I do a lot of reading - some medicine, some more esoteric stuff. I try to stay sort of current with several subjects including business.
Digital marketing has been a thing for many years. Big computer companies can easily experiment with any aspect of web design. Tiny tweaks can sometimes make a significant difference to how long people spend on a website, how satisfied they claim to be, sometimes how much they spend. Marketers may exaggerate these effects but they surely exist. Data is usually helpful but also can be misinterpreted.
This is accepted as business dogma. In addition to neuromarketing tools, some businesses uncritically accept the idea they need to collect as much data as possible in their customers which can be analyzed quickly with powerful techniques, including foundation techniques, to glean competitive insights not otherwise available. Netflix is able to make credible recommendations based on what you have enjoyed. Since big companies can offer this, gurus suggest every business must also do this to offer or claim to offer a personalized experience. Coupons and deals, just for you! Not to offer this is to be uncompetitive as the technology becomes cheaper and diffuses to smaller companies.
But sometimes the dogma is questioned. Recently, British Airways passengers were decidedly nonplussed when staff were encouraged to learn about their customers, ask about their families, know their names and interests. Privacy was always valued by individuals, and is becoming more highly valued in general after realizing its worth.
Sure, business analysis can be powerful, but the assumption more is always better is becoming questioned more often. So let me ask:
Do you personally value the recommendations, “picked just for you”, offers, etc. From specific places only or more generally? I tend not to value them much even if they may be quite good.
Marketing is hype and people often have to do the best they can with a mediocre product. Better products can be easier to sell and lack of manufactured hype is sometimes a plus. Do you generally trust web reviews? More than people you know?
Marketers use pop-ups, time limits, artificial counts (just 2 more hotel rooms left!), are you sure? bubbles, etc. because they presumably work. Which of these techniques annoys you the most?
If you have a business or work for a place which has thrown its eggs into the digital marketing basket I would like to hear about your experience. Was it really all that? Is it greatly worthwhile? I can see some businesses are primarily digital and rely on good pictures and procedures (online used cars). This seems less necessary for a beloved local restaurant. Maybe the benefits grow as you enter more markets?
My eyes glaze over anything labeled “personalized just for you” because it’s usually anything but.
The ideal recommendation scheme would be one I don’t notice, but just invisibly figures out what I wanted at that instant. Even highly data-driven schemes fail pretty badly at this, with the most obvious failure being how they often recommend the very thing that I just bought and don’t need another of.
I think it’s inevitable and in fact desirable to have working recommendation systems, but they’re pretty far from achieving this in the best cases, and not even trying elsewhere. Plus there is the tension that advertisers are trying to sell me a specific product, regardless of whether I want or need it. Someone like Amazon has less of this problem, since they just want to sell anything. But makers of a specific product want everyone that is even plausibly interested in their product to see the advertisement.
I think it goes without saying that customers will be turned off by invasive information gathering, like your British Airways example. But there’s a lot that can happen without customers noticing.
The star rating of web reviews is completely gamed and untrustworthy. I do have some trust in individual reviews if they sound like they’re from a human and present a balanced picture. I may be optimistic, here. It could be confirmation bias at work. I notice the obvious astroturfing, but by definition I don’t notice the successful astroturfing.
I agree with pretty much everything @Dr.Strangelove said above.
As for web reviews, I look at the average number of stars, and a product with thousands of reviews may get more of a second look than a competing product with only a hundred. But I always read some of the one- and two-star reviews, to see if they are stupid reviews (“the package was hard to open, the product is hard to assemble”), revenge reviews, or they make genuinely good points. Sometimes I read three-star reviews to see what people think is average. I never read five-star reviews, mostly because they have no useful information. The fact is, if I am looking on Amazon for something (for example) it’s because I want to find a good product to buy. The reviews are the only way to have any idea what’s acceptable or horrible. If there is someone I can ask personally and who knows something about the product, that’s great, but rare.
Marketing gimmicks only annoy me when they get in the way of the process of shopping. I guess pop-ups are the worst at that. The one I see a lot in my hobby (not on eBay, but other retailers) is the popup that gives me a huge 10% discount if they can only have my email address. No thanks, not ever.
