A month of work that was not this
Between mid-June and late July, Slingshot sent zero messages to trainers.
The month was not idle. We incorporated. We opened the account. We rebuilt the business model around what V1 could actually do. All of it needed doing.
It was also all work where nobody could tell me no. The first round of outreach had stalled out, and restarting it was the thing I kept finding other things to do instead.
The engineers were not blocked on hardware
In July, Ryan and Kai did not need a customer. They needed to know what a trainer means when he says a kid looks tired.
The Session Load Score has to put something on a page that a trainer reads and recognizes on sight. Pick that field from a desk and you get a number that survives a spec review and means nothing on a floor.
So outreach came back as an engineering unblock. There was no sales target attached to it.
Nineteen since July 24
Every row is a private basketball trainer or an academy operator. Thirteen sit in the LA basin, Orange County, or San Diego. Five are out of state. One we could not place.
Seventeen of the nineteen went out in the eight days between July 24 and July 31, from a standing start with no list and no warm path.
Five replies from nineteen cold messages is a 26 percent reply rate. Nineteen is far too small a sample for that percentage to mean anything, which is the first reason not to be pleased about it. Here is the second.
What a reply rate does not measure
Reply rate measures whether the message works. It says nothing about whether the conversation produces anything.
Nineteen messages produced one usable answer to the question the engineers were blocked on, and zero answers to the second one.
The shape of it is the part worth publishing. Four of the five replies came off the same ask: can I come watch a session. All four of those trainers wrote back, and all four of those rows have an empty answer column. The one thread that produced answers ran differently. One question, then a wait, then the next question, across four days.
That is a sample of one, so we are calling it a lead and not a finding. It still points somewhere uncomfortable. Getting a reply and getting an answer are two different skills, and we had only been counting the first.
Three tells, all of them movement
The thread that answered belongs to Jonathan Simmons, a trainer in Atlanta with a following that dwarfs ours. He has never seen our hardware. He answered because we asked him about the one thing he is the expert on.
We asked how he tells the difference, mid-session, between a kid who is tired and a kid who is being lazy. He named three things he watches:
- -They are not jumping the same.
- -The ball is losing energy and power.
- -Their mechanics are changing.
Then he described the test he runs when watching is not enough.
"Sometimes something as simple as a 'go harder' helps too. Then if they give a burst of speed they have more to give. If they don't usually it's still fatigue."
That is a trainer describing a sensor operation without calling it one. Apply a stimulus, look for a response, treat the absence of the response as the reading.
All three of his tells are movement. None are internal. That matters for what goes on the page, because movement is what an arm-worn pod has the best chance of seeing. Whether we can actually see them is unproven and stays unproven until we have run it on a real arm in a real session. What changed is that we now know which three things to aim at, and we did not choose them ourselves.
What parents get today is film
We also asked what he sends parents between sessions. The answer was video. He breaks down film from a session and sends it over, and outside of that he says he does not text them much.
The artifact already living in a solo trainer's practice is video, and he makes it by hand, one kid at a time. Anything we hand a parent is landing next to that.
Two thousand miles from a pilot
Atlanta is two thousand miles from the only place we can currently run a pilot. A pilot means standing in the gym, handing over hardware, and coming back the following week. We cannot do that from Los Angeles.
So he stays where he is. He is the most useful conversation in our system and he appears in no pipeline number we report, in this post or in a deck later. Discovery does not need geography. A pilot does.
The same rule is why the football and soccer conversations that have come our way stay parked. One sport, one region, until one of them works.
We are writing the rule down because the pull runs the other way. A responsive, generous, credible trainer is exactly the row you want to count.
All of it ran on one channel
Every message above was an Instagram DM. There is no second channel warm behind it. If that account gets limited, or trainers stop opening messages from strangers, the pipeline goes to zero and we start from nothing.
We know that. We have not fixed it.
Four things we are changing
- 01The ask changes on discovery threads. One question at a time, spaced out, with something given back between them. The visit ask stays for in-region trainers, where a visit is the actual goal.
- 02The floor moves from sends to answers. Three rows carrying a fatigue or parent-message answer per batch of roughly twenty. Three is not arbitrary. One trainer saying something is a lead, and we already treat three saying it as a finding, so three per batch is the rate at which a batch is worth running. We are at one. Under the floor, the ask changes again, not the channel.
- 03At least half of every batch stays in Los Angeles or Orange County. The pilot is the thing that has not happened yet, and it has an address.
- 04The parent-message question is still at zero. We have asked nineteen people what they send a parent after a session. We have one answer, about film, and none about what the text itself says. That is the next hole to fill.