The session ends. The driver climbs out, still processing what the car did, and talks for six or seven minutes about balance: where it pushed, where it came loose, what changed as the tires went off, and the one corner where the thing that happened does not have an obvious name.
Three words go into the notebook. Loose off two.
Six weeks later the team returns to the same circuit and consults the log. It says loose off two. Nobody can reconstruct which phase of the corner, at what point in the run, with what fuel load, or whether the driver had said the same thing about turn four and been ignored because the notebook was full.
The setup log is the team’s only durable asset from a season of testing. It is routinely built from the least reliable step in the process.
What goes in a race debrief?
A useful debrief records handling behavior by corner and by corner phase, on a consistent scale, together with how that behavior changed over the run and what was altered on the car beforehand. Practitioners who work on driver feedback recommend breaking each corner into phases and rating understeer or oversteer on a fixed scale, commonly one to five, so that reports can be compared across sessions rather than interpreted fresh each time; the approach is set out in guides such as Your Data Driven’s process for driver feedback. The same guidance stresses recording feedback immediately on exiting the car, which is where most of the difficulty lies.
Driver feedback is language, and the language is the data
Telemetry answers what the car did. It does not answer what the car felt like to drive at the limit, which is the input the engineer needs in order to choose between two changes that would produce the same lap time on paper.
That input arrives as speech, and it arrives in a form that resists compression. A driver saying it turns in fine but then it just stops rotating halfway, and I have to wait for it, is describing something specific about mid-corner balance. Written down as understeer mid-corner, it has lost the sequence, which is the part that distinguishes a mechanical grip problem from an aero balance problem from a driver technique issue.
The distinction matters because those three have different remedies, and the notebook entry supports all three equally.
What gets lost between the car and the notebook
Four things, consistently.
Sequence within the corner. Whether the behavior appeared on entry, at the apex, or on exit is usually the single most diagnostic detail, and it is the first casualty of summarising.
Progression over the run. A car that is neutral on new tires and loose after eight laps is a different problem from one that is loose from the first lap. Debriefs conducted verbally and recorded in summary form tend to capture the state at the end of the run, because that is what the driver remembers most vividly.
Conditional statements. Drivers hedge accurately. It is only a problem if I get on the throttle early is a precise piece of information. The notebook records it is a problem.
Everything the driver said after the person writing stopped writing. Debriefs are not evenly informative. The useful material often arrives in the fifth minute, once the immediate complaints are out of the way and the driver starts thinking rather than reporting. By then, whoever is taking notes has usually moved on.
The last one is worth dwelling on. Nobody writes for six minutes continuously while also asking follow-up questions. The person conducting the debrief is either listening properly or recording properly, and in practice they choose listening, which is the right choice and leaves the team without a record.
The vocabulary problem
There is a second failure that only becomes visible across a season.
Drivers and engineers use different words for the same behavior, and the same words for different behaviors. Push, tight and understeer all describe the front axle running out of grip, and a log that alternates between them cannot be searched. Loose covers everything from a stable rear slip angle the driver is comfortable with to a snap the driver is not.
The fix is not to insist the driver change how they talk, which does not work and wastes the debrief. It is to agree a small set of canonical terms for the log, and to translate into them during the write-up rather than during the conversation. A shared vocabulary and a fixed scale is precisely what the follow-up piece on leveling up driver feedback is describing when it treats quantified, phase-by-phase reporting as the goal.
Translation on paper afterward requires the original words to still exist, which brings the problem back to the record.
A debrief that produces a usable log
The mechanism that solves both problems is to record the debrief as audio and build the log from the recording afterward.
That changes who does what. During the debrief, the engineer asks questions and listens, which is what the debrief is for. After the debrief, someone converts the audio to text and writes the log entry from the text, in canonical vocabulary, at leisure, with the driver’s original phrasing preserved underneath.
A structure that holds up over a season:
Say the session identifier first, out loud. Circuit, date, session, run number, tyre set, fuel state. This is the metadata that makes an entry comparable to any other entry, and it takes fifteen seconds.
State what was changed on the car since the previous run, before discussing how it drove. Debriefs conducted the other way round produce logs in which effects have no causes.
Work corner by corner in track order, and within each corner by phase: entry, mid, exit. Order is what makes two sessions comparable.
Give the number on the agreed scale, then describe the behavior in whatever words come naturally.
Cover progression explicitly at the end. What changed between lap two and lap ten.
Finish with the driver’s own priority. If one change could be made before the next run, which one? This sentence is the most valuable in the whole debrief, and it is the one most often not asked.
From recording to log
Phones record to .m4a by default, which is AAC audio in an MPEG-4 container, and most transcription services take the file as it is with no conversion.
A tool that will transcribe M4A recordings returns punctuated text with the speakers labeled, which matters in a debrief involving a driver, an engineer, and a crew chief talking in sequence. Vomo divides longer recordings into timestamped sections and produces a summary with action items, which maps onto a debrief where the action items are the candidate setup changes, and exports to plain text, Word, PDF or Markdown so entries can be pasted into whatever the team uses for its log.
Two checks before an entry is filed, and both concern the same category of error. Numbers and corner designations need verifying by eye, because figures and short alphanumeric labels are the least reliable part of any automatic transcription, and a rating attached to the wrong corner is worse than no rating. And the canonical vocabulary needs applying deliberately, since the point of the exercise is a log that can be searched in six weeks.
One procedural note. Debrief recordings capture team members discussing performance and occasionally each other. Participants should be told the session is being recorded, the purpose should be stated, and the audio should be deleted once the log entry is written. This is a thirty-second conversation to have once, at the start of a season, rather than an argument to have in September.
What the log is for
Across a single weekend the benefit is modest. The engineer remembers the debrief.
The value appears across a season and across circuits. A searchable log answers questions that are currently unanswerable: whether this car has always been loose on exit at this type of corner, whether a change made in April produced the effect it was supposed to, whether the driver has reported the same sensation at four different tracks and had it attributed to four different causes.
Those are the questions that separate a team that is developing a car from a team that is adjusting one. Both work from the same debriefs. Only one of them can still read them in October.
The short version
Driver feedback is the highest-value data the team collects that no sensor produces; it arrives as several minutes of speech, and it is conventionally preserved as three words written by someone who was also trying to listen.
Recording the debrief and building the log from the text afterward separates listening from documenting, which is the only way both get done properly. The additional cost is a few minutes per session. The thing being bought is a log that is still useful at the end of the season.
