The verdict engine

One call. And the reasoning under it.

Any tool can list what a code might mean. The hard part, the part a 30-year veteran does without thinking, is deciding which one it actually is, and knowing how sure you’re allowed to sound.

AceDxReading from Autel MaxiSys
RO-4471
2019 Silverado 5.3L · 128,442 mi
Best call
87PERCENT
Ignition coil, cylinder 1
Swap and retest before you touch the plugs.
Part D513CLabor 0.6 hr
Evidence
DTC P0301DTC P0300Cyl-1 misfires 2,914Freeze frame 1,842 rpm
Not confirmed: compression on cyl 1 untested. Call does not rest on it.
Amber · awaiting retest
Why one call

A list of possibilities is the problem, not the answer.

Hand a tech six causes ranked by frequency and you’ve handed the diagnosis back to them with extra steps.

What a code gives you
  • Ignition coil, cylinder 1
  • Spark plug, worn or fouled
  • Fuel injector, cylinder 1
  • Compression loss in that cylinder
  • Wiring or connector fault
  • Vacuum leak on that runner
6 candidates · no ranking · your call
What a verdict gives you
87%
Ignition coil, cylinder 1
Swap and retest before you touch the plugs.
DTC P0301DTC P03002,914 misfiresFreeze frame
1 call · evidence attached · you can overrule it
Same vehicle, same data. On the left, six things it could be. On the right, what AceDx returns instead.
Evidence

Not all evidence is worth the same.

A code is a claim. A measurement is a fact. AceDx weighs them differently, and shows you which kind it used.

Measured
A value read off the vehicle right now: a misfire counter, a fuel trim, a sensor out of range.
Corroborated
Two independent signals pointing the same way. A code plus 2,914 matching misfire events is a different claim than either alone.
Captured
Freeze-frame data from the moment the fault set. The car’s own testimony, about a moment that has passed.
Reported
What the tech observed and typed in. Real evidence, but human and unmetered.
Absent
What AceDx expected to see and didn’t. A missing signal narrows a call as honestly as a present one.
Relative weighting, not published coefficients. The ordering is the claim. A measurement outranks a code because the vehicle is telling you what is true now rather than what was true once.
Confidence

The number is an instruction.

It isn’t decoration and it isn’t a score. It tells you how much weight the call will carry, so you know whether to swap the part or run the test first.

What pushes it up

  • The vehicle’s own live data agrees with the code
  • Independent signals corroborate instead of repeating each other
  • The fault pattern is specific to one part, cylinder, or circuit
  • The freeze-frame conditions match how the complaint actually presents

What pulls it down

  • The only evidence is a code, with nothing measured behind it
  • Two plausible parts explain the same data equally well
  • The complaint is intermittent and nothing is faulting right now
  • The fault is upstream of something that could be causing it

A high number says the cheapest thing you can do is fix it. A middling number says the cheapest thing you can do is test it. AceDx names which test: the one that would prove its own call wrong fastest.

Restraint

Sometimes the right answer is “not yet.”

A tool that always produces a confident answer is a tool that is confidently wrong on a regular schedule.

01It hedges instead of inventing
Won’t sayIgnition coil, cylinder 1. Part D513C.
SaysIgnition-side, cylinder 1. Not enough to name the part.

When the data supports a direction but not a component, AceDx says the direction and stops. A fabricated part number is worse than an admission.

02It asks for the one test that matters
Won’t sayCould be the coil or the injector. Start somewhere.
SaysSwap coil 1 and 2. If the miss follows, it is the coil.

When two causes explain the same data the tool does not coin-flip. It names the single measurement that separates them, so ten minutes of testing replaces an hour of parts-swapping.

Trust the color

Green and red are earned.

Color is the fastest thing a person reads across a bay. If it can be spent cheaply, it stops meaning anything.

Green means it held

A verdict goes green only after the repair is confirmed on retest and logged to the repair order. Not when the part goes in, not when the tech is fairly sure. It’s a record of what worked, not a hopeful guess.

Red means stop

Red is a safety stop: a fault that says don’t hand this vehicle back yet. It’s rare on purpose. A tool that flashes red at everything gets tuned out by Thursday, and then it’s useless on the day it matters.

The difference

This is not a chatbot with a scan tool bolted on.

Plenty of tools will now discuss your car with you. Discussion isn’t the job.

A chat assistantAceDx
What you get backA paragraph listing several possible causes, in no committed order.One ranked call, named specifically enough to order the part.
Where it came fromSomewhere in the training data. You cannot check it.The codes, counters, and freeze-frame values are printed under the call.
When it isn’t sureSounds exactly as confident as when it is sure.Says so, with a number, and names the test that would settle it.
When it’s wrongNothing happens. It never finds out.The retest catches it, the call never goes green, and the miss is on the record.
What it leaves behindA chat log nobody reads twice.A repair order the next tech can actually use.
Fair questions

What techs ask us first.

What happens when it’s wrong?

The retest catches it and the call never goes green. That’s the whole reason the loop exists. A tool that can’t be caught being wrong can’t be trusted when it’s right.

We’d rather ship a tool that visibly misses sometimes than one that quietly misses always.

Does it override the tech?

No. The tech signs the repair order. AceDx makes a call and shows its work; overruling it takes one tap, and the disagreement gets logged too. If the tool is wrong in a particular way on a particular platform, that’s worth knowing.

Do you have accuracy numbers?

Not yet, and we won’t print numbers we haven’t earned. We’re pre-pilot. The point of running with working shops this fall is to generate exactly that: measured against real ROs, on real vehicles, with the misses counted.

When we have them, they’ll go on this page with the methodology next to them.

Does it replace ALLDATA or Identifix?

No. Those are libraries of procedures, diagrams, and confirmed fixes. AceDx is the part that reads this specific vehicle and decides what’s wrong with it. It sits beside what you already pay for, not on top of it.

Put a call to the test on your own vehicles.

The pilot runs on real ROs in working shops. If the engine is wrong on your platforms, we want to find out on your bay floor, not in a slide deck.