Get Cosmic

The topic of the ‘gig economy’ and worker status is already a big talking point of the presidential campaign, centered on the classification of workers as 1099 contractors at companies like Uber and Postmates . Being a contractor makes workers ineligible for the benefits of full time employees, but is also a much lower cost structure for companies that, in some cases, is essential to make the model possible. (For example, Homejoy recently cited pending employee lawsuits about classification as a contributor to their failure.)

At its core, the issue of worker’s classification boils down to an issue of social justice. It would be unfair to have an underclass stuck in indentured servitude, making a break-even living in a gig job after recurring expenses (like car payments, health insurance, and gas) and with no path of advancement toward a better future. One could imagine a future where millions of workers are stuck in this endless economic limbo of break-even employment, outside the normal protections of full time employees, and the group growing over time as more low-skill jobs are automated away. 

There are some tools emerging to help empower workers. Zen99 and Sherpa are helping gig workers understand and reduce the complexity of income tracking and expenses to better understanding their ‘full picture’ profit. This is very important for gig workers as the responsibility is completely on the worker to understand how much they are really making from their gigs and how to figure out their taxes. Having better tracking and transparency will help workers to make more informed decisions about their options. 

But there are still two big problems missing a good solution: the inability for workers to efficiently maximize their hourly pay, and the lack of a career path for gig workers.

In maximizing hourly pay, there’s an information asymmetry between the companies providing gigs and the workers fulfilling them. The companies that set prices have all the insight into existing demand, and this market-making is happening at the individual service level. Gig workers need a tool agnostic of the fulfillment partner, designed to maximize the hourly rate given a worker's credentials (cleaning certification, programming skills), assets (car), and preferences (willingness to travel, clean, or deliver food). Workers need their own hourly-rate-maximization algorithms to go toe-to-toe with company hourly-rate-minimization algorithms in order to properly align employment incentives in the ecosystem. This would help the fungibility of gig employment cut both ways: while contractors would be less costly, they would also be empowered to be ruthlessly mercenary in maximizing their wages. This is only possible if information is sufficiently low friction and fast to inform decisions at on-demand speed. 

We also need a new way to think about a career path in the gig economy. Without better visibility to the tradeoffs in investing time in new skills, workers are at risk of ignoring opportunities that may better optimize their long term career outcome. A good solution would help workers take an impartial view of the right skills to develop in the short-term (eg, investing in a cleaning certification) and the long-term (eg, learning programming skills to start taking higher paying jobs from Upwork). It would also give visibility to an expected return over time, so workers could make more informed decisions about when to defer short term wages for skills with long term payoff, and make it easy by integrating with skill building resources like Udacity. Providing greater visibility into these options could help make long term upward economic mobility the norm within the gig economy.

With the right tooling we could have our ‘gig economy’ cake and eat it too. And now is a ripe time to start building, because as the presidential campaigns get up to speed these issues will quickly move to the mainstream. Companies like Uber will be under pressure to fix their model or point to solutions, and presidential candidates will be hungry to find a middle ground solution. If a tool could truly balance empowering gig workers without forcing a change to worker status, it would help build a more socially just future -- and maybe find itself at the center of a broad reaching national conversation.

I was thinking about the Amazon Dash product and trying to unpack what it says about the strategy unfolding in relation to fast delivery services like Amazon Fresh, Google Express, and Instacart. 

What’s fascinating about the Amazon Dash product is that it is a test of the limits of convenience around the data entry flow provided by a smartphone vs custom hardware built ‘from the ground up’ around a specific software product. It’s an early data point to the question of whether the future will be a small number of smart thin pieces of glass pushing data to our software, or whether instead we’ll have smart glass for some things, a wand with a big button on it for other things, and other custom hardware for yet other things. 

I can’t imagine we'll have splintered hardware for our different use cases of manual data entry because it feels intractably ugly and inconvenient to the consumer. I’m not convinced that what I gain from entering the data with a wand really ‘beats’ the convenience of entering it with my smartphone (especially if that phone has magic native support for it, e.g. Amazon Fire support, where press-button-and-talk-to-order could be supported without having to navigate to an app). Amazon wants to find a way to wrap themselves around the data entry process because it would be an enviable moat against others in the instant-order-at-home business. Convincing a consumer to order and use a new piece of hardware is expensive, and I don't see the value here. But we’ll find out.