The idea an entertainment site can recommend similar popular products is not really exciting. Millions of users and reviews - this does not seem a big ask. When discussing unique insights not available except from data, examples are scarce and tend to recall Target and pre-pregnant patients. Are you aware of exciting and unusual stuff actually gleaned from novel looks at data?
Especially annoying when I can tell by the recommendation exactly which datapoint they were basing their crap on. Which is almost never relevant to what I am looking for at that particular moment.
I’ve very occasionally found stuff on Amazon that I wouldn’t have found otherwise because of the “people who bought this also bought” feature. Of course, that’s a pretty primitive recommendation system, but it does have the advantage of being hard to game.
To me, a useful recommendation system would have to be somewhat… smarter than me. Or at least somehow has more domain knowledge than I have.
Some of my favorite YouTube videos are ones where they demonstrate a tool that I never knew existed. Applied Science and Adam Savage are two good ones that come to mind. Recently I saw a video by Savage about sanding tools. He showed off a particular tool that was exactly what I was looking for, but had no idea actually existed or what I’d even look for. Those are the kinds of recommendations I want, but it seems that we’re a long way from achieving that.
Sometimes I’ll scroll through hundreds or thousands of listings for general search criteria, just so I get more exposure to what’s out there. That’s an inefficient approach, but often the direct route never gets you what you want at all.
This thread, on this board, appears to currently be showing me an ad for a charity 5k in another state, as well as an ad for a degree program at a college in another other state; clicking over to another thread on this board, the first ad I see is for $700,000 custom homes in yet another state; clicking over to another website entirely, the first ad I get is for — uh, a shiatsu foot massager?
For me to discuss whether this could be an intelligent future path forward, I guess my first question is: are they doing it yet? Because, near as I can tell, (a) they’re giving me what I’d expect to be personalized-marketing doses, and (b) they’re instead giving me stuff that, as it happens, is actually of no interest to me.
(Granted, maybe I’m misreading the situation and as of yet I’m only getting stuff that doesn’t involve shooting for customization. But — here, let me look at some other website to see what ad pops up: curtains from Wal-Mart? Yeah, that’s another swing-and-a-miss, either because they’re swinging blind or because they’re trying but failing, I dunno.)
Obviously, Netflix tailors its suggestions, and the home screen grid, to people’s tastes. There’s just too much stuff in their catalog; if they want us to be satisfied with their content, they need to find out what we like, at least broadly, and show that first. We watched a silly Korean sci-fi movie once, and for the ensuing months our home screen had a fair percentage of Korean movies and shows; when “they” noticed we didn’t watch any of those, our profile became less Korea-centric.
We’re a gay couple. So sometimes we will choose a movie that has, say, Ryan Philippe in it, rather than another one with George Burns. There are a couple of LGTBQ-themed movies on my Netflix queue.
Over the past 2 years, I’ve noticed that the thumbnails on our home screen increasingly feature young and attractive guys, often shirtless – even if the guy is just a secondary character. We’ve watched all episodes of Black Mirror, but our thumbnail for the show is always one of 3 variations of the shirtless guy from the VR Street Fighter episode. The Good Place was an excellent show with lots of high-flying concepts and weird comedy and literal unicorns, and Chidi’s a central character, but somehow our thumbnail is usually about that time he became a nihilist and went to the supermarket shirtless. It’s become annoying, I feel like I’m missing out on some interesting stuff because of this.
I used to work in ecommerce. I spent time at two different companies — one a major online services platform you have all heard of and used, and one the digital marketplace for a niche brick-and-mortar chain some of you with a particular interest will have known and used. At both places, we attempted to use known information and data modeling to influence people’s behavior and steer them into increased consumption.
It was … not great. At the bigger place, we were attempting to invent our own wheel, and at the other place, we didn’t have the R&D money to do that so we flailed through a series of vendors who promised to spin our customer database into gold.
