/ Uncategorized

THE TOP CLASS WEDNESDAY UPDATE IS ALGORITHMICALLY SOUND

It’s reassuring, in a way, to know that however bad a week you’ve had, it’s nothing compared to Gavin Williamson’s. Thankfully the lang kittens are still too young to be caught up in the exam fiasco, but lots of you reading this will have offspring who are affected. Full sympathy for you and yours from this particular corner of North Embra.

When you read around the subject a little bit, you quickly get into Algorithm Land (second worst theme park ever after Platform Land, though much more mathematically satisfying) and you’ll read all sorts of people saying that the algorithm used to assign grades failed. But from what I can glean, that isn’t what happened. The algorithm was sound and did exactly what it was expected to do.

What was wrong here was way further back up the chain. The problems came with what the policymakers (whether in Ofqual or central government) asked the algorithm to do. The maths was fine (although ‘fine’ was downgraded in 37% of cases to ‘tolerable’). But what the algorithm was set to achieve wasn’t to give a satisfying grading experience to the tens of thousands of teenagers going through the process. It was set to avoid grade inflation from overly optimistic teachers and to ensure that the societal stratification of results generally seen in previous years was, for the most part, replicated.

Within any large-scale exercise like this, there will always be margin cases at the far end of the statistical playing field; that’s normal. The real test is whether the algorithm achieved what it was trying to do for the majority of results, and it did. The problem, once again, is that what it was asked to do is unpalatable in the cold light of day.

They say there are two things you don’t want to let people see the making of: laws and sausages. I think algorithms can go into that too.

Why is this our subject this week? Well, more and more of what drives our sector’s innovation is and will be driven at least in part by algorithms. I’ve spoken to a bunch of wealth managers in the last few weeks on a research job, and even guys who you would associate more with oak panelling and heritable wristwatches are getting into the game of seeing if the computers can do more of the heavy lifting. And of course the robos at the other end of the wealth spectrum have led the way.

We often ask, when thinking about this rise of the machines, what happens when the model breaks and the algorithm fails – will there be a person there to pick up the pieces and mitigate the potential harms that clients suffer?

I wonder if the better question isn’t who is ensuring that the messy, icky, inefficient, biased bags of bones and meat and other stuff who are giving the instructions in the first place are really doing the right thing and asking the right questions – and if the questions they asked in the first place remain relevant as time goes by? We’ve seen in the last few weeks what happens when the answer to that is ‘no’.

STATISTICALLY PROVABLE LINKS

  • After OpenMoney’s brutal advice gap report the other week, we now have a new report looking in more depth at the supply of advice. It’s from Octopus Investments and is well worth a read. Available here.
  • Cash on platforms is a thing again – any of you holding ver’ haigh balances on platforms might like to read this Dear CEO letter from our friends at Stratford.
  • No #HomeGames this week as we take a barely-earned rest. We’ll be back at 12.30 on Wednesday next week when Steve will be talking to Rohan Sivajoti of Postcard Planning and NextGen about lots of things, but especially what happens when you go behind the headlines about diversity. Register here and view past sessions on our YouTube channel.
  • And a wee reminder that we’re doing an open demo of the new platform pricing engine inside Platform Analyser tomorrow. 11am, 45 mins, all welcome, even if you’re not a subscriber. Register here.
  • Your music choice this week is a bit different. Right now it should be the Edinburgh Festival, but of course it isn’t. So a bunch of venues, including my beloved Leith Theatre, have been lighting up the skies and running socially distanced closed-door performances under the banner of “My Light Shines On”. Here’s one of those performances with the remarkable Aidan O’Rourke and a band of insanely talented musicians. It’s a mix of East and West, and when the Skype pipes start it will give you chills. It’s also 18 minutes long, but I promise it will reward you if you stay with it.

See you next week

Mark

/ Blogs

Impact of poor service

/ White papers

The Impact of Poor Service

We provided the research for a report, in conjunction with Parmenion, which reveals how far short of expectations many adviser platforms are falling. The research found that over the last 12 months, 88% of advisers needed to apologise to at least one of their clients on behalf of a platform, and that poor service delivery from platforms impacts 91% of advisers every day.

Impact of poor service

/ White papers

The Impact of Poor Platform Service

We provided the research for a report, in conjunction with Parmenion, which reveals how far short of expectations many adviser platforms are falling. The research found that over the last 12 months, 88% of advisers needed to apologise to at least one of their clients on behalf of a platform, and that poor service delivery from platforms impacts 91% of advisers every day.

/ White papers

Answering the Call

Service means a lot of things to a lot of different people. It’s so subjective it can be hard to put your finger on. This paper aims to challenge the status quo and inertia that’s built up in the sector for many years.