An interesting move for Amazon might be to remove the manual part of the data entry. Maybe Amazon could figure out how much stuff I have and order new stuff for me (with some notification so I can cancel when it’s wrong). This would be possible in an ‘information perfect’ environment, so the technical challenges feel much bigger than asking the customer to do it themselves. You’d need a bunch of ‘internet of things’-esque sensors (on the products, in your fridge, in your cupboard) telling you in real time what you have and what you need, or you need very accurate algorithms that can predict, given what you have, what you are likely to consume and how quickly, along with a full understanding of new products entering the home. Or probably some combination of both. Amazon is interestingly positioned as a distributor of these products that they could start slapping RFID or BLE sensors on products going to homes, assuming they could get the costs low enough. This might sound nuts, but when you factor in things like potential optimizations of inventory stocking in localized warehouses to support greater same-day delivery fill rates and the falling cost of active RFID chips it might start to look less crazy. Then we can finally have that smart fridge from the movies, except the intelligence will be located in software instead of the appliance, and our phone would bridge the gap.

The fun part is that as long as these large-scale technology companies are competing I’m sure we’ll continue to see experimentations with new hardware strategies to capture data for new use cases. So far our phones and tablets and computers have done a good enough job of connecting us to our software, and as sensors get cheaper and more ubiquitous they'll get better before getting worse. But we'll see if someone can find the killer use case to buck the trend.

Instacart recently announced a $44mm fundraising round to help scale grocery delivery. A16z, who led the round, had an interesting podcast about the round and the space. They emphasized the role of cloud-based logistics in helping the company scale (ie, algorithms optimizing a distributed delivery force; this workforce already has cars and smartphones) vs. other models like Peapod or Amazon Fresh that rely on brick-and-mortar distribution systems (ie, algorithms optimizing warehouses, trucks, hired drivers). 

Beyond 'unpacking grocery', as a16z put it, these logistical systems are starting to unpack food as a whole. The current distribution model for food across restaurants and grocery is broken in some fundamental ways, and distributed delivery systems have the potential to fundamentally change the way people purchase and consume food. Broadly, I think these trends will drive us toward a greater proportion of food purchases being for pre-made meals vs. grocery.

There are a bunch of companies focused on how to best get food to people, and they cover a broad variety of use cases. But I see 4 main 'value levers' that are helpful to understand the diversity: speed of getting the food, variety and choice for the consumer, price, and healthfulness. If you could create a meal that's delivered to me almost instantly, gives me a lot of variety meal-to-meal, is price competitive to my other options, and is good for me, you would capture nearly all of my food spending. I'm leaving out things like eating with friends in restaurants, or enjoying a tasting menu, or cooking for pleasure, which collectively make up less than 10% of what I spend on food. 

Companies like Spoon Rocket, Sprig, and Munchery get close to delivering on this proposition using pre-made meals. The key insight is to maximize speed, healthfulness, and reduced price, because the importance of variety only matters from the perspective of an individual consumer. The offering can be standardized across an entire customer base as long as my individual consumer experience is varied and gives me choice. 

The moat around grocery exists because buying groceries for food is by far the cheapest way to eat in a minimally pleasurable and healthy way. But there is an enormous intractable inefficiency in grocery stores: lots and lots of perishable inventory exists to preserve the variety of choice needed to fulfill the centralized distribution model. Some grocery stores challenge this traditional model (like Trader Joe's) by carrying far fewer SKUs and focusing on single 'good' pieces. For example, you may not have 10 choices of raisins in different brands and sizes, but you do have one good bag of raisins at a decent price. As more grocery players more online (like Amazon Fresh) the centralized distribution inefficiencies will improve. But the preservation of vast, splintered choice is the root cause of this disadvantage. 

