Quantitative Investing in Shampoo

My partner Andy wrote a blog post on USV.com with this title today. I like the title so much that I want to feature the post on AVC today too.

I have been a skeptic about data-driven venture capital investing for a long time. However, I do think CPG is a sector where this could work very well.


Can a machine help you invest in shampoo? Coffee? Another consumer product?

Last week, the USV portfolio company CircleUp announced the closing and launch of CircleUp Growth Partners  – a $125 million fund that will use a quantitative machine learning approach to invest in early-stage consumer and retail brands.

We believe this is an important evolution towards using data technology to make investment decisions – a theme we at USV have invested in many times ranging from Lending Club to Funding Circle to Numerai. CircleUp Growth Partners is slightly different. The Fund’s thesis is that one can use machine learning to determine early-stage equity decisions in consumer companies. This machine learning platform, Helio, identifies and evaluates companies across billions of data points. The Fund is live right now – Helio recently analyzed 3,400 vitamin and supplement companies and flagged HUM Nutrition as being in the top 3% for brand score. This ultimately led the Fund to make one of its first investments in that company.

The provocative proposition is that a system like this can run these types of analyses at scale and pinpoint brands earlier and with more efficiency than traditional investors. Consumer investors historically have had to spend around 75% of their time sourcing deals manually. Helio is able to automate this entire sourcing process and provide data-driven insights to help companies grow.

Helio has also been applied to two other business lines – credit and marketplace.  CircleUp originally operated solely on a marketplace model but has recently launched a credit arm that provides working capital to consumer companies. These three business units all provide data back to the model, which in turn makes each better in its own domain. This is a data network effect – Helio is continually improving.

The focused industry of consumer goods should lend itself well to this approach; consumer packaged goods all share the same business model, and data proliferates across the industry.

Could data-driven investment models like that of Circle Up be extensible to sectors beyond consumer goods? It will be interesting to see how these approaches might affect capital formation more broadly, as data applications move to designing new financial products and services we have not yet even considered.

#VC & Technology

Comments (Archived):

  1. jason wright

    “This is a data network effect – Helio is continually improving.” – reminiscent of PageRank?

  2. LIAD

    pincer movement for VC from both sides.data-led decisions from the right, token sale funding from the left.seems similar to Capital as a Service from Social Capital – just feed it data and it will make a decision.3,400 vitamin and supplement companies! unbelievable. thank you capitalism. i guess.

    1. fredwilson

      i think applying data driven investing to tech startups will be a lot harder than CPG

      1. Riley Rodgers

        I agree, but there are firms doing this now and we’ll have to wait to see how those investments play out. I wonder if anyone has a list of the funds doing this now and what inputs they are using.

      2. JamesHRH

        I don’t think you have much to worry about from either part of this particular lobster claw.

    2. Richard

      Truth is that there are a handful of vitamin copackers and even a smaller number of vitamin manufacturers. One can start a vitamin label with a a 30 bottle minimum.

    3. Salt Shaker

      I’m somewhat of a skeptic of that number. SKU’s yes, companies no.

  3. JimHirshfield

    Lather, rinse, repeat. That’s basically it, just with data and code.

    1. jason wright

      and head and shoulders above the competition. although what actually stops the competition from competing?

      1. Girish Mehta

        Maybe the competition Dove into the data as well…?

      2. JamesHRH

        Nicely done.

      3. AngelSpan

        Right now there is not enough good data, but the same could be said about vcs the way it took Major League Baseball teams so long to catch up Sabermetrics. Anchoring Theory…..

    2. Richard

      Good opportunity for a Russian Vodka startup

    3. JamesHRH

      Brilliant.

  4. LIAD

    how soon before starting seeing jobs posts looking for data scientists to reverse engineer the algorithms so enterprising entrepreneurs can game the system and score investment.

  5. Chimpwithcans

    Wow – that’s a lot of vitamins….trying to get my head around this – Would Helio update a database or alter its own algorithm when new information is learned?

  6. Tom Labus

    Does it also assess management talent?

