This was not what I expected. I thought it would be about toxic customers. It's actually about a set of very good customers who happen to like products that are doomed to failure. The idea is that some people have a distinct out of the mainstream taste. Kind of like when I tell my daughter I like her outfit so of course she won't wear it again.
It's really about a basic misunderstandingof statistics.
Some people prefer to buy novel products. Most novel products fail to get popular. Therefore, there is a correlation. There was no finding that such customer prefer unpopular new products over popular ones.
I interpret the article as a finding that there exists some "bland/mainstream vs novel/unique" scale that actually is objective - i.e., corresponds to measurable and consistent preferences for groups of customers.
There actually was a finding that the analyzed customers historically preferred novel-flavor products over mainstream ones. This means, that if those customers really like your product, then that's some evidence that it may fall into the same group as the other niche products
With a focus on consumer goods rather than services, toxic customers is not as relevant a concern in the context of the article as it would be say for a consulting business. Likewise repeat customers are more of an unambiguous signal for a service business.
Toxic customers are a fault of a business model--not the customers. eg. some biz models are not robust enough to be "indifferent" to the customer--they have exteranalities that are not compensated for--customer support, say. Commodity products typically lack such externalities, and/or are compensated for by other tactics at the industry-structure level.
1. Are these people just general 'early adopters'? Anyone who's willing to try new stuff is going to use products that ultimately fail in the marketplace. This is the whole 'Crossing the Chasm' story.
2. Are these people merely flitting from product to product? In other words, they drink the new Cola A now but will immediately switch to even newer Cola B tomorrow. Thus, they're not really reliable long-term customers. The article gives the impression that this is not case but it's not that clear to me.
In any case, this comes down to a question of market size. If that niche is big enough, there may still be a profitable business in it. For example, how many people do you think need a service that makes Bingo Cards?
WalMart et al stock bingo cards. Teaching stores have them 10+ SKUs deep. That's one reason I thought BCC would sell - people were empirically happy to pay money for an inferior substitute.
From the article: "The worst of the harbingers? Repeat customers, those with a tendency to purchase a product like Diet Crystal Pepsi not just once, but over and over again. These customers really, really like the products that end up failing. Harbingers with a history of making four or more repeat purchases of a failed product are nearly twice as likely as other customers to buy another product that fails."
So no, it doesn't seem like it's as simple as early adopters or novelty-seekers.
That might explain that they're not really novelty seekers but it doesn't negate the point that they're just early adopters. The article goes to some lengths to make these customers look bad but do they also buy products that later succeed? The article doesn't clearly say.
Moreover, How do you distinguish an early adopter from a harbinger? I'd argue that there's very little in this article (as presented) that helps with that.
The stats (as they are quoted here) are just exactly what you would expect to see if you looked at early adopters.
New products typically fail. If you are a consistent early adopter, you will buy a lot more failures than someone who sticks to tried-and-tested products.
If anything, "twice as many" is lower than I would have thought. In my opinion, this suggests that early adopters are rare - that there isn't a novelty-seeker type of person who goes around trying new products for the sake of it.
What they should be measuring is slightly different than this. If these people buy a new product, does that make the product more likely to fail than it would otherwise? (Given that the product is likely to fail anyway)
As I've understood, this stricter statistic is not seen in the article. Even if we could identify people like that, I would put those people into the category of "temporarily unlucky early adopters" and would not expect them to have predictive power.
With that in mind, the conclusion of the article seems nonsensical.
This is a customer group that would be really good for an online service.
For a physical retail store, items that are well liked by a 0.1% of population are a failure - since they carry much of the same fixed costs as a similar item liked by 40% of population, and extra variety costs them; and the best target audience of a physical retail store is "everyone who lives close enough" or "everyone of socal demographic X who lives close enough".
An online store can serve only that 0.1% audience based on their flavour preference. If there are a million people worldwide who really like Diet Crystal Pepsi? Great, there's a lot of money to be made on that product - even if it's unprofitable to stock it in Walmart and expect that a Diet Crystal Pepsi fan will wander in today.
Yeah, the economics for a SaaS application are in some ways very different than a supermarket. I imagine that many of the products that eventually get cut are still "profitable" in the narrow sense that each unit sold is sold at some margin above cost. But when you take into account the enormous opportunity cost of stocking that item and taking up shelf space for it that then cannot be used for something else that moves at greater volume, it becomes uneconomical.
A SaaS app doesn't have the same sort of physical opportunity cost -- every dollar earned is a good dollar. The only real opportunity cost involved is the creator/employees' time.
I want to take issue with "Every dollar earned is a good dollar" at essay length. It's widely believed to be true by devs starting SaaS companies and sometimes dooms good people/products by e.g. makin them scramble to find 50 $20 a month accounts when they'd sleepwalk to 3 $10k a month ones.
