Why We Built Base: The Spreadsheet That Kept Breaking

What the reconciliation actually looked like

Three questions, every week, all requiring data from at least two places.

Is my surplus real? Requires: average intake from the macro tracker, weight trend from the scale app, and endurance volume from the running platform. The macro tracker's calorie target came from a dropdown that had never heard of a long run.

Should I deload? Requires: lifting volume and RIR drift from the lifting app, endurance load from the running platform, HRV and sleep from a fourth source. The lifting app's fatigue model — where it had one — assumed lifting was the only stress.

Are my easy runs easy? Requires: threshold from the running platform and the actual paces of the last four runs, compared. Nothing computed this, because "easy" was a label you chose rather than a band you were graded against.

None of these is hard arithmetic. All of them need a spreadsheet, because no app had all the inputs.

Why the spreadsheet stopped

Not because it was wrong. Because it required a weekly act of will, and the plan required months.

The first six weeks it is interesting — you are learning things about your own numbers. Around week six the answers stop being surprising, and the twenty minutes of copying figures between apps becomes a chore with no immediate reward. So one Sunday it does not get done, and then it does not get done again.

The failure mode is not that the athlete stops caring. It is that the reconciliation is the least rewarding part of the process and the part with the longest feedback delay. You do not find out you skipped it until eight weeks later when the scale has gone somewhere unexpected and you cannot reconstruct why.

What broke first, in practice

The surplus. Every time.

The arithmetic is brutal and it is worth writing out. A pound of lean tissue costs roughly 2,500 kcal, so a trained lifter gaining 0.5 lb a week needs about 179 kcal a day of usable surplus. An advanced lifter at 0.25 lb a week needs 89.

A 40-minute easy run costs 400 to 500 kcal.

If the maintenance figure came from a week without running — or from a calculator, which is worse — then adding three runs a week moves expenditure by 200 to 400 calories a day while intake does not move at all. The surplus is gone. Not reduced: gone, and possibly inverted.

And the diagnosis from inside is always wrong. The scale has not moved and the squat has stalled, so the conclusion is "my metabolism", or "I'm not training hard enough", or "interference is real after all". The actual answer is that a number in one app never learned about a number in another.

The two things that had to be true

Maintenance had to be measured, not predicted. Every method that adds up your expenditure inherits the error of every estimate in the sum — the watch's active energy, the running economy assumption, the unknowable non-exercise movement. Measuring the outcome instead sidesteps all of it: trend weight change over fourteen days against actual intake, and every calorie the training cost is in the number whether or not anyone logged the session.

Fatigue had to be read across both disciplines. A deload trigger that only knows about lifting will miss the athlete whose problem is a spiked endurance week. The signals that work are boring and shared: recovery against your own baseline, RIR drifting up at fixed load, short sleep, flat strength trend. Two firing at once is a pattern; one is noise.

Everything else in Base follows from those two.

What was deliberately not built

Not a coach. It does not write your programme. There are good coaches and there are good templates, and a piece of software confidently prescribing an eighteen-week block from four data points is neither.

Not a session runner. Apple Watch and Garmin already do live pacing, zone alerts and structured intervals, natively and well. Building a worse version helps nobody. The half they do not do is deciding what today's session should be given everything else in the week, and grading it honestly afterwards.

Not a social network. No feed, no kudos, no leaderboard. That is a different product with a different business model, and the business model is the part that changes what gets built.

Not a blended fitness score. The reasoning is here.

The thing that is actually hard

Not the arithmetic. The arithmetic in this whole app is a few hundred lines and none of it is clever.

The hard part is making a daily loop cheap enough to survive years. Sixty seconds a day is the budget, and it is the constraint that every screen was designed against: bodyweight pre-filled with your trend, a whole day of food as one tap on a saved template, each set as a confirmation of last session's numbers already in the fields.

A logging habit that costs five minutes a day survives about three weeks. One that costs one minute survives long enough to answer the questions that need six months of data — which are all of the interesting ones.

What to do next

What hybrid training is is the full version of the problem. How Base works is the workflow that came out of it.