    1. fredwilson

      I don’t think so. Likely used performance as a proxy for that

      1. Twain Twain

        There is very serious analysis being done to examine biases in Machine Learning algorithms: https://uploads.disquscdn.c…* https://internetofbusiness….John Giannandrea, Google’s SVP of Search has said this: “It’s important that we be transparent about the training data that we are using, and are looking for hidden biases in it, otherwise we are building biased systems. If someone is trying to sell you a black box system for medical decision support, and you don’t know how it works or what data was used to train it, then I wouldn’t trust it.”* https://www.technologyrevie…Meanwhile, “The best example of [bias in data] is the 2008 crash in which the models were trained on a dataset,” said Shervin Khodabandeh, a partner and managing director of Boston Computing Group (BCG) Los Angeles, a management consulting company. “Everything looked good, but the datasets changed and the models were not able to pick that up, [so] the model collapsed and the financial system collapsed.”https://www.informationweek…

      2. Twain Twain

        We’re also moving towards ethical AI and Joi Ito has written a piece on resisting reductionism (which is over-quantification) and retaining the qualia.Over the next couple of years, game theory mechanics will be challenged because it doesn’t enable like-for-like values comparisons. Price and other quant performance OKRs are not necessarily good proxies.There’s a recent paper from Duke University ‘Moral Decision Making Frameworks for Artificial Intelligence’. The values they want to measure are: “They ask whether an act would be unfair or selfish (because they are not sharing a good with someone who is equally deserving), ungrateful (because it harms someone who benefited them in the past), disloyal (by betraying a friend who has been loyal), untrustworthy (because it breaks a promise), or deserved (because the person won a competion or committed a crime). In these ways, moral reasoners typically look not only to the future but also to the past.”Yet that gets reduced to a non-like-for-like mathematical model where $ are substituted in as proxies for qualitative moral values. https://uploads.disquscdn.c….That’s loss of data signals, the creation of biases towards quantification rather than quanta+qualia for insights and an example of AI not representing or supporting human decision-making.

    2. Twain Twain

      I met Numeral’s CTO in Q1 2016 in SF and asked him how they’re dealing with non-quant qualitative information like management talent, legal environment and consumer preferences which aren’t reducible to neat numbers.I explained to him that when I was 22 like him, I worked in a hedge fund Econostat (now part of IHS Insight) where my Research Director was a Professor of Neural Networks (PhD from Cambridge, Chair of Applied Mathematics at Kings College my alma mater). We had 5+ AI models that processed a lot of quant data from the stock exchanges.In 2011, that Professor of Neural Networks co-founded Commonwealth Capital Management LLP together with Nathaniel Philip Rothschild (@fredwilson:disqus). https://uploads.disquscdn.c…Quant data is useful but we need to be mindful that it’s only 50% of the insights, at most. The other 50% is in the QUALIA and in the language and culture of people that the prob+stats based AI CAN’T (YET) SOLVE:* https://www.fastcompany.com…* https://motherboard.vice.co

      1. Tom Labus

        This is great info, thanks. I thought the interview the founder did with Jason C was great since he asked basic questions

        1. Twain Twain

          Thanks, Tom. Yes, Jason C’s interview with Richard of Numerai was helpful.I particularly liked Jason’s question about whether existing sentiment analysis is “snake oil”.Well, as can be seen from that link to Stanford and Google’s sentiment analyzer:https://uploads.disquscdn.chttps://uploads.disquscdn.c…Sentiment analysis is wide of the mark. The systems automatically assign negative to Jewish people, women, gay people etc.Motherboard article: “Sentiment analysis technology grew out of Stanford’s Natural Language Processing Group, which offers free, open source language processing tools for developers and academics. The technology has been incorporated into a host of machine learning suites, including Microsoft’s Azure and IBM’s Watson.”@fredwilson:disqus — So when MS Taybot is racist and sexist, no one should be shocked. The very foundational NL frameworks from Stanford and other institutions and techcos are badly flawed.Very few people understood why I invented my system and filed a patent. The market is finally waking up to the data biases and inadequacy of NatLang algos.My system doesn’t have the legacy problems that Stanford, Google, FB, MS, IBM Watson et al have.

          1. PhilipSugar

            I love your posts. I would like your viewpoint on using AI as a tool with human interaction. I am a huge believer in that we can use computing power to augment not replace the human brain.A great example are wearable robotics http://www.wearablerobotics.com (full disclosure I am an owner) My believe is the human brain is very hard to replace but augmentation is so powerful.

