Transmission diagnosis: the sequence that survives a comeback

A diagnosis is a sequence, not a hunch

Transmission complaints arrive as conclusions. The vehicle "needs a transmission". The customer has been told it is the torque converter. Somebody's cousin says it is the solenoids. None of that is evidence, and the fastest way to spend a customer's money badly is to start with the most expensive possibility and work backwards.

The sequence below is the one that stands up in a shop: every step either rules out a class of failure or produces a measurement that the next step depends on. Run it in order, write down the result of each step, and the decision to remove the unit becomes a conclusion the evidence supports rather than a bet.

The order is not a suggestion

Two rules matter more than the rest of the sequence.

The first is that electrical tests and self-tests come before the pressure test. Ford's workshop procedure says it directly: certain sensor failures can cause high line pressure through failure mode effects management, so the self test and electrical repairs must be completed first or the pressure test results will be wrong. A pressure reading produced while the module is in a limp strategy is not a measurement of the transmission.

The second is that the line pressure test comes before the stall test, always. The same procedure carries a warning: if line pressure is low at stall, do not carry out the stall test or further transmission damage will occur. It also caps wide open throttle at five seconds per range and requires the converter to cool at about 1,000 rpm in Neutral between ranges. A stall test on a unit that cannot produce pressure is how a diagnosis turns into a repair.

The other ordering principle is cheaper and simpler: work from what rules out the most, for the least money, first. Fluid level, codes and live data cost almost nothing and eliminate whole categories. A rebuild costs thousands and eliminates nothing if the pressure problem was upstream of the clutches.

Getting the complaint right

Step one is the step most often done badly, because it is done from the driver's seat of the customer's description rather than from the driver's seat of the vehicle. What you want on paper is specific: it slips when it shifts from second to third, hot, at light throttle. It shudders at 45 mph only when the converter should be locked. It bangs into reverse first thing in the morning and then behaves. It has no third gear at all.

Those four sentences lead to four different test plans. "It slips" leads to none of them.

Adaptives, and when clearing them is the wrong move

Late units adapt shift pressure and timing as they learn wear. That adaptation hides problems: a unit that has been correcting for a leaking circuit can look fine on a scan until the correction runs out, at which point it slips in every gear. Two practices follow:

• Read and record adaptive values before touching anything. A large correction in one clutch tells you where the wear is, even when the shift feels acceptable today.
• Do not clear adaptives as a diagnostic step. Clear them only when the repair changes the hardware the adaptation was compensating for, and expect a relearn drive cycle afterwards during which shift quality is not representative.

The pan is a decision, not a ritual

Pulling the pan early costs a gasket and fluid, and it can destroy the evidence you need. Pulling it late wastes the customer's money on tests that the debris would have answered.

Read the pan the way a builder does. Fine grey paste on the magnet is normal wear. A dark film of friction material means clutches are wearing and the unit is living on borrowed time. Brass or bronze glitter points at bushings or washers. Hard, magnetic chips point at gears, bearings or the pump, and a converter failure often announces itself as debris in the pan and a scored pump rather than as a code.

In some unit families the pan is more than a wear gauge. On the 68RFE, debris in the pan is a documented contributor to solenoid switch valve bore wear, and the valve body cross-leakage that follows shows up as burnt clutches and lockup problems. On a 4L60E rebuild, the pan decides the scope of the job: fluid and a filter, a valve body repair, or the whole unit.

A diagnosis has a price, and it should be a line item

Published 2026 estimates put a comprehensive transmission diagnostic at about 100 to 200 dollars at independent shops, with dealership flat fees starting around 115 and running to about 200 dollars, and one to four hours of labour depending on how far the sequence has to go. Many shops credit the fee toward the repair when the work is authorised.

Charging for it is not a money grab; it is what makes the sequence possible. A shop that gives away diagnosis must make the money back on the repair, and the pressure to find a repair worth selling is exactly how a fluid level turns into a rebuild. Selling the diagnosis as a defined deliverable, with the tests named and the findings written down, aligns the shop's incentive with the customer's: the goal is the correct root cause, not the largest invoice it can support.

Documenting each step is what prevents comebacks

A comeback is rarely a technician who cannot diagnose. It is a job where the diagnosis existed and did not survive the handoff, or where verification was skipped because the vehicle "felt fine" on the way out. Three specific failure modes, and what stops them:

• The diagnosis lived in one person's head. The repair is done by someone else, from a verbal summary, and the actual measured value never reaches the bench. The fix is a record that names the test and the number.
• The repair was approved before the evidence was complete. A confident guess moved straight to parts. The fix is a case that cannot look finished until it is complete, and that separates completeness from confidence rather than treating certainty as proof.
• Verification was optional. The vehicle was road tested at walking pace and invoiced. The fix is a gate: the test that reproduced the complaint is the test that closes it, and the invoice does not open until it passes.

That last point is where a comeback gets its price tag. A comeback is not a new job; it is the original job arriving again, with the bay occupied, the parts gone and the labour already spent. If the original conclusion is on the record, the second visit is a comparison: what changed, what was not tested, what was ruled out on evidence that turned out to be thin.

The workflow this sequence implies

Every shop that runs this sequence already knows what the record should contain. The problem is that it lives across a ticket, a group chat, a scan tool session and a technician's memory, and the next person to look at the vehicle gets a summary rather than the evidence.

That gap is the one BayStacker is built around. The complaint is captured as a complaint, by voice or by typing, and it is explicitly not a diagnosis. Analysis A will not run without a decoded VIN, so the unit is identified before any hypothesis exists. The case is the primary object that every later step reads from. Analysis A returns a ranked hypothesis with the evidence behind it and the next test, and stays a draft until a person confirms it; after the drive and scan, Analysis B returns a revised stack and the difference between the two.

The confirmation gate is structural: a person confirms the root cause, and nothing below that line moves without it. Completeness and confidence are tracked as two separate numbers, so a confident but thin case cannot look finished, and the next test is the gap that most moves them. Rework risk is computed from the evidence that is missing, before delivery, while there is still time to run the test. Parts, the build and the repair read from the recorded conclusion rather than a verbal handoff, a reading nobody took shows as missing rather than as a guess, and a managed job does not invoice, deliver or close until verification passes. A comeback reopens the work order it came from, which means the comparison happens on the same record instead of in an argument.

None of that replaces the sequence in the table above. It is what makes the sequence survive contact with a busy Tuesday. The tests run in order, the measurements land on the case rather than in a memory, and the decision to pull the unit has a reason attached. When the sequence turns up a companion code instead of a symptom, the P0700 guide covers how to read it, and when a specific gear is the finding, the slipping diagnosis works that case from pressure to pan.