Both Ben and I have had a lot of conversations about robo-advisors lately. I wrote an ET In Brief piece a few days ago to discuss some of my own thoughts about Vanguard moving into the periphery of the space. Robos have also come up in discussions with early-stage and VC investors, individual high net worth investors, traditional competitors and staff from the robo-advisors themselves.
The sentiment I’ve gotten from these discussions? Concern.
In some cases, that’s exactly what you’d expect. Traditional wirehouses and RIAs obviously dislike the model, because even if it isn’t directly competing away all that much business, it is starting to influence margins. But among fintech investors and company principals we have spoken to, the level of concern is similar, even if the issues are different. I am hearing about painful customer acquisition costs, rapidly accelerating competition eroding whatever margins there might have been, and a general sense of fear about what a real downturn in equity markets – if such a thing ever happens again – would do to a client base whose stickiness has yet to be proven.
Maybe I’m wrong, and maybe I’m projecting, but I haven’t had a positive conversation about robo-advisors with anyone in months.
So I thought it would be interesting to run the topic through Quid’s natural language processing engine, as Ben and I have been known to do from time to time. It clusters news stories from a wide range of sources around general themes based on various measures of similarity, links them to other nodes, and then qualifies the language to assign sentiment.
Below is the Quid map for Q2 and the beginning of October 2018 for robo-advisors. The boxed categories are mine.
My first observation is that when the financial and general media cover robo-advisors, the stories they tell cluster around one of two distinct Narratives:
- Robo-advisors are an exciting part of a machine-learning and AI-fueled set of innovations, including blockchain applications, that will revolutionize banking (the 3 clusters on the right).
- Robo-advisors are forever changing how financial services companies marry product, technology and advice (the 3 clusters on the left).
The only strong topical link between these two similar but clearly distinct Narratives? Millennials. C’mon.
My second observation, and probably the more important, is that the news treatment of robo-advisors isn’t just positive. It is incandescent. Of all the stories written, Quid’s engine categorizes fewer than 3% as carrying generally negative sentiment. That is very, very unusual for anything relating to financial services. In fact, I’m not really sure that I’ve ever seen it before.
I don’t have a strong take from this, other than to say that topics where different sources have vastly different perspectives tend to be the most interesting. It may also simply be the case that my anecdotal evidence is exactly that – just anecdotal, and not at all representative.
But I don’t think so.
Your earlier piece, “The Power of ‘And’ and the Walmartization of Advice,” provides part of the answer to the question raised in this piece. The disconnect is that those in the trenches - those you have real conversations with - know the dirty little secret about retail investment advice:
“If someone tells you that you need to pay a lot for their advice on these topics [portfolio construction, asset allocation, diversification, etc.] they are misleading you. But here’s the thing: none of those topics are why you hire a financial adviser. You hire a financial adviser to keep you from doing something stupid…”
All your conversations reflect concern because we are at elevated market levels after a looooong run with a big correction/crash/recession/depression out there and who really believes a computer program can replace the psychological support a good FA can provide to keep a client “from doing something stupid[?]” Especially since most clients don’t know that the real reason they hired an FA is specifically for psy-ops.
But on the other hand, and as we know from Ben’s Gell-Mann work, the media knows just enough to be dangerous and probably doesn’t really understand the above core FA-client relationship nuance, so instead, it plays up the wiz-bang new-technology, isn’t-this-exciting, out-with-the-old-in… (you get it) robo-adviser angle.
Also, and this needs a “Gell-Mann” appellation, much of what gets hyped in the media is the result of the need to fill up media websites, blogs, news pages, updates, what have you - not some thoughtful balance of objective importance or relevancy. The maw for this material is relentless, so every somewhat-viable topic gets written about to death.
One plus one equals two: the business people intimately know the challenge (hence, your conversations) + the media has a surface knowledge and an insatiable need for “stories” (hence, your map) = the disconnect you observe.
I think that’s part of the story anyway.
Mark’s third-to-last paragraph raised this question for me: what screens does Quid have for including articles into the data set? Hypothesis is a large volume of SEO-term heavy content (Robo/AI/ML/Blockchain concentration) from the Newsweeks/Forbes(es?) that has little in substance or critical thought, and consequently skews positive in sentiment. Or is such article selection irrelevant because volume and repetition of message is more important in narrative construction than “quality” of individual points?
A couple of thoughts in response to some good questions:
Their brush is a very broad one. This period included about 1,400 articles from publications that range from national publications to trade rags and even press releases (about 31 in the set).
Your last point is mostly right, at least in the way we think about it. Missionary institutions ARE important if you want to predict ex ante whether a narrative will form, but if we are looking at the content on an ex post basis, we can observe the effects of narrative in the similarity and linkages between otherwise distinct clusters. The example Ben and I talked about yesterday was “TMZ and WSJ using the same words to describe Elon Musk.”
For this reason, when a network has strong clusters of linked narratives and language, on balance we usually feel more comfortable about the sentiment being representative as well. When the existence of any governing narrative is less clear, I think that gives us pause about sentiment, too, and requires us to be a bit more granular in examining the quality and reach of underlying sources.