          2. Twain Twain

            Thanks, Phil, that means a lot to me. I’m a fan of AI that Satya Nadella’s called “assistive technologies.” In my hardware adventures, I’ve mostly focused on enabling visually impaired people to navigate their homes and to be out+about with external-to-body technologies:* https://devpost.com/softwar…* http://digitalcane.comI agree with AI augmenting our intelligence. The main bottlenecks at the moment are to do with whether human intelligence and language understanding is probabilistic and statistical.It’s a non-trivial problem. In 2010/11, Google’s Director of Research took the strategic position that everything is prob+stats:* http://norvig.com/chomsky.htmlIn 2017, there’s clear evidence that prob+stats can’t solve the language understanding problem and that’s vital for dealing with “fake news”, trolls and a bunch of other things:* https://www.fastcompany.com…So if we mean for the AI to support human decision-making, do comprehensive risk management, foster fair+democratic+representative society and enable us to communicate with and understand people of different cultures, that problem needs to be solved.

          3. JamesHRH

            PayPal already figured this out.Max Levchin couldn’t keep the Russian mafia out of their transaction stream and AI couldn’t balance the ‘catch the crooks’ need and the ‘do not piss off custeomers’ need.So, they used AI to flag funky looking transactions and HI to find Boris in the flagged transactions.And then they started Palantir, from that insight.

  7. Adrian Bezler

    Very skeptical towards their approach, but happy to be surprised in a couple of years when we see first results. Why? I am confident that CPG sales will shift to Amazon in the mid-term, hence it will be most important to understand how to game the Amazon ranking algorithm to drive sales.

  8. William Mougayar

    This seems to make sense when there are millions and billions of data points, but maybe less when there aren’t.Regardless, I’m curious as to how the algorithm maintains their data on these winner picks.

  9. awaldstein

    I know of roughly 20 supplement companies that I’m personally tracking and most are either too tiny to have datapoints, too disruptive to actually transfer industry stats to their upside or spinouts.So I wonder though interested in this.

  10. Pointsandfigures

    On the plane from Chi to NYC this morning I continued reading a book on strategic intuition. AI seems to follow the Joumani theory on strategy. But the big breakout companies are created using a more Napoleanic view first identified by von Clausewitz

    1. JLM

      .I am in NYC for a couple of weeks. Ping me if you want to get a drink or coffee.JLMwww.themusingsofthebigredcar

  11. David Semeria

    I think the real challenge here is sourcing the data, in the sense that the earlier the company the less of a trace it leaves…

  12. Salt Shaker

    The CPG industry is full of parity product w/ very limited brand differentiation. Success/failure is often determined by how well a product can deliver on a combination of meaningful attributes/benefits, price, packaging, positioning, distribution and marketing/promo support, all in the context of a brand’s competitive frame. CPG is also very much in transition w/ retail consolidation and increasingly DTC distribution w/ the likes of Amazon and Wal-Mart. I think for sure there are ways to calculate regression and probability curves (hence success) based on the above variables, but the notion that there are billions of meaningful and relevant data points out there is kind of hard to swallow. There are fundamental reasons why historically strong CPG companies like P&G, Lever, J&J, etc., still dominate, though their biggest challenges today are adapting from legacy models.

    1. PhilipSugar

      We should talk sometime, we do collect tons of data on CPG companies (can’t disclose much about the actual data) but we collect a ton.The thing that interests me is that I agree DTC is great for those in high density areas that care about convenience more than price, but it is still way cheaper to ship a truckload of stuff to a warehouse then ship a full truck of different stuff to a store and then have the consumer be the “last mile” deliverer.I think Charlie and Arnold would agree distribution is the tough part.Three stores that interest me are Aldi, Lidl, and Dollar General.They all opened near me here in the flyover states. They are game changers. Small format, low number of SKU’s, cheap prices, smaller packages.Not name brands but I have to think the major CPG companies are white labeling products.Now I think there is going to be a major shift to focusing on price which is what technology does to everybody and everything. Those giant HQ with tons of layers of management are going to be tough to sustain.

      1. Salt Shaker

        Lidl is the new Trader Joe’s. Well established in Europe. Love the model. (TJ’s is so well run, wish it was public). Do you package and sell your data w/ anonymity? Would think that’s a complementary biz for you, assuming it can be packaged and marketed w/ out breaching any NDA’s.

        1. PhilipSugar

          The NDA’s are the tough one (and they should be) I think Aldi is more “Trader Joe” like the Lidl. Lidl is strange in that they sell non food items.

      2. JamesHRH

        I now use a Tic Tac Toe model (or Hashtag framework) when talking to erstwhile entrepreneurs.Paul Lynde – for the young’uns https://www.youtube.com/wat… -is where Distribution sits.