What's interesting is what would happen if the price wall were suddenly scaled, so pre-made meals became price competitive to grocery shopping. Spoon Rocket is currently around $8 a meal (up from $6). If they were at far larger scale, could that go to $4? If it could, it would start competing not with ordering out from a restaurant, but with grocery stores and Instacart. None of the players are there yet - Spoon Rocket is at ~$8, Munchery is around ~$12, Spring is ~$10 - prices that are basically on par with 'fast casual' restaurants. And when Spoon Rocket increased their prices some customers noticed.

Could pre-made meals ever get cheap enough to scale the price wall? I think they could. As food becomes more standardized in a large area, it would allow for economies of scale that aren't otherwise possible. As technology scales and the distributed delivery and logistics systems grow in sophistication, it will start driving down prices for distribution. And if it drives it low enough, we would see major changes in consumer behavior toward pre-made meals making up a larger part of overall food spending relative to grocery shopping. 

And what would happen if we got there? Well, for starters I'd probably convert my kitchen to something else.

I saw some interesting data on e-sports prize pools, graphed below. Both bars on the right are for Valve's International Dota 2 championship, for last year and this year (2014 has a $9mm prize pool). Because the vast majority is crowd funded through the sale of in-game digital goods, the 2014 prize is still increasing - yesterday alone it increased by almost $100,000. With 42 days until the tournament, the prize could top $10mm. 

This obviously blows every e-sport prize before it out of the water. The 5 person team that wins The International will be splitting almost $5mm. For comparison, the 2014 Masters golf tournament awarded $1.6mm to the first place winner.

Traditionally player payouts in sports mostly come from sponsorships, not prize pools. It will be interesting to see how the sponsorship ecosystem for e-sports evolves. These events will be distributed in new ways: live-streaming on Twitch (recently acquired by Youtube) and in-game viewing within Dota 2 itself. While ads run on Twitch, the ad dollars are a small fraction of the ad dollars spent for televised sports. I could imagine these events making it to television and growing toward a traditional model, but there's an important difference between Valve and other professional sports: Valve makes its money directly from the players of the game by selling them digital goods. This means their primary interest isn't in ad sales, but in creating as many happy Dota 2 players as possible. It would be like if the NBA made most of their money selling basketballs and shoes to amateur players instead of advertising. When they don't need sponsorships as much as they need happy players, crowd funding the prize pool might be scalable and preferable for Valve, players, and fans. 

Services like Uber / Hailo (transportation logistics) and Homejoy / Exec / Handybook (home services) are creating more efficient marketplaces of labor. Their software adds value to both agents performing the service (appointment scheduling, payment processing, marketing) and consumers (consistent quality and pricing; customer service; insurance; easy to order). It's obvious these companies are changing how local services work.

When I get into a cab, I always see a smartphone on the dash with Uber or Hailo installed (often both) next to the radio dispatch. I've asked several drivers about which they prefer, and they always sound agnostic. They don't care if the customer hailed from the street, booked on an app, or called the dispatcher (though some felt customers from the apps were better tippers). 

It must be similar for home cleaners: they have a smartphone with some combination of booking services like Homejoy or Exec, and maybe get some customers from their own local marketing. Like the cab drivers, the goal is to maximize booked time at the highest average price point while avoiding double-booking. But they're agnostic - a customer is a customer. I wonder if it gets tricky to remember which t-shirt logo to wear when going to appointments. 

What these services have in common is the structure of a network of laborers with some thin software wrapped around to manage and book appointments, some local operations to provide value like basic quality screening of agents and customer service for customers. From the consumer perspective, the services are fungible. As long as my ride is minimally safe, my house is cleaned, and I'm  not paying too much, I don't care who is providing the service. 

A key challenge for these companies is marketing -- the first to reach the consumer is the first to get their app on the phone / their cleaning subscription locked into their platform. (Side note: it's no wonder Rocket Internet is launching Helpling, a Homejoy clone, to snatch these customers before Homejoy can make it to Europe. It's their standard MO and is a reflection of the importance of first-mover marketing to the model).

