Why I Don’t Believe in the Hybrid Advice Model


Editor’s note: This is a guest post from the CEO of LendingRobot, Emmanuel Marot. The views expressed in this post are his own and do not necessarily reflect the views of Lend Academy.

Hybrid human/robo advisors are all the rage now, with several key players expected to announce a hybrid solution in the forthcoming months.

On paper, it looks pretty good: let the robot do the simpler stuff, like a modern-portfolio-theory allocation between a few ETFs, and have a human being intervene to provide more sophisticated and personalized advice.

In practice, I think it’s nonsense. If you believe the market is truly efficient, then there’s no point in using an advisor, robot or human. Just invest in a broad market ETF and be done with it (except for tax harvesting).

If you think the market is efficient-ish, then low cost optimization is the solution. Go robot. If you think you need an active manager, you should read the trove of statistical analysis that demonstrate you’re simply paying for someone’s yacht. Indexes beat stockpickers [92% of time]… But even if you’re still in need of human intervention (or if you’re willing to pay a lot of money for smart-sounding but after-facts market analysis) ask yourself the question: how good can investment advice be when they’re cheap? It’s hard enough to find someone actually generating value, finding someone willing to do that for cheap in a corporate environment is like finding a virgin nymphomaniac.

It does make sense for robo-advisors to move to the hybrid model, since it allows them to differentiate and de-commoditize their service, but for their clients, not so much.

Don’t get me wrong, associating humans and machines can be shockingly efficient. The best chess teams are hybrid. PayPal famously used a mix of algorithms and human judgements to identify fraud. But investment advising, at least at the scale sought by robo-advisors doesn’t require truly personalized products. It’s like your banker asking you tons of questions, and ending up advising you to invest in the one fund they’re pushing nowadays, anyhow.

How many factors should be taken into account? Risk tolerance, time horizon, and maybe the accrual / disbursement rhythm. Factoring these in a broader investment algorithm is easy. Not trivial, but easy. And once it’s coded it becomes infinitely scalable, fully traceable. It can also be improved over time without any human retraining.

Machine learning algorithms have become so good in the last 10 years, that any number-crunching and quantitative decisions a smart but junior employee can do, the machine will do better, faster, and cheaper.

Big parts of the finance industry are built on convincing clients they need sophisticated services. Because it’s easier to sell an expensive service if it looks complicated. Neither complexity nor human whims have proven to bring any value.

Or, as a human adviser would say “this alpha-seeking strategy, when conjugated with a beta-stable position, is successful in 5% of the case, with a 5% error margin.”  Because so far, that’s one of the few areas where humans are vastly superior to robots: spouting meaningless justifications.

Peter Renton is the chairman and co-founder of LendIt Fintech, the world’s first and largest digital media and events company focused on fintech.

LendIt Fintech conducts three conferences a year for the leading fintech markets of the USA, Europe, and Latin America. LendIt also provides cutting-edge content all year long via audio, video, and written channels.

Peter has been writing about fintech since 2010 and he is the author and creator of the Fintech One-on-One Podcast, the first and longest-running fintech interview series.

Peter has been interviewed by the Wall Street Journal, Bloomberg, The New York Times, CNBC, CNN, Fortune, NPR, Fox Business News, the Financial Times, and dozens of other publications.

Subscribe
Notify of
6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Todd
Todd
Apr. 19, 2017 4:32 pm

You’re just some sort of cynic who doesn’t want all the millennials to herd up and put all their money into similar trading algorithms (or you haven’t come up with an anti-algorithmic-advice trading algorithm, yet).

Which is it? Talk, heretic — or it’s off to the rack with you! (Where do I send the anti-algo-advice algo-fund check, again?)

🙂

Victor F
Victor F
Apr. 19, 2017 9:38 pm

The truth is the hybrid model is a proven success. Check out Financial Engines, the largest independent RIA and the very first hybrid advisor. The firm manages over $100 Billion in AUM for more than a million customers based on automated investment algorithms but with a human advisor available by phone. The firm is a publicly trade company & actually makes a profit. It’s a wonder that it’s never brought up when people mention Wealthfront, Personal Capital, Betterment, etc.

Emmanuel
Apr. 20, 2017 7:51 pm
Reply to  Victor F

It’s true that Financial Engines is an impressive success, probably due to their focus on retirement plans. Besides, I haven’t seen any data showing their performance beats the market, or pure robots. Many fund managers turn tidy profits for themselves, but, alas, it doesn’t correlate with providing superior returns for their clients.

Victor F
Victor F
Apr. 21, 2017 9:46 am
Reply to  Emmanuel

Financial Engines doesn’t try to beat the market, it aims to achieve the market with passive, low expense & low turnover funds–the foundation of Bill Sharpe’s work. They make no claim of out-performance or alpha. The company’s profit comes from implementing this investment advice across many different 401k plans for many different people via scale & technology. It’s one of the rare examples of a fintech success.

John F
John F
Apr. 20, 2017 1:56 pm

Completely agree with the other commenters. Emmanuel is grasping at straws here trying to defend his robo model. If you read his article, he doesn’t actually state any facts against hybrid.

Emmanuel
Apr. 20, 2017 7:55 pm
Reply to  John F

John, Agreed, I didn’t include any hard data, hence the ‘I believe’ in the title. Besides, nothing would prevent us from running an hybrid model as well. In a way, all models are hybrid, since humans are still needed to guide the robots’ learning process (munging the data, choosing a model, etc…).