      3. LE

        However the consumer does not generally factor in their cost to get things from the store in terms of time, car expense or transportation and so on.So they will drive across town to save a very small amount of money. Money that the same person would not care about if the same effort went into earning the money hypothetically. Look at how people buy cars. They will scheme to drive down the price spending a ridiculous amount of time. I am talking about people who aren’t making an hourly wage also.One thing that is interesting with the retail model vs. online purchases that I don’t typically see being addressed is impulse purchases. You know in retail there is shelf height and end caps and all sorts of manipulations to get you to buy things that you don’t need. While this is also done online I don’t see it as the same as retail.For example I go into WF to buy a bagel and inevitably end up buying many other things that I see or think that I need. If I got that bagel delivered online I wouldn’t be in the ‘casino’ and fall for the lights and sounds.Ever been to Dim Sum? You just want to get things off the carts that come around. Not the same as ordering off a menu. How many times have you bought things you didn’t need at Lowes or Home Depot because of the packaging or thinking ‘I might need this some day…” while cut wood wafts in the background.Yes there is no question that people buy things online (because of ease of doing so) that they don’t need. But I wonder if it is the same as they do when they actually see and touch in a store.

        1. Salt Shaker

          “People that bought x also bought y.” You purchase a new grill online, they also tout grill covers. Correlation and power of suggestion. Prob more powerful than end caps at retail, cause thought process is linear (more targeted & focused).

          1. LE

            I think this would depend on the business model and merchandise. Although I haven’t been in a dollar store or Walmart in the longest time, I somehow feel seat of the pants that selling that type of merchandise is way more lucrative in a store than if done online. I mean in the sense of people buying what they didn’t come in for or need.Why? Well for one thing I remember as a kid an electrician doing work told me that if someone brings you into their home for an estimate and wants a price on work and then starts to ask for pricing on other work ‘what about this? Can you also do that?’ don’t give it to them. Why? Because they might then decide to do nothing because it will add up to more than they want to spend and they will want to ‘think about it’. So you lock up one order prior to offering pricing that might cause them to mentally go bust. This is in a way equivalent to shopping car abandonment.My point is if you go around a dollar store and put stuff in the basket (where you don’t even know the grand total, right?) and go to the register there is a high amount of probability you will finish the transaction. How many people want to get out of line when they here they have spent $153 when someone is behind them. Or at the Whole Foods want to give the wrapped Chilean Sea bass back to the cashier? But if you put things in a shopping cart you know the price total and might bail out the transaction entirely. It’s so easy to do so.Also suggestions are not the same as impulses. We could both be correct but my feeling is impulses in shopping are a huge contributor to profit nobody goes to a mall for one thing and exits with only one thing. Milk is in the back of the supermarket for this very reason (and is priced as a loss leader).Nothing lastly that Arnold Waldstein (I think) has noted that wine is not sold well online and while there are state laws and reasons why other than impulse I somehow feel that the wine store experience is part of what people are buying that doesn’t work the same when bought online (a guess not my thing or knowledge in any way).

          2. Salt Shaker

            Well, that’s why retailers are so big on loss leaders, to drive in-store traffic. They count on people filling their cart w/ impulse purchases. Not a choice, it’s a necessity. We’re starting to see a lot of big box stores downsizing to maximize rev per sq foot and extract value from RE ownership (e.g., Lord & Taylor flagship in NYC). Not necessarily divesting of space, but subleasing. Overhead is killing them.

          3. sigmaalgebra

            Special case of general result in selling: If you can close the sale, then do so, then, and get that sale CLOSED, then, ASAP, without more discussion about anything else.

        2. PhilipSugar

          I have said this a million times there is supplying, shopping, and buying. Yes, people will go out of their way when supplying to save money, when you are online you are buying and there are few impulse purchases and you see the price and can abandon the cart, when you are in a store you are shopping,I love Dim Sum, and when I go to Home Depot (walking from our office) I know I am going to buy something I didn’t plan on. My two favorites were Green Gobbler (enzyme drain cleaner which is totally safe but dissolves hair), my daughter has hair to her waist, And a battery powered Ryobi sprayer to spray my organic pesticides (I had the father’s day pack of cordless tools already)Never would have bought either.

          1. sigmaalgebra

            If you have a daughter with hair to her waist, then you may have problems no CPG or enzyme could ever solve!E.g., maybe she will decide she wants to dress like Melania! Or win in Olympic figure skating. Or needs a new violin, one she has her eye on, $2 million. Or wants to go Harvard and then get her Ph.D. in biochemistry under E. Lander at MIT and then win a Nobel prize in biology from from curing cancer!Still, no doubt you’d rip out all the plumbing and even take apart half the house just to make her smile!You wanted to be a Daddy and hear her say “Daddy, Daddy”, and you’ve got it!:-)!

