How your data
was meant to work

No model is perfect, but we pride ourselves on leading the industry. Our early models were approximately 2x more accurate than competitors, and our newest SPT model has widened that gap to roughly 4x for marathon predictions.

Statistically, the SPT model yields less than 3% error for over half our users and stays under 5% for the vast majority.

Some error is natural as models evolve, but much of it stems from external factors currently outside our dataset—like race-day weather (heat, humidity, wind) or sleep. As we incorporate these features, our precision will only increase.

Other variables are harder to measure, such as shoe choice, fueling execution, or mid-race injuries. Data quality also plays a huge role; broken sensors or erratic heart rate data can impact results. Considering these real-world variables, a 3% margin of error is exceptional. We have found that the cleaner your heart rate data, the sharper your prediction.

If your predictions ever seem off, please reach out on Discord or email support@kailo.fit. It doesn't happen often, but when it does we genuinely want to hear about it. Your feedback helps us get better for everyone.

They solve different problems, and Kailo is built to sit alongside them rather than replace them.

Strava is where your runs live and where your friends are — we sync from it, and we're not asking you to leave. What Strava doesn't set out to do is model how you responded to the work you actually did. That's our half: where your fitness stands right now, which way it's trending, and what it adds up to on race day.

Runna and apps like it are built around handing you a plan to follow. Kailo starts from the training you've already done. If you're on a plan you like — theirs, your coach's, or your own — Kailo tells you how it's actually landing and what today's session should look like given how you're recovering.

Tldr; Strava records it, and Runna prescribes it. Kailo explains it and predicts what it adds up to.

Beacon and LiveTrack answer one question: where are they right now? It's for a short list of contacts, for the length of one activity. Kailo Live is built to be watched.

Anyone you send your link to opens it in a browser: no app, no account, no invite list. The link is permanent (kailo.fit/watch/yourname), so your people bookmark it once and it works for every run after that.

What they actually see: your route drawing itself in 3D as you run it, with live pace, heart rate, distance and elevation. They can send messages and reactions mid-run, and the words land on your watch. You choose who's allowed in — public, followers, clubs, or unlisted — and sharing your heart rate is its own separate switch. Followers get a push the moment you go live, so someone always knows you're out there.

And if you'd rather broadcast than share a link, the same session drives an overlay you can drop straight into OBS for Twitch or YouTube, with your live stats on screen.

One thing worth being straight about: Kailo Live is not an emergency or safety service. It lets the people who care about you follow along and cheer — it doesn't call anyone on your behalf.

You can, and you should. That's the point. Kailo isn't competing with Claude or ChatGPT; it's the layer underneath them. Our job is to organize your training history so the answers you get back are grounded in your actual data instead of generic advice.

Ask an assistant about your training without that layer and it's working from whatever you remembered to paste into the chat. Connect Kailo and it can pull your runs, your splits and streams, your sleep, HRV and resting heart rate, and your race predictions, then reason across all of it at once.

Use the Strava MCP alongside it if you like. It'll read your activities and your plan. What it won't do is map you a route, run a real analysis across hundreds of activities, build a training plan and push the session to your watch, or take into account how you've been sleeping. We've built Kailo around assistants since day one, so those are first-class tools rather than something bolted on afterwards.

We try to give away as much as we can. Everything that tells you where you stand is free: marathon race predictions (the real number, not a teaser), live tracking and streaming — your watch link, viewers, chat and the OBS overlay — the Kailo Connector that puts your training data and Kailo's read tools inside Claude or ChatGPT, and the apps themselves on iPhone, Apple Watch and Android, syncing Strava and Garmin.

Premium is $10/month, and it's mostly the training side — the tools that tell you what to do next: training plans and training status, course-specific race pacing, unlimited Claude chat inside the Kailo app, the deeper analysis and dataset tools in the connector, and your full training history (free accounts see the last six weeks).

Every account gets a 28-day Premium trial, no card required.

RaceSPT uses your last 30 Strava activities, but the typical marathon race predictor uses your last 1-year of runs.

Yes, but most of our focus is on building tools for marathoners at the moment. The marathon is an aerobic event, so most training recommendations are transferable to shorter distances, usually by running less mileage.

Your data is your data. When you delete an account, we delete all your data. You also have the option to delete your data without deleting your account. We use your data in an aggregated way to improve our products, but users can opt out of this in the settings page.

No. Strava forbids us from training models on Strava data.