I came across this interview via my fellow-alumnus, and friend Swaroop CH‘s twitter feed.
For many reasons the story of this startup “PlateIQ” is interesting to me. In silicon valley code-speak of “we are the X of Y”, “PlateIQ is the Mint of restaurant industry”.
They essentially unlocked the data hidden in receipts, invoices of restaurants and made it easy for the businesses to understand their cost of doing business and uncover inefficiencies that affect the bottom line.
This is essentially what my previous company(ENthEnergy) was trying to do with Building energy. The typical energy bill of a large commercial building is obscure for a layperson. Add it to the vagaries of different occupancy patterns, geographical variations, etc., and you essentially have a large bill, that you, as a building owner/occupant have little understanding of.
Anyway getting back to the interview of Ram Jayaraman, CTO of PlateIQ with Garry Tan, these are some of the things that stood out:
- Restaurant business is famously very thin-margin. Moving the margin even a bit can double the profits they take.
- They are trying help restaurants understand the money they are spending. And that information is kind of hidden in their invoices.
- They started in early 2015. It has taken them about 5 years to get to ~10,000 customers.
- Restaurants is a $750B/yr business in US. So, that covers the “if we can sell footwear to X% people in China” angle ;)
- They do claim to use ML and AI to digitize the data out of invoices (PDF, paper scans), but they started out doing this manually. This actually is a good strategy. For one, you will experience the actual pain in automating stuff. Two, if you can find an economic way to do it “manually” you have a long runway before you need to build the buzzword friendly ML/AI solution. ML/AI/Deep-learning is the current VC catnip. So, as a startup, you can use this judiciously to attract attention/funding.
- Not denying they are using Mechanical Turk (the ‘cool’ answer), but “we have teams in three different countries” tells me that they have a back-office in India (very likely), that can already primed to do this kind of digitization by the dint of having medical-transcription experience.
- I liked the story about showing the restaurant owner that they have been overpaying their Vodka supplier. This is a common thing we saw with energy billing too. The person who pays the bill and person that negotiated (assuming they are big enough to do that) the rates are not the same. And over the years the bill diverges from what people think they are supposed to pay. Usually for the worse.
Overall, I think what PlateIQ is doing – liberating data and exposing inefficiencies caused by information asymmetry can be model for building similar “eat the world” software companies in other niches.