Score Before You Ship

The loop is 5 steps and it runs the same way on a still, on a cut and on a page of prose. Lock first. The template, the identity, the voice, defined before a single frame or sentence exists, because everything after that point is judged against it. A piece created with no locked target cannot be scored. It can only be argued about. Create second, fast, badly, whatever it takes to exist. The first version has 1 job, which is to give the next step something honest to look at. Score third, against the numbers rather than the feeling. Diagnose fourth, and name exactly 1 gap. Rebuild fifth, that gap and nothing else. 3 disciplines, 1 spine. About 100 per cent of the value sits in steps 3 and 4, which are the 2 steps almost everyone skips, because creating feels like progress and scoring feels like criticism.

8 numbers carry the score and each 1 has its own floor bolted underneath it. Identity at 85. The same face and the same body across every frame, the anchor carrying it, no drift between shot 1 and shot 40. Light at 85, meaning a source that exists in the world, a temperature that is named, a direction that holds all the way through. Composition at 80, the eye landing where you intended and nowhere else. Continuity at 85, with the 30 degree rule, the 180 degree line, screen direction and eyeline all clean across the cut. Anti-AI at 90, no waxy skin, no extra fingers, no invented lettering, no morph between frames. Impact at 90, because the piece has to do something to the person watching it, and doing nothing is a fail at any level of polish. Grade at 85, 1 palette, flat capture, no drift between shots. Craft at 90, which is the bar of the discipline itself, the reason for every cut in film, the register in image, the fusion in writing. 8 dimensions. 60 per cent of them are already hard gates before anyone looks at whether the thing is any good.

Score all 8, average them, then throw the average away if any single dimension sits below its floor. A still with perfect light and perfect composition and a face that drifted is still a fail. A story with flawless prose and 1 clinical word in the wrong phase is still a fail. That is not harshness. That is the only reading that matches what an audience actually does, which is forgive 7 flaws and stop dead at the 1 that breaks belief. So a 90 average over 8 dimensions can still be a 0 in the only column that mattered. The floor is not a preference. It is the gate the piece has to walk through, and 1 number under it closes the whole thing regardless of the other 7. This is also why the score has to be written down honestly. Score 85 when it is 85. The rubric is a mirror, not a certificate, and a mirror that flatters is worth nothing.

Diagnosis is 1 clause. The face drifted between shot 3 and shot 4. Not this looks off. Not everything is wrong. Everything is wrong is a description of a mood, and it gives the next loop nothing to aim at. 1 clause is a target. The face drifted between shot 3 and shot 4 tells you exactly which 2 frames to hold against each other and exactly which piece of the asset library let go. Then map the gap to the file that owns it, because every gap has 1 home. Identity belongs to the production template, light to the lighting canon, continuity to shot alignment, the artefacts to the pitfalls list, impact to the shot brief, the grade to the grading notes. Without that map, a correction becomes a re-roll, and a re-roll is a 2nd attempt at the same mistake with different luck attached to it.

Score 85 when it is 85, or the loop never closes.

Fix the source and not the symptom, then change 1 thing and close the loop. A drifted face is a broken anchor, so fix the anchor. Do not re-prompt and hope. A jump cut is a shot pair that was never designed, so redesign the pair. Do not lay a crossfade over a mistake. A shot that came out cold gets regenerated warm or it gets graded. It does not ship cold with a note attached. Then only 1 fix per pass. Change 2 things at once and the score cannot tell you which 1 of them moved it, so the next loop is guesswork wearing the costume of work. The cost of doing this properly is small and known. About $0.01 to $0.10 for a still. About $0.20 to $1.00 for a video clip. Against $37.50 a round of billable time for a correction, the loop is the cheap part and the ship is the expensive mistake.

The loop has 3 exits and only 3. Exit 1 is the number. Clear every floor and it ships. Film and image clear at 90. Writing clears at 95, because a bad sentence is cheaper to fix than a bad frame and there is no excuse for the extra 5 points being left on the table. Exit 2 is the plateau. If the score has not moved for 3 loops, the form is the ceiling rather than the draft. A single viewpoint single night story tops out far below the sky, which means the correct move is to change the form and not to rewrite the same object a 4th time. A rubric that is not moving is a rubric that is wrong, so re-read the canon instead of re-rolling the prompt. Exit 3 is the budget. Stop, log what the loops taught you, and resume later. Resume-safe logs make that trivial, and a run that ended with 40 scored versions and a written record beats a run that ended with 60 unscored ones. Built companies across twelve countries. Restructured a industrial group. Deployed 210 energy systems across Africa and Asia. Took a company through a Frankfurt IPO. Entrade built 200 wood-gasification machines across the UK and Europe. 16 years of field work from rural Nigeria to Fukushima. 393 pages in paperback and Kindle, published 8 April 2026. The probability that an unscored first version clears the bar is close to 0. The probability that the 3rd scored version does is close to 1, and the only reason it ever gets there is that someone was willing to put a number on the draft they had just made. Lock before you create. Score before you ship. Name the gap in 1 clause, fix the source and not the symptom, 1 fix per loop, and never ship below the floor.