Chapter 03 · Predictive RSA

The best breakdown is the one that never happens.

We're not scared of roadside assistance being disrupted. We're the ones doing it. 365 uses live vehicle signals + probability to catch failures before they strip a driver on the side of the road — protecting the moment, and the brand behind it.

Scroll

Everyone else is protecting the tow truck. We're trying to make it unnecessary.

Traditional roadside assistance is a reactive business: something breaks, a driver panics, a call comes in, a truck rolls. The whole industry is optimised around responding fast to failure. That model made sense when the car couldn't tell you anything.

Modern vehicles are rolling sensor arrays. Tyres, batteries, thermal loads, fault codes, charge cycles — the car is talking constantly. So the real disruption isn't a cheaper tow. It's using that data to eliminate the breakdown entirely: monitor trends, spot the pattern, run the probability, and reach the driver before the failure — with best efforts to route them to a dealer while the car is still moving.

Yes — that cannibalises a chunk of our own reactive callout volume. We're fine with that.  A premium OEM doesn't want more successful rescues. It wants fewer roadside moments, full stop. The brands that own the prevention layer own the customer relationship. The ones that don't will be explaining to their drivers why the car "should have known."

Live vehicle telemetry
the car is talking constantly →
Four data streams, one verdict. The signals are already there — the question is whether anyone acts on them.
The old model

Reactive Roadside

  • Waits for the car to fail
  • Driver stranded, then rescued
  • Brand moment = a bad day, remembered
  • Success measured in response time
  • Data ignored until it's too late
➜
The 365 model

Proactive / Predictive RSA

  • Reads the car's signals continuously
  • Driver guided to a dealer before failure
  • Brand moment = "my car looked after me"
  • Success measured in breakdowns prevented
  • Every signal feeds the probability model

Same light. Smarter car. Same broken deal.

Half a century apart, the oil light still does the exact same job — it tells the driver something's wrong and then waits for them to do something about it.

1972
The analog era

A single amber bulb behind the speedo and tacho. The driver has to notice it, understand it, and act — pull over, check the oil, get to a garage. Miss it, and the engine's gone. The car warns; the human carries all the risk.

2026
The connected era

A full digital cluster, hundreds of live signals a second, a beautiful amber icon. And still — the driver is the one expected to act. Same warning. Same burden. Just rendered in higher resolution.

50 years apart. Same warning light. Same broken deal.

The '70s driver got an oil lamp. The 2026 BMW driver gets a beautiful digital cluster streaming hundreds of signals a second. The technology transformed — the contract never did. The concept hasn't moved an inch: the car still just warns, and the driver is still the one expected to notice, decode, and act. Or break down.

We're in the golden age of AI and data — yet the data mostly just sits there, watched but unused. Collecting it isn't the win. Data you collect but never act on isn't intelligence — it's just a prettier warning light. If we have the signal and don't turn it into a change, an action, an intervention for the customer, then why is it even relevant?

That's the gap 365 closes. We don't just show the light — we see the signal, run the probability, and act: reaching the driver and routing them before the light ever mattered. The data finally does something.

The failure-probability model — live.

Drag the signals. Watch the probability of imminent failure recalculate in real time. This is a simplified view of the model 365 runs across a fleet — each live signal is weighted, decayed over time, and fused into a single risk score P(fail) per vehicle.

Live model

Vehicle · WBA-8E9C-2291

P(fail) = 1 − Π(1 − wᵢ·sᵢ)
Live vehicle signals — drag to simulate
🔋 Battery health degradation 62%
Cranking voltage drop + charge-cycle age vs baseline
🛞 Tyre pressure / wear anomaly 40%
Slow-leak trend + tread depth vs fleet cohort
⚠️ Recurring fault-code frequency 28%
DTC repeat-rate + severity class from telematics
🌡️ Thermal / cooling stress 18%
Coolant temp excursions under load, ambient-adjusted
— P(fail) · 14 days
Est. time-to-failure window: —
Healthy
Monitoring only. No action needed.

Illustrative model for demonstration. Production weighting is trained per make/model/component on real fleet failure history and continuously recalibrated.

From signal to saved — the interaction.

When the model crosses threshold, a coordinated chain fires across the vehicle, 365's AI workforce, the customer, and the dealer — before the breakdown. Press play to watch the hand-off.

📡
Step 01

Vehicle emits signals

Battery, tyre, fault-code & thermal telemetry stream continuously from the connected car.

Vehicle
🧠
Step 02

365 model scores risk

Signals fused into P(fail). Trend crosses threshold → predicted failure flagged with a time window.

365 · Predictive AI
📲
Step 03

Proactive outreach

Customer contacted in the brand's voice: "We've spotted something — let's sort it before it strands you."

Customer
🏬
Step 04

Directed to dealer

Booking pre-loaded with the predicted fault + parts flagged. Dealer ready before the car arrives.

OEM Dealer
✅
Step 05

Breakdown avoided

Fixed in a service bay, not on a motorway shoulder. No tow. No stranded driver. Brand protected.

Outcome
Breakdown prevented. The driver never knew how close it was — and that's exactly the point.

Get on board — or explain why the car didn't know.

Predictive RSA isn't a feature OEMs can add later. It's becoming the baseline expectation of a premium connected vehicle. The brands that embed it define the standard. The ones that wait inherit the complaints.

🚀

OEMs who get on board

  • Own the "my car looks after me" moment — the strongest loyalty driver in premium automotive.
  • Convert unplanned roadside failures into planned, profitable dealer service visits.
  • Fewer stranded-driver incidents = fewer brand-damaging stories, higher NPS.
  • A data flywheel: every prevented failure sharpens the model for the whole fleet.
  • Position the brand as intelligence-era, not mechanical-era.
📉

OEMs who fall behind

  • Drivers left stranded by faults the car had already flagged — an unforgivable look in 2026.
  • Reactive-only roadside gets commoditised and undercut on price.
  • Competitors' cars "predict and prevent"; yours "break down and rescue."
  • No prevention data = no model = a permanent lag they can't buy back quickly.
  • The customer relationship at the moment of failure belongs to whoever reached them first.
0
Signals monitored per vehicle, continuously
0
% of breakdowns that are predictable
0
Days of early warning on typical failures
0
Roadside moment turned into a dealer visit

The industry's about to split in two.
Predict, or get predicted around.

365 already runs the operational intelligence. Adding the prevention layer to your fleet is a partnership conversation, not a research project. Let's build the roadside that never rings.