I normally like your comments, but I believe the OP addressed this. They said the opportunity cost is the creator's time, which I believe takes into account your objection.
(Of course, responding to the a standalone phrase, you're absolutely correct)
Providing support has an opportunity cost as does developing features requested by users. If your early users are "harbingers of doom" and you listen to them for feature requests, you're going to have a rough time.
Recently I signed up to test a new SaaS app that I thought would be useful. The marketing language on the site was a bit different from what I found inside the app once I got going. I asked the support team for help and instead of leading me on with promises of adding things in the future, they just told me the app was a bad fit and refunded my money. I appreciated that. I was not going to be a productive user for them. Focus is key.
I think this is one of the reasons apps like Stripe take so long to go international. The user needs are very different and there's a real cost to taking on the responsibility of grappling with them.
I think there is also a cohort of customers - people who are really, really attracted to 'new and weird' who are constantly on the lookout for products to fill their super-specific pain point. When something new comes along, they jump on it and get enthusiastically engaged, to the point of likely trying to alter the product direction/roadmap.
Going away from the supermarket example and onto software - it's easy to become engaged with these types of customers, particularly if they are well funded - and lose a connection with the general market.
That is where the real danger lies. Going back to the supermarket - if these customers could affect the Cola recipe, they'd be saying 'maybe add some salt' to tailor to the unique taste they have been looking for. As they increase their purchases when you add the salt, it's easy to think you're finding the market- when in fact you're moving further away from where everyone else is.
Yes, thank you. There are some insane conclusions drawn in this article that are the result of statistical/logical over-simplification and incorrect intuitive thinking.
This is pretty interesting but may not be so relevant in the context of digital products or doing a startup. Yes, we want our products to be mass market from day 1 but if your low-capital 2-person company can get traction in a 1% niche market then you'll have money to focus, retarget, and grow to 2%, 4%, 8%. Pepsi, on the other hand, needs to pay for inventory, shelf space, marketing, and distribution so their profit margins probably depend on mass market adoption. They're also in a position where it's more profitable for them to just axe less popular products and try again rather than steal shelf-space and mindshare from their big sellers.
Yes, I had a similar issue with this article; sometimes, appealing to a niche market IS a path to success. There are countless successful business that cater to smaller market segments.
The article is about product development in retail. If you're not planning on selling your product via a major retailer then this article doesn't apply to you.
While I agree with your sentiment, I would like the article to drill down on the reported behaviour. What happens to those people who gave good early traction (the 4 quick repeat purchases). ie do they continue to buy (niche) or do they move on to the next bright new-fangled thing. If it's the latter, as a two-person company, the fad may wear off quicker than allowing you to get the refocus, target etc.
Identifying who these buyers are is the key ... and ultimately very difficult to ascertain. Any tips on that would be incredibly valuable.
Well-said. I've heard that one of the main reasons Sequoia invested in Airbnb was because they looked at what they were doing at the time, helping people rent out air mattresses in their apartments, and felt it could be generalized out to renting houses, apartments, and all kinds of spaces.
Even the hardcore Whedon fans didn't like Dollhouse very much. Just a handful of episodes, really.
I haven't watched Agents of SHIELD at all yet, because the early episodes have had an even worse reception on whedonesque.com, though that seems to be improving lately.
"Harbinger of failure" seems to be a misnomer, given that the article does show any way to pragmatically identify these users - the only way seems to be with the benefit of hindsight. A 'harbinger' requires predictive value, not post-mortem value.
Edit: To be a bit clearer, they do mention a sort of weeding out of users, but this requires deciding beforehand what 'mainstream' is and then (effectively) looking for confirmation bias.
This seems, at least, to be aimed primarily at market research professionals at the sort of level who would be able to access things like loyalty program data and so forth to provide an initial pool for focus groups, etc. The advice is predictive in the sense that it says "eliminate people whose purchase patterns have consistently exhibited product loyalty to failed products in the past from the pool", and that the people who insisted that "butterscotch sardine supreme" would be a killer flavour for potato chips in your last market test, losing you millions (since your hundred positive testers were the entire market for the flavour), probably shouldn't be part of your next market test.
Product tests are a lot like private software betas - people who actively participate and provide more than the minimum feedback tend to be called on again and again. The advice here is to stop calling on people who tell you to do stupid things.
Your explanation makes far more sense than the article: when you're looking at your pool of informants, weed out the ones with a track recording of previously liking failed products. Basically you're selecting for selective fitness. The article doesn't put this very clearly at all.
I felt the same way. I basically got that point from the article, but there's a lot of verbiage to weed through. Could have been much more concise and direct.
