Baseline before you modify

The measurement people skip is the one before.

It is understandable. Baselining a stock car costs money and produces a number nobody wants to frame. The exciting run is the one after the parts go on. But a number after, with nothing before it, tells you almost nothing, and the internet is full of confident claims built on exactly that.

Why the "after" number alone is close to meaningless

Say a car runs a given figure after a modification. To know what the modification did, you need to subtract something. Without your own before-figure you are left borrowing one: from the brochure, from a forum post about a similar car, from the vendor's claim.

Every one of those substitutes is measured on a different car, on different equipment, in different conditions, possibly in a different unit convention. The difference between them can easily exceed the change you are trying to detect. So the "gain" gets attributed to the part, when a share of it belongs to the weather, the equipment, or the arithmetic.

The fix is unglamorous: measure your own car before you touch it.

Two runs are only comparable if you controlled what changed

A baseline is only useful if the second measurement differs in the one way you intend. That means recording the conditions, not just the result.

What varies between runs and can move a number:

• Ambient conditions. Air temperature, pressure, and humidity all affect what an engine makes. Elevation matters. Correction factors exist to normalise for this, but they are a model, not a measurement, which is why the raw conditions belong in the record alongside the corrected figure.
• Equipment and setup. Different dynos read differently, and so does the same dyno on a different day with a different strap-down or tyre pressure. For timed runs, surface, grade, and traction dominate.
• The car's own state. Fuel: grade, ethanol content, how much is in the tank. Coolant and oil temperature at the start of the run. Tyre condition and pressure. Brake and accessory drag. Whether anything is heat-soaked from the previous pull.
• Method. Which gear, what launch technique, how many runs, whether you took the best or the average.

You do not need laboratory control. You need to write down what you did not control, so that when two numbers disagree you can tell whether the car changed or the day did.

What to record for every run

Whatever the measurement is, the record needs the same shape:

• the result, and what kind of result it is
• the date, and the odometer
• the conditions you can observe
• the car's configuration at that moment: what modifications were on it, what fuel was in it, what software it was running
• how the run was performed
• who measured it and on what

That last pair is what makes a number citable later. "It made this" is a claim. "It made this, on that equipment, that day, in that configuration" is evidence.

Baseline at the right moments

Once is not enough, and constantly is a waste.

• Before the first modification, so there is an origin point.
• Before and after any change you expect to move the number. Immediately either side, with as little else different as possible. One change at a time is the only way to attribute a result to a cause.
• Occasionally with nothing changed at all. A repeat run on a different day, in different conditions, on an unchanged car, teaches you how much noise your setup has. Everyone underestimates this, and it is the single most useful calibration you will ever do on your own data.
• After maintenance that could plausibly matter, so a later surprise has a nearer reference point.

Beware the comparison you did not intend

Two failure modes account for most bad conclusions:

Changing several things at once. Hardware, software, and fuel all at the same time yields one number and no attribution. If the result disappoints, you cannot tell which change let you down; if it delights, you cannot tell which one to credit or repeat.

Comparing across configurations you have forgotten. A run from last summer on different fuel, before an unrelated repair, is not a clean reference, unless the configuration was recorded with it. This is precisely why the run record has to carry the car's state, not just the result.

Progression is the point

The value of baselining is not any single comparison. It is that after a few years you have a series: what the car made, when, in what state, under what conditions. That series answers questions no single run can. Whether a change held up, whether something is quietly getting worse, whether last month's disappointing pull was the car or the weather.

That is also the honest form of a performance claim. A range, with its conditions and its provenance attached, is worth more than a single big number with none.

Where BuildBay fits

The Performance Lab is built around exactly this shape: dyno sessions and timed runs stored with their conditions and the car's configuration at the time, kept as a progression rather than a leaderboard, with estimates always given as ranges carrying their provenance rather than as a single confident figure.

Whatever you record it in, record the conditions. A number without them is a souvenir.