Launch HN: Bottomless (YC W19) – Coffee Restocked with a Smart Scale
Bottomless automatically re-stocks coffee using a smart scale. Users leave their coffee on the scale, then we detect the perfect time to trigger re-orders. We ship the scale for free when customers buy their first bag.
We met in college, and bonded over talking about businesses we could build together. You could say we've kept in touch since then: we're now married. Bottomless was born out of our frustration managing our household stock levels. We always seemed to be running out of one thing or another.
When we thought about it, we realized that restocking was a universal problem.
But if this was such a big problem, why was there no great solution? Subscriptions should be a solution, but they don’t work well for items that aren’t used on a set schedule. It seemed that if we could capture data on usage and stock levels in a passive way, we could solve the problem. Thus, Bottomless, the concept, was born.
The market for stuff people repeatedly buy is enormous. (We'll leave an exact estimate up to the reader's imagination.) We decided that to start we'd establish a beachhead with a single market. We landed on selling premium coffee because it's cheap to ship and has good margins. It also is much better shipped straight from the roaster than bought at the grocery store.
In the beginning, we built the simplest thing possible to test if the concept would work. We hacked together a scale prototype, made five of them and got them into the hands of friends. We bought coffee from roaster websites with our customers’ addresses to bootstrap supply.
The goal was to test if people would leave their coffee on a scale, and if we could reorder at the right time. It turns out they would and we could!
Since then, it's been a matter of making larger batches of scales. We bought a few 3D printers and acquired quite a few burned fingers from soldering.
We've benefited from a few technological tailwinds. For one, smartphone supply chain has driven down the cost of components quite a bit. We've been able to build hardware that works for this business model out of super cheap WiFi modules and LiPos. Also, the level of open source software for ML is quite powerful and well-documented.
We're aware that we are just scratching the surface of re-ordering hardware. We'd be interested to hear ideas that the community might have about this space!