  13. JLM

    .Long Term CapitalOld wine, new bottles.Using massive amounts of data to arrive at an “apparent digital winner” has been around for a while. I think LTC had access to a Cray.In the seed world, it seems difficult to come up with significant enough data to make well grounded, deep assessments.JLMwww.themusingsofthebigredcar

    1. JamesHRH

      I had the same thought – by the time you have data to decide Uber is disruptive, the investments has been made.It may work in package goods (which non-jumbo consumer credit is, really).My next thought was ” can you use MI to tell you when to sell (which is the higher value decision in investing, to me). “

    2. AngelSpan

      I would suggest more of a ‘smart beta’ filtering process vs. picking winners. Filter out obviously bad ideas with disciplined, thoughtful filters, then buy the beta of a basket. Then, with the right longitudinal data, you can make more ‘alpha-seekng’ decisions.

    3. sigmaalgebra

      LTC looked at the central limit theorem (CLT), concluded that the market prices were independent increment processes and, from the CLT, concluded a Gaussian distribution, calculated some tail probabilities, concluded that they would have a license to print money with nearly no chance of going bust, promptly did go bust, and had been so highly leveraged that they threatened the world financial system.The CLT is rock solid with several versions, some with fewer assumptions than others. Likely still the most powerful (least powerful assumptions) is the Lindeberg-Feller version, not so easy to state and much more difficult to prove (when I was in grad school, the prof, star student of E. Cinlar long at Princeton, spent the whole hour on the proof, did a grand, tour de force, covered both the front and the side board — sure, I still have my notes).But, still, the CLT is an asymptotic result. The collection of asymptotic results in probability theory is just astounding, beyond belief that any such things could be true, but they are.Still, there’s nothing to say that a “black swan” can’t happen on some fine day. And that’s what happened: Something about defaults in Asian credit from over active US loan making — rickshaw drivers buying Gucci shoes on credit.If are going to put big bets on asymptotic results in probability theory and stochastic processes, then need to understand the math. As above, work your way carefully through Rudin, Royden, Loeve, Breiman, Chung, Neveu, Doob, Karatzas and Shreve, etc., and, then, let’s talk again. Otherwise, keep your money in your pocket or, say, in an index fund.

      1. Adam Sher

        My take on LTCM was that they failed because of management style drift. Their fundamental thesis held but speculation into areas such as foreign credit, which were not part of the original analysis for Black-Scholes. When Genius Failed is one of my favorite books.

        1. sigmaalgebra

          I won’t stand by my guesses about LTCM — right, need the ‘M’. You are likely better informed on LTCM than I am. E.g., while on their very bad day, I did drive by their office, I never read a book on the disaster; it was bad enough without reading a book on it.As much as I don’t like some of the writing of Mr. Black Swan, it’s clear enough that in the stock markets on any fine day anything can happen, including disasters, i.e., can get “black swan”. Maybe on that day, don’t want to be highly leveraged.As soon as we’re talking a stationary, independent increments process for investing, I hold on to my checkbook. Let’s see: We’ve got a stationary, independent increments process. So, we’ve got a martingale. But we believe that it’s bounded. So, we drag out the martingale convergence theorem (stronger than the strong law of large numbers, that is, with a super short proof of the SLLN) and conclude that the market converges almost surely to some random variable. Tilt. Bummer. We don’t believe that. So, we don’t believe stationary, independent increments.E.g., with stationary, independent increments, the thing stands to grow like Brownian motion, and from the law of the iterated logarithm we know a lot about that growth. Ballpark, eyeball, we don’t see growth like Brownian motion.Fun stuff, but my computer’s acting up, and I have to get back to it!

  14. Thomas Luk

    I would be genuinely surprised if that approach alone works well. In so many strategic challenges I have observed that it never has been about data to make the right decision. Typically the guys with tons and tons of data can be easily misled by it.