Companies with large marketing channels have an opportunity to take market share here. Imagine if Google integrated booking an appointment and ordering a car into search. I ask Google for the services I need, and it connects me to a consistently priced, high quality agent to perform the service. As a consumer, I don't care what's happening behind the scenes to make the connection. So why pass the interested customer directly to a 3rd party vendor? The answer is the extra value these companies currently provide in hard-to-scale areas like customer service and screening of agents. Scaling local operations across all the cities you want to operate in is hard. Screening lots of candidates is hard. 

But these challenges aren't insurmountable. Agents will continue to become unbundled from these services. For example, agents will have de-facto credentials by virtue of another company's screening processes (if you see a cleaner that worked at Homejoy, you can at least know they passed the Homejoy screening process). I can imagine some simple cross-platform credentialing happening organically when enough of these competing services exist. 

As these thin layers of software continue to wrap around local providers and make the supply of labor more efficient and standardized, it'll become easier to offer 'store brands' of localized services. It could be the same agents and products, similar packaging, less expensive, a better purchasing experience, all with higher margin for the seller. In cases where there is more money from a house cleaning sale than a PPC click-through, it's a tempting value prop. 

So will we see Google opening up cleaning services? I doubt it. Google likes being horizontal (they build platforms, they mostly don't sell widgets), and this would be a 'vertical integration' play. A better fit might be review aggregators like Yelp, Angie's List, or daily deal companies. A more 'out there' possibility would be Apple: they could install a 'booking app' as a default on their phones and blow out the market (ignoring the legal issues a la Microsoft and IE). When Apple has a local provider of the service with preferential margin, they could promote it in the app; when they don't, just fall back to Homejoy or Uber. 

Some friends and I have been spending the past 4 weeks thinking about how restaurant inventory tracking works. As a consequence I've been having lots of conversations with restaurant owners and operators trying to better understand the problem of tracking inventory, shrinkage, and measuring current inventory. The process has been interesting; this is a summary of what I've learned so far.

First, there are a few concepts that helped me understand the problem space:

1. Shrinkage = Initial Inventory - Sold Inventory - Current Inventory

Initial inventory and sold inventory can be tracked through the vendor order and your POS system data, respectively, but your current inventory count is harder to track.

2. "Balance Sheet" Food cost % = (Initial Food Cost - $(Current Food Inventory)) / Sales

I'm calling this "Balance Sheet" food cost % to distinguish from "Cash Flow" food cost %, which would take the net food cost over a period of time and divide it by the net sales over the same period. The former is a 'snapshot' in time, the latter is an average over a longer period.

3. Re-Order Inventory = "Par" Inventory - Current Inventory

"Par" inventory is a set level of inventory to keep on stock at a given time. Typically static, this value can change based on projections of sales (eg, if greater foot traffic is expected because of nicer weather, the restaurant may want to set higher par values for that period). 

Imagine the landscape of restaurants as a line, shown below:

As you move along the line inventory optimization becomes more important and the operational problems get harder. I heard feedback that 3 important events occur as you grow: opening a second location; inventory costs cross a line that tracking shrinkage becomes a burning need; and sales grow enough to justify building a total custom inventory tracking solution. 

Starting out, a single location restaurant can track their inventory mostly by paper / on a computer relatively easily. Calculating 'balance sheet' food cost and placing reorders are the primary concerns, shrinkage is less of a problem. Most shrinkage can be managed through the owner/operator's personal supervision.

When you get the second location, suddenly inventory becomes abstract. You need to rely on someone else to tell you what your current inventory is at the other location. This is required for order placement (need to know what current inventory is to know what to restock against a projection of sales for the next period) and tracking things like shrinkage (less important). iPad with a google doc spreadsheet is a pretty good solution here. Getting your shrinkage is "another data point" that's nice to have, but not yet urgent.

At some point your sales will be large enough that it will make sense to focus on managing shrinkage down. This has now become a burning problem, not a 'nice to have' optimization. As current inventory is tracked and gaps emerge between expected inventory vs. reality, management needs to happen on accurate portioning, theft prevention, safer handling to avoid spillage, etc. More accurate understanding of current inventory is more important, so weighing some inventory vs. counting may be preferred. This person wants something that works "like the minibar in Las Vegas" (close to real time insight into when inventory is consumed and in what amount). 

