What’s My Data Worth: A History of Restaurant Data Markets (Part 2)

This is part 2 of a 2-part series. Part 1 can be found here.

While GuestMetrics was acquiring data through long distribution cycles of its commoditized reporting products, Restaurant Sciences decided to buy data upfront to reach projectability quickly, figuring they could migrate to offering ‘value-added’ information products once they’d reached critical mass. Bankrolled by Jeff Katz, one of the founding partners of Mercury Payments, Restaurant Sciences paid between $10 and $20 per month per rooftop for restaurant locations that had material adult beverage sales. The theory was that paying for data outright would speed up the data collection, and Restaurant Sciences would sell the data – after cleaning it – for more than they paid, netting themselves the spread. And in 3 years, Restaurant Sciences had indeed built a book of over 28,000 establishments providing full data.

Unlike GuestMetrics, Restaurant Sciences used much more robust modeling techniques to clean the dirty POS data. While some manual cleansing was always needed, a model was being trained so that a high percentage of new entries would be automatically and correctly identified as the right item.

That was all good news for Restaurant Sciences. But they hadn’t planned for the fact that their market – very large enterprise-scale distiller, vintners, and brewers – required multi-year sales cycles to penetrate and close, analogous to the brutally long sales cycles GuestMetrics was dealing with by relying on POS and reseller channels to gain traction.

The entrance of Restaurant Sciences also forced GuestMetrics to start paying for data. Now there were two companies burning large sums of money in a race to bring to life a potentially lucrative on-premise data market. GuestMetrics seemed to get the edge in 2013 when they locked up a data deal with NCR’s Aloha division, giving them access to data from roughly 10,000 additional restaurant locations.

But it wasn’t meant to be.

GuestMetrics still supported an unsustainable cost structure in their cleansing methodologies and both companies were running out of money. In 2014 the companies merged, consolidating data from roughly 25,000 restaurant sites, and creating one broke startup in place of two. Immediately after the merger the combined entity (still named GuestMetrics) stopped paying for data (except for NCR, where their contracts required them to pay $2M annually for several more years). This dropped the combined restaurant count to 15,000 and in late 2015 IRI acquired exclusive rights to the merged data set. As far as we’re aware, IRI has quietly wound down the efforts.

We estimate that the companies burned a combined 20 years and $40M trying to get the data market to work. Their efforts give us several key learnings in regards to the future of the on-premise data market:

1. Paying for data can get you out of the gate fast, but long-term is a fool’s errand
2. Cleansing must be automated
3. Sufficient data must be available, in the right mix, or it’s all for naught

We can further combine these lessons with our own knowledge of data markets to share some more useful information.

First, raw data is the lowest form of data. Companies tend to pay much more for information when it’s immediately useful. Raw data is the last place anyone should focus their efforts. Would you rather get $5 per month for your raw data or let a third party turn that data into a product that earns you $50 per month? It’s a no-brainer.

Second, producing a statistically viable sample size means cooperation is necessary across the industry. Resellers are often asking what “their” data is worth but the reality is that data needs to be aggregated at the ISV-level to achieve scale. The channel is much better suited to be handed products built from the raw data as those products a) make merchants more successful, b) demonstrate higher value for the channel, and c) avoid the accounting nightmare of pro-rata payouts to thousands of resellers. A rising tide lifts all boats, and we can’t forget this in light of the below reality.

Lastly, merchants are still dumb as hell and irrational about their data rights. Companies like Amazon, Opentable, Uber, Grubhub, and others are already consuming merchant transaction data for a wide variety of uses. It’s absurd to think POS companies shouldn’t be given the same data rights especially when some of the aforementioned companies are actively working on business models that will compete with merchants! It’s one of the most asinine things we’ve ever seen and perfectly sums up why on-premise is so far behind every other industry.

We borrowed heavily from Chuck Ellis for this article. Chuck served as CEO of both Gazelle and Restaurant Sciences. He has since founded Foodservice Analytics, a company set on foodservice data cleansing through their taxonomy, attribution, and normalization automation IP. Chuck routinely deals with the three major types of data in this industry: POS transaction data, purchase orders, and menus, which all obviously interact.

So no, your data isn’t worth anything. But it’s also worth billions when productized correctly. Are you able to spot the right partners to get you there?