It rarely rose more than a notch or two above voodoo. (Which puts it two notches above SEO, but that’s a separate conversation.) Occasionally there would be some obvious direct correlation we could leverage, but they were fleeting and transitory and they weren’t “emergent” in the sense that they arose as a surprising conclusion from the data instead of an old fashioned marketer’s instinct.
This was a lot of years ago, and the sector has evolved and become somewhat more refined. But it taught me to maintain a healthy skepticism about the gap between what advertisers know about me and what they claim to be able to derive from it.
And none of the targeted marketing I’ve been shown in the years since has much changed my mind.
To the op in general, I’m not convinced at all.
To build off the general theme of @Heracles’ netflix experience, it seems to me more that “they” are trying to force me/us into niches, trying to force us to accept what they are telling us is correct even if it is blatantly wrong, so that they can sell us to… someone?
It’s beginning to feel like some sort of digital indentured servitude or something.
I have reached a point in my life and outlook where I resist the digital infomining of my life as much as I can and know how to.
Okay, but: you’ve just now prompted me to see what Netflix recommends as Top Picks For Me, and they’re, um, a documentary about a vegan restaurateur, and one of the Jackass movies, along with Bridgerton, plus a flick I didn’t enjoy ten years ago.
The most generous read I can give this is maybe that they’re trying to sucker me into thinking they don’t know me — the better to sell me on other stuff once my guard is down — because, otherwise, what, they’re just taking titles they have no particular reason to figure I’d like, and slapping a ‘For You’ label on them in hopes of getting me to try them when I otherwise wouldn’t? Or they’re honestly trying to appeal to me with these, chosen from among all others, and they genuinely have no idea how to do it?
You say it’s like digital indentured servitude, and wonder if they’re trying to sell us to someone. I say I’m not seeing any evidence that they’ve noticed who I am — in terms of what, specifically, to market to me — and I wonder if they’re selling my info to other entities who can, in turn, get it wrong as often as, oh, say, everyone else who puts irrelevant online ads in front of me.
Take Netflix. They have lots of users who have seen lots of movies and often written reviews and rated them. Netflix has this data and likely owns this data. Obviously they make enough recommendations it could not be done manually.
If you were to take on the task of recommendations, you would see the genres they most preferred, and which items in that genre were highly rated by enough people. That would give you some recommendations which work. You could add preferences for newer or older items, certain topics and formats.
The next step would be to go deeper. You could see what people who watched certain specific movies liked and rated highly and recommend that. You could give similar scores to every watched movie rated highly enough and the people involved like actors, directors, studio, etc.
The third step would be to integrate what you know about the customer. Before the Internet, I had a used book market research book which described every Canadian community and divided them into about thirty arbitrary categories of people based on age, wealth, status and to a lesser degree things like politics and religion, to come up with labels like “Frugal Family of Faith” or “Urban Conscientious Consumer”. Based on these and other lifestyle variables you could further tailor selections.
The fourth step would be to try to predict moods, what one watched on certain days, political and social trends, etc. At some point, your predictions are mediocre and probably make the selections worse. But this could also be measured.
But they are recommendations. It does not much matter if they are all good, but presumably you can track how often they are used and their ratings and improve them with an iterative approach. Claim success. Let the consumer think this advanced approach yields better suggestions than a simple one. Probably it often does. But why shouldn’t it?
Doing this mathematically means turning reviews into numbers. I presume they give a score to positive and negative words and use that as a good enough approximation. This probably works for moods less well but solidly enough, and certain phrases are probably quite strongly associated with mood.
You just don’t know you want it yet. After you are assimilated, it will be obvious you were mistaken only ordering one item of everything just because it fits your needs.
The problem with data intensive marketing is that it seems to be pretty easy for dishonest people to subvert it.
I’m interested in technology; I read a few technology blogs and watch some technical videos. I get marketing for technology scams.
I’m interested in traditional British cookery. At some point, I think I clicked ‘like’ on a post that was about scones with jam and clotted cream, not realising that post happened to be a piece of bait from some right wing activist page. Now, I am getting ‘recommended for you’ suggestions for nationalist and fascist pages (and for some related reason, conspiracy nuttery and ancient astronauts).