  15. sigmaalgebra

    They are trying to do statistics that will yield valuable results.From history, there are many successful examples.But from history, much more often statistics yields nearly no valuable results or no significant results at all.There are two main challenges:(1) Statistical Methods. There are a lot of statistical methods. Too often, for a particular problem, selecting or inventing a sufficiently powerful method is too difficult.(2) Data. Commonly there is nowhere nearly enough good, relevant data.For investing, a lot has been tried in statistics. One data point stands out: There is only one James Simons, a darned good mathematician, good at both pure and applied math.E.g., the H. Markowitz work on minimizing a convex, quadratic function subject to linear constraints is nice, got a Nobel prize, for some cases but generally needs better data on the covariances than is readily available.The W. Sharpe capital asset pricing model (CAPM) also won a Nobel prize and would be terrific if, again, we had the covariances.The idea of covered call writing from E. Thorpe, F. Black, M. Scholes, the S. Kakutani observation that first exit times of Brownian motion can be used to solve the Dirichlet problem, etc. was nice before enough people removed the statistical arbitrage opportunities from the simpler cases.As usual, a crucial step for success is the first one, good problem selection. That someone is using “big data”, AI, ML alone are useless and meaningless. Instead need much better evidence that the techniques are powerful enough and the data good enough to yield valuable results.For the techniques, really, need a lot of luck, a lot of rare intuition, some great luck from some rare experience, and/or rock solid, relevant, fully carefully done theorems and proofs, and in the business world all of these are less common than hen’s teeth.Next, more specifically, for such theorems and proofs, need those in probability theory and stochastic processes. Okay. For this little poker game, the ante is J. Neveu, Mathematical Foundations of the Calculus of Probability, from, say, E. Cinlar at Princeton. For more, sure, Karatzas (Columbia) and Shreve (CMU). And there’s much more, e.g., consider E. Dynkin, student of Kolmogorov and Gel’fand and long at Cornell. Now the world of business is down to a meeting in a Boeing 737 restroom. That material has strong pure math prerequisites; those are readily available in essentially all the high end US research universities. But the additional work, Neveu, etc., are not popular in US universities and are much more popular in France, Russia, and Japan.There’s a long history of people getting intuitive ideas and believing in stuff. We can look back and see stuff, just stuff. E.g., leach bleeding, potions from boiled rat tails, poking with needles, just small examples from the horror stories of the long history of bad medicine. But, wait, there was more: The Mayans killed people to pour their blood on a rock to keep the sun moving across the sky. But, wait, there’s more in our time — Saint Laureate Al Guru and his totally wack-o, faked-up, couldn’t read a simple graph, next to zero for solid evidence, quasi-religious, guilt-trip, flim-flam fraud scam, NYT Goebbels style propaganda, shoot the US economy in the gut, kill lots of people in the third world, total BS about CO2 from human activities significantly warming the planet.As a result of such horror stories, we need good evidence. At times, some of the best evidence is solid, relevant math theorems and proofs.

  16. Robert Metcalf

    As the owner operator of a consumer products retailer who was approached by CircleUp to start a discussion about our growth plans, I’d say that their algorithm is spot-on at identifying great companies (completely unbiased opinion).I was also grateful to hear from them and read here that they have a credit arm, since I have very little interest in taking on equity investors.Considering my personal credit card from Citibank (that I’ve had for 18 years) is still a more useful tool for my business than any lines of credit or credit cards that have been made available by Wells Fargo (our business bank by default), there definitely seems to be significant room for improvement when it comes to financing small companies with strong fundamentals. Lots of room left in the world of finance!

  17. Salt Shaker

    Different biz models, but obv there’s increasingly convergence w/ the WF and Jet acquisitions, as both aggressively pursue DTC revenue. Walmart has over 5K stores in the U.S. They’re in the real estate as much as retailing biz, and they can negotiate favorable financing like no other cause of their enormous cash flow and the power of the brand. (AMZN’s P/E ratio is 288, WMT’s 22, so one is trading more on actual performance, the other on mkt potential.) They’re both machines though, and there’s certainly enough room for both to thrive, though their customer bases will likely have a diff demo skew. I’m a Prime customer and it’s hard to beat what they offer in price and convenience. I have not experienced Walmart first hand. No question WMT is very big into “district building” (in the traditional sense).

  18. Salt Shaker

    LOL. That certainly explains why I’ve never been. No FOMO here.

  19. PhilipSugar

    95% of Americans did last year: http://www.businessinsider….The average person is a woman with family income of $53k http://www.businessinsider….For every dollar Amazon sells Walmart sells nearly four.If you are in retail you can’t ignore. You can not like, but not ignore.

  20. PhilipSugar

    I hope you read the comment below each picture, they are funnier than the pictures.

  21. Salt Shaker

    Yeah, of course I read the captions. Def the funniest part. What started as a small prank, evolved into a biz. The site offers to remove any photo upon request, but I’m surprised no one has sued for using their image w/out permission in the 1st place. (Maybe they have.) Does Walmart offer walk-in legal services, like flu shots? How ironic would that be? Full circle.