Eventually your sales are large enough that it makes sense to just build your own proprietary solution to better track and manage inventory for more accurate reordering and shrinkage tracking. If you're a big national smoothie chain, for example, you may have a scale connected to the internet that you wheel out and weigh all your fruit containers, ice cream, etc. each night, and the data from this tracking (along with original order information and POS data) is collected back at HQ. A more middle-ground solution might be paying someone like Crunchtime to build a custom version of their software for you.

What could a good solution look like?

The most opportunity probably exists at the 'burning problem' level. A good solution for this owner would gather the data of current inventory as close to real time as possible, and has access to the vendor data (original inventory) and the POS system (sold inventory). It should be flexible to support weighing or counting where it makes sense -- the more things that can be weighed, the more accurate the output. Ideally it would also help with reorder (simplistically refilling inventory to pre-defined par level; more sophisticated solution may help predict projected sales for the next period and adjust par values). 

To get current inventory levels counted close to real time (without figuring out how to put a scale under everything) you need to build a sufficiently optimized workflow tool that drastically shortens time-to-count while maintaining or increasing accuracy, thus allowing for more frequent and accurate counts. Partender is an example of this workflow optimization applied to bars working well.

For restaurants this solution would be dependent on weight (for accuracy) and fast-search (for speed). 

A hardware strategy could be packaging passive RFID tags into daydot stickers (stickers used to keep track of perishable information) and selling a wifi scale that has an RFID reader. BLE would be another route, if it were less costly, to tag the containers. Once set up, you could imagine the workflow going down to seconds per container weighed (as you place the container on the scale and it registers, it could be as fast as checking out from the grocery store). 

A software solution might be simple streamlined data entry tool, a little more streamlined than the excel-data-entry-on-an-iPad solution. Simple features like working well on an iPhone with one hand, the ability to quickly 'tap' to enter count information, instant search, and creating a to-do workflow might be enough to push time-to-count down enough to justify counts more often. 

We created a 'design draft' of what the software product for this could look like: http://signup.getcosmic.com/ 

Other thoughts

1. It seemed like everyone needed to predict future sales to determine par, but were all using their own systems that varied in sophistication. If you had an algorithm that looked at future events like weather and sporting events (for foot traffic) along with historical sales, I bet you could create a product here.

2. Interacting with large food vendors (like Sysco) is a mess. Lots of PDFs in e-mails, opaque pricing, re-orders are a nightmare. There is definitely an opportunity in fixing this process. 

If we're on a trend of software eating the world, I wonder how technology will shape entrepreneurship in the future. I have an intuition that in the future web applications will become the most common vehicle of entrepreneurship in general. This isn't to say that all new businesses will be technology companies, but they will involve technology and web applications. The driver for this is simple: the internet offers too many advantages for new businesses vs. other modes of distribution. Whereas the path used to be to create the product or service and market it to your local geographic area, the future path will be to create a web app to support your product or service and market it online.

Right now the only barrier for entrepreneurs to do this is the thin layer of technical knowledge needed to spin up a simple web app, and this layer is getting thinner and thinner. Standard web frameworks and API's-as-a-service make it nearly (but not quite) dead simple for a non-technical person to set up a rails app, make it look professional with a CSS frameworks like Bootstrap, collect credit cards with Stripe, manage e-mail collection and management with Mailchimp, and handle deployment with Heroku. 

Two areas where I see opportunity in this area are in removing the last remaining gaps in that (already thin) technology barrier, and in increasing the depth of what you get without technical knowledge. I see IFTTT driving the latter: by helping people create 'recipes' hooking together APIs from around the internet, they help increase the depth of what's possible for non-technical people on the web. They are essentially a programming language for non-programmers. It's a great increase of depth of what you can do as a non-technical person without any programming (though I don't think they seem themselves as targeting non-technical entrepreneurs specifically). 