It's worth skimming the (still incomplete) paper rather than just reading the article. The article misses a key point made in the paper: repeat purchases by 'harbingers' was found to be an informative feature, in predicting whether or not a new product would succeed.
tl;dr but did they actually show that the 'harbingers' were predictive or were the 'harbingers' just correlated with failure? It is really easy to find correlation when you start trawling through a large dataset but which all dissipate to nothing once you start using the correlations to make predictions.
It seems to me yes they are niche customers, but also having ever changing tastes in things. Not like early adopters, but will buy a unique product for short time until they find the next unique product they like or get bored and move on.
More like short attention spans more than anything. I mean, how long can someone who's never had diet crystal Pepsi drink until they get bored with it? Or until their friends stop asking them what it is their drinking and hop on the next "cool" product coming to town.
Sounds to me like they just analyzed a bunch of data and found that there were some people, in that historical data set, who consistently bought products that fail.
For this to be science, instead of just finding random correlations, they'd have to take those people and see if they _continue_ to buy products that fail.
Otherwise, sure, I have no doubt in any big enough data set, there will be some people that happen to have done whatever you want to find. Doesn't mean there's any predictive power in it.
I suppose the basic thesis is: given support by a potential prospect for a new product, check what other products they have lent support to. If those products are still around, great. If not, you might have a problem. Or more succinctly: "You don't want people that buy failed products to buy your product--and if they want to, you should change your product"?
I suppose I could believe this. But there's probably a lot of ways to interpret that data. And it seems to be about CPGs (not that that makes it worth more or less, just that CPGs have different dynamics than a lot of other industries).
Parallels to software industry: fickle users that don't want to give you negative feedback for fear of discouraging you (when in reality, they may be saving you from blowing your life savings on an ill fated concept)? Be careful with interpreting the results of customer development (or the CPG parallel--focus groups) Certain types of early adopters should be avoided (how does this square with Crossing the Chasm concepts?)
Maybe similar to how some people find themselves consistently in bad relationships, some people find themselves buying products that are doomed to fail time and time again? Perhaps they wonder why they can't just find a good product that's willing to stick around for awhile...
The general idea is reasonable ("in product study, make sure your sample is representative") but the recommendations are problematic.
Obviously if you just put a product in a store for a certain period of time, only pay attention to the sales numbers, and then try to make claims about the number of potential buyers in a larger population, you're going to have a bad time. (hello selection bias)
But watching out specifically for people with niche preferences in some area is not at all the answer. Barring some very strong data showing otherwise, there's absolutely no reason to suppose that a person with niche movie tastes also has unorthodox tastes in dish cleaning liquids. Now, of course, if you're about to market a Swiffer-lookalike product, whether a certain study participant likes Swiffer is an interesting variable to record. But then, you probably don't need that study all that much in the first place compared to someone who's about to bring to market a bold new product.
And of course this is just basic statistics -- do a good job randomizing your sample, and pick sample size that's large enough for your desired confidence interval.
Oh ok, there's a linked research paper and it's more nuanced. It might actually have some data to support their claim that if a buyer has kinky tastes in toothbrushes, it also spreads to toilet paper.
I think there's a lesson here that might be applicable to, say, JavaScript frameworks, devops tools, or maybe even programming languages. If you're building a new JavaScript library, and you're getting lots of positive feedback from people who have tried everything except jQuery, you're probably not creating something with mass appeal...
My guess is that the "harbingers of failure" are just anti-popular-item folk. They like an item because other people (read: the huge market a company actually wants) don't like it, or because it's "different." Not because it has actual inherent new value, or actually tastes good. Perfect setup for failure -- tons of people dislike your product, and the only ones who like it are the ones who like it because others don't. Cases in point my guess are Crystal Pepsi (as the article uses), another is Zima[1], a third is Orbitz[2].
"These customers do not seem to be shopping at odd hours, and they are not any more likely to pay full price for the products, both of which might indicate that they are less alert or savvy than other customers."
Did anyone else read that as "too drunk or stoned to know the difference"?
Fairly trivial article which fails to state its primary assumption: it is talking about mass-market FMCG (fast moving consumer goods) which want immediate success. This is why we have tens of varieties of instant coffee but no 'cold concentrate' type liquid coffees.
Claiming that niche consumers are the kiss of death is obviously news to many makers of luxury goods (and companies like Apple who do not cultivate mass markets but end up creating them because the niche users ended up being right).
Let's be clear: this is about bringing mass-market products to market, not niche products like most people on HN are building. This article considers a niche product to be a failure.
> Just as positive feedback from lead users signals that a new product will likely succeed, positive feedback from harbingers signals that a new product is likely to fail.
Straight from the "creepy similarities" department:
