Before you feel left out, know that this product will be made available to retail establishments once we finalize our partnerships with retail POS companies (COVID has delayed a number of initiatives on both sides). Consequently, this product is first available to North American restaurant operators, with international restaurant POS partners turning on later this year.
Over the past nine years, our core company (which we almost never talk about) has been partnering with restaurant POS companies (among others) and now sits on top of full POS data from 100,000 restaurants. 30,000 of those will probably die before COVID is over – and that number might be too conservative, frankly – but that’s still the largest repository of restaurant POS data that the market has ever seen.
Additionally, we’ve secured data from the card networks that allow us to ascertain performance for every global merchant that takes a card.
But before we get ahead of ourselves, just exactly how big is 100,000 restaurants worth of data, and why did it take so long to acquire?
For the first question we can look at OpenTable, which counts 50,000 global restaurants as customers. Granted they’re limited to restaurants that take reservations, but even OpenTable does not have full POS data. In fact, when we talked with Jeff Jordan, OpenTable’s once-CEO and now-Andreesen Horowitz partner at a Stanford lecture some years ago, he shared that his biggest regret was ignoring POS integration because the data could have created a lot of value. OpenTable has started integrating to POS systems but it’s not all 50,000 and it’s not 100% of historical and ongoing data.
To bolster this point you could look to today’s largest restaurant POS companies by unit count:
- NCR Aloha, peaked at 70,000 restaurants and losing at least 10% a year since they’ve apparently refused to invest in anything that’s not JetPay and Aloha Essentials
- Micros, peaked at 60,000, and down to 40,000 by our estimates
- Toast, raised over $1B, counts 30,000 restaurants as customers (pre-COVID, and this has likely dropped significantly)
100,000 is comparatively enormous.
As for the second question, logic tells us it shouldn’t take long to get POS data from 100,000 restaurants because the POS companies who contribute data would stand to benefit much in the way of new, defensible product which brings with it new, defensible EBITDA streams. But that assumes a rational market with rational operators which, as we rally on this blog consistently, is not the case at all. Here’s the truth.
Family Business Conundrum
Most POS companies are run as disorganized, dysfunctional family businesses with a hefty splash of incest. This was on full display at PAR / Brink (really driven home by activist investor Adam Wyden’s “Sammon family discount” comment), and keep in mind PAR is one of the industry’s largest POS companies. You can’t imagine how bad it is as you move downmarket, where “executives” literally cannot perform basic arithmetic.
Channel Blockers (And Not The Helpful Kind)
Many of the legacy POS companies are reliant on resellers who won’t work more than a 20-hour week. They interpret their jobs to be lifestyle businesses then become angry when they get their asses handed to them by POS companies willing to work more than 20 hour per week. It’s totally their prerogative to be lazy, but they can’t become indignant about losing to people who are willing to work harder. This becomes problematic when, even if the POS company wants to do something with the POS data, the channel prevents them from successfully distributing the resulting products to the end merchant.
In a grand twist of irony, this makes the merchants of those lazy resellers more prone to leave for a competitive POS. You can’t. make. this. shit. up.
Cloud Is Here… But Still Early
Many POS companies who claim “cloud” aren’t really cloud at all. They might put the right buzzwords on their marketing collateral but they’re not replicating data to the cloud, and merchants need to pay for the privilege of remote access like it’s 1998. So while you might think a legacy company with 10,000+ sites has data above store because that’s how they market things, the reality is far different.
The newer cloud companies do have the data in the cloud, but they take things to another extreme, which gets us to the last point.
Greedy Omnipotence
Too many of the cloud POS companies think they’ll do everything. The same could be said for the larger, legacy POS companies actually, making one wonder if the industry is just replacing one master for another (we’ve heard a number of merchants call Toast the new Aloha; shudder).
Or, if the POS has been acquired by a payments company, you can wholeheartedly bet that the POS company is now doing the bare minimum required to win the processing business, usually under some variant of “free POS“. That almost universally means that they can’t be bothered to invest in creating actual value – look at payment company R&D budgets and tell us that we’re wrong.
Of course the online ordering and delivery aggregators will have the last laugh, but the POS companies are too dumb to think that far out.
Now that we’ve sufficiently covered the disaster that is POS, let’s talk about card data.
Depending on the vertical, 70% of US spend was previously made with cards. Thanks to COVID, that number is closer to 100% than it’s ever been, and we don’t know if it will ever go as low as 70% again. It also means that every merchant that cares to stay solvent is taking cards, and card data now gives a more accurate picture of total sales
Through our partnerships – which took many years, mind you – we have secured data from the credit card networks that give us access to spend and customer data for every merchant that takes a card globally.
Why does that matter? Two-fold.
First, it means that the benchmark is relevant to any merchant, in any vertical, in any geography. Because we’re sticklers for creating value (i.e. a benchmark is more useful when it’s coupled with data science to provide prescriptive recommendations, and one must train the disparities between card data and POS data – which is the universal truth – with measure of error) we’ve limited the the application of the benchmark to North American restaurants to begin. But for those North American restaurants who worry that 100,000 restaurants wouldn’t provide sufficient local market coverage (you guys don’t understand statistical sampling but that’s a whole other conversation), we will have you covered.
Second, and perhaps equally important, is the customer data we’re collecting. We know who your customers are because we have their card data. We understand:
- How often your customers are patronizing your business, and how that’s changing
- Where else your customers are spending money, and how much they spend
- Who your customers are, and how to target them on digital marketing platforms
- When customers return after seeing your marketing
And best of all, we use data science to prescribe actions based upon the changes we’re seeing.
Think you can build this? The card networks do make some flavor of data available through their consulting groups, although it costs ~$100,000 per data set. Since we update data daily, you’d spend $36,500,000 just to even the playing field, and even then you wouldn’t get all of what we get. Then you’d have to hire data scientists and spend months training the data set on top of POS data from 100,000 restaurants, which will take you nine years to acquire.
How does the benchmark work?
By default, we use card spend to add competitors to your compset – short for competitive set – for each location of your business. This is important because the default compset is actually more statistically valid than anything any merchant could choose for themselves. Why? Because we know your actual competitors based upon your customers’ card spend.
Here’s an quick example to show what we mean.
Let’s say you’re a steak restaurant. You might be inclined to think that your competitors are other steak restaurants. What if most of your customers actually went to a nearby seafood restaurant, a high end Asian bistro, and an Italian restaurant? If you pigeon-holed your worldview to steak, you’d be missing the bigger picture.
If you don’t like our default compset you can build your own using a Google map interface: just search for restaurants around you, add the minimum number of required locations, and you’ll have generated your own compset.
Additionally we compare you to your industry segment (think quick service, casual dining, or fine dining) in your local market (generally defined by an 8-minute drive around your location) and all dining establishments in your local market as well. We’re working on a cuisine comparison but restaurant cuisine is… very dirty.
We filmed a video with some sample data just so you can see how this works.
We know it’s the best tool on the market because 1) nobody has this scale of data, and 2) nobody is willing to invest in data science. We don’t care what you hear otherwise – it’s not true.
Want to know what the data science prescriptions on top of this do? Want to use the tool? Want to white label the tool? Get in touch jordan [at] whatsbusy.com.
Add comment