There are lots of individual companies going for APIs-as-a-service to help with the 'removing technical barrier' problem, but I'm not sure I see anyone going for closing it holistically. Some companies have created closed-source versions (like Etsy or eBay storefronts) or a temporary, raise-initial-cash version (like Kickstarter or CrowdTilt). The only thing I see coming close are MOOCs and resources like rails apps on how to build your own rails-app-with-X-feature. But that's a long way from completing the circuit. I wouldn't be surprised if we see a company work to completely close this circuit in an open source way in the near future.

What would happen if we could close this thin layer completely, and if the depth of what you get without technical knowledge got bigger? I think you could see a fundamental shift in the number of online companies and entrepreneurs. As soon as you create a way for people to create a web app that supports up to the first few hundred (or thousand) customers without any technical knowledge, you could fundamentally lower the barrier of entrepreneurship to include everyone that has a great ideas but doesn't have the technical connections, knowledge, or ability to put it online. It's going to be a small incremental shift once it arrives, but I think it'll be a step function increase once it does.

One picture that shows the importance of product innovation, via asymco.

Looking at the company in 2007, who could have guessed the future revenue impact of the iPhone + iPad....

From the article: "Markets are effective in anticipating the path of a technology as it ascends into the hands of users, especially if it is destined for ubiquity. But markets are completely incapable of anticipating the emergence of new technologies.

In this regard markets are a reflection of common sense rather than uncommon wisdom."

I was watching an Marc Andreessen interview where he says (at the 52 minute mark) that the internet provides much easier access to intelligent peoples' opinions by creating a platform for publication without barriers. He calls out several technology bloggers and Nate Silver as prominent examples. The number of smart people in the world that can express themselves and reach a large audience is increasing and there won't be an arbiter (eg, newspaper) blocking them from an audience.

I think this phenomenon will express itself not just in tech analysis or election prediction, but also in broader cultural areas like art, food, and fashion. In aggregate, greater access to more sophisticated cultural voices will raise the average level of sophistication in aesthetic tastes and preferences in ways not previously possible. It used to be, for example, that to be active and up-to-date in the contemporary art world you needed to be in New York City. Publications like Artforum helped provide a platform for critics and galleries (through ads) to broadcast to a wider group, but it was still relegated to major capital cities. Now websites like Contemporary Art Daily (which I used to work for) flatten this access to shows from all over the world. 

Following the technology sector has similarities to following the world of contemporary art: they both involve understanding a certain culture, the existing players, and the current state of an ongoing conversation. Without access to sophisticated voices, the only way to keep on top of it is something like direct industry experience or connections through a personal network.

Recently the art world has seen a huge boom, reflected in both market prices (interest from rich collectors) and museum attendance (interest from the general public). And museums are not getting this interest just from older, conservative works -- they are seeing huge interest (and investing heavily) in contemporary art. The Economist wrote this up in a recent special report.

There's been a meme for a while that the rising prices in contemporary art reflect the irrational exuberance of a bubble about ready to burst. See Felix Soloman's recent article as an example. And it's true that when you compare the size, prices, and amount of money in the contemporary art world today to the art world 10 years ago, it's vastly larger. 

But I think something different is happening. This wider access to cultural information is driving a fundamental expansion of the size of the contemporary art world, and exponentially increasing the size of the market. It looks like a bubble to those looking from the inside (who are used to numbers of a certain size), but it's actually a step change.

For one data point, look at attendance at Art Basel: Miami. After some quick googling, it looks like the 5 year attendance numbers (after a big dip in '07/'08 after the financial crisis) have been:

2009: ~40,000

2010: 46,000

2011: 50,000

2012: 50,000

2013: 75,000

Does this mean we won't see a contraction in interest in contemporary art (or art prices) in the future? No, it'll be a gradual rise with normal dips and retractions, but I feel strongly that the macro trend is up -- something like the Dow Jones Industrial Average from 1980 - 2000.

The flattening of cultural sophistication will take years, but I think we are already seeing it play out. There's no reason to think it'll be limited to just contemporary art -- for example, it's widely said we're in a "Golden Age" of television -- but I'm closest to art so I'll keep my speculation  to that area. It's always hard to see a sea-change from the inside, but I think we are in one now and will continue to be over the next 10-20 years.