
|
Mike
Member of DD Central
Posts: 651
Likes: 446
|
Post by Mike on Oct 24, 2016 9:29:32 GMT
Don't be fooled by people who seem to know what they're talking about! I have recently left academia (applied maths) and feel it's important to point out that the methodology used in the report appears to be heuristic and I can't see any rationale behind the way points have been awarded and subsequently multiplied. Since this is the cornerstone of the resulting table then it's tough to imagine in what way it can be described as 'objective' when the calculation seems to be entirely 'subjective' - at least based on what I can see.
I now work closely with adwords, and in my opinion (which is at risk of being as if not more subjective then the points I originally complained about) given the unknown-unknowns nature of the metric which is supposed to be extracted [Google's black box] this kind of analysis is of limited use - no matter how one attempts to combine different observable/measurable metrics - without undertaking a huge data science project.
A less arbitrary (and more insightful?) approach maybe to do some analysis on the text in the posts on this forum and elsewhere on the net, although this [probably machine learning] too has potential to be subject to some similar pitfalls.
|
|
Neil_P2PBlog
P2P Blogger
Use @p2pblog to tag me :-)
Posts: 355
Likes: 209
|
Post by Neil_P2PBlog on Oct 24, 2016 14:34:55 GMT
registerme - I just record the organic search results, but perhaps google gives a little organic boost to its paying customers Mike, very interesting to hear your thoughts on this given your applied academic background: The reason I went for such a discrete function of estimated search frequency (just multiples of 1x, 2x or 3x) was to keep things as straight forward and transparent to understand. I don't intend for this to be a perfect measure, but another tool to pick out strong or fast risers among other indicators. The points for 1st, 2nd place etc were not just picked out of thin air: I considered different things such as studies, my own data for websites I run and also with a view to start recognising new and upcoming sites early on (the 1 point for 200+ place). There was no bias on my part for different platforms, I have no preference or affiliation with any. If you look at some other popular measures online, for example alexa analytics which estimates user visits, they have built an entire business off a number which is just a relative estimate. I'm open to suggestions to tweak these numbers or the process. What I can also do to improve it is to add a drill through on each figure to explain exactly how it was calculated: the exact search placing for each keyword and the multipliers that were applied for when I do the November data.
|
|
invest
P2P Blogger
Alive
Posts: 33
Likes: 17
|
Post by invest on Oct 24, 2016 15:53:50 GMT
p2pblog - Interesting work!
It would be interesting to see the following (or a derivation thereof)
1) The ratio between the size of the platform (loan volume), the age of the platform and the number of mentions. This ratio would give an interesting metric to judge the popularity from another metric.
2) The number of mentions of the platforms on specific forums or specific social media platforms.
3) The effects of affiliate programs, these tend to increase the number of links and have an effect on authority.
Cheers!
|
|
