Where is the denominator?#

SLOP2805 · Week 3 · Studio framing

When your bar says 40%, forty per cent of what has happened?

Where we were, and where we are going#

Week 2 gave you states. This week turns them into a number — and finds out what the number leaves out.

By 17:00 you can:

• name the denominator behind any percentage, or show there isn’t one
• measure where two honest denominators disagree, on the same run
• say in one sentence what your chosen denominator excludes

Back to the case from week 1#

40 MB. Bar reaches 99% in nine seconds. Then eleven seconds of nothing.

One ordinary explanation: the bar was counting bytes handed to the network stack. All 40 MB went into a buffer in nine seconds. The eleven seconds were the server hashing and writing.

The bar was not lying. It was answering a different question from the one the user was asking.

Two denominators, both honest#

Byte-counted — loaded / total, straight from the platform. Cheap, smooth, frequent. Excludes everything that is not transfer.

Stage-counted — stagesDone / stagesTotal, from your week 2 model. Covers the whole process. Moves in lurches, and assumes all stages are the same size, which they never are.

Weighted stages are honest only if the weights were measured. A guessed weight is a number with no source event.

The platform is blunt about this#

xhr.upload.addEventListener("progress", (e) => {
  if (!e.lengthComputable) return showIndeterminate();  // no denominator
  setPercent((e.loaded / e.total) * 100);
});

lengthComputable is the platform telling you whether a denominator exists at all. Most code reads e.total without checking.

That flag is next week.

Worked example — where they part company#

Simulated run, trace-office-wifi.json, both readings on one timeline:

Largest disagreement: 49 points, at t = 14 s — the moment the bytes finished and the verification had not started reporting.

Your turn — the build#

1. Two bars over one simulated run: byte-counted and stage-counted, both visible, both driven by the same events.
2. Log both readings on one timeline.
3. Find your largest gap. Write down the number and the moment.
4. Decide which one ships, and say on screen what it excludes.

The mistake to expect#

Writing the exclusion as a category instead of a consequence:

✗ “Shows upload progress.”

✗ “Percentage of transfer.”

✓ “Bytes sent. Does not include the server-side check, which on this trace took a further eleven seconds.”

The test: could a waiting user predict the 99% pause from your sentence? If not, the sentence is decoration.

What this week cannot tell you#

Which denominator is right.

Nothing measured today settles that, because “right” depends on what the user is waiting to find out, and nobody has asked one. You now know what each choice omits and what that omission is worth in points — on one recorded trace.

The first outside reaction to your choice is week 11.

What leaves the room#

• two working bars and one timeline log
• one number: your largest disagreement
• a denominator statement on screen

Where it goes: the shipped denominator is condition (a) of Assignment 2 — Behaviour across the four conditions, 35 marks — and your exclusion sentence is the first row most people write in the week 10 claims register.