Two Ways to Read a Session

You said it should make the Neurosity console less confusing. It cannot do that, because the confusing thing never appears on screen.

The console gives you focus: 0.407 and no idea what to do with it. The app’s answer is “2 hr 23 min of focused work”, a different number, asserted, with no visible relationship to the one you were confused by. There is no before and after, so there is no act of translation. It reads as a second opinion from nowhere.

• ✕No raw data anywhere on the main screen Not one figure the Crown actually produced. A translator that hides the source text is just a page of assertions.

• ✕No provenance Nothing says which headset, which recording, how many readings, how many were thrown away, or what turned them into the answer. The derivation chain is the product, and it is buried in core/stats.js.

• ✕No introduction Nothing on screen states what the app is for or what it just did. “How it works” is a page you navigate to, which means leaving the thing you were confused by in order to read about it.

• ✕The dashboard is a stack of unrelated boxes Cards in a column with nothing showing how any number relates to any other. That is a list wearing dashboard clothing, and you are right that it fails even as a dev panel.

The correction

Every claim needs a visible path back to a number the Crown produced. Either show the raw data next to the reading of it, or lead with the reading and let anyone open the working. Those are the two mockups below, same session, same numbers, opposite starting points.

The screen is split. Left is exactly what the Crown recorded, in the form the console gives it to you. Right is what each of those things means. The layout is the argument.

localhost:5273, Session Debrief

A work session, translated

Neurosity’s app gives you these numbers. This tells you what they mean.

Crown-2B1 · 8 sensors 09:00-17:30 13,501 readings one every 2s synthetic sample data

What the Crown recorded

Raw stream · first rows

09:00:00 focus 0.3811 calm 0.4702 09:00:02 focus 0.3794 calm 0.4688 09:00:04 focus 0.3823 calm 0.4651 …13,501 rows

Focus · all readings

0.407 median

10th, 90th percentile 0.287-0.578

Power by band · session mean

1.00 DELTA

0.51 THETA

0.58 ALPHA

0.35 BETA

0.28 GAMMA

Sensor contact

CP3 C3 F5 PO3 PO4 F6 C4 CP4

1,338 readings discarded

What it means

The raw stream

Every two seconds the headset writes down two scores. Neither is a measurement, they are a model’s estimate, from your brainwaves, of how focused and how calm you were at that instant. Nothing here is interpreted yet.

Your focus number

0.407 is a normal working level for you. Neurosity says anything over 0.3 is meaningful, so this is not “40% focused”, that reading is the most common mistake. Most of your day sat between 0.29 and 0.58, and that span is what “your usual” means everywhere else on this page.

The bands

Brainwaves sorted by speed. Delta being largest is normal for everyone and does not mean you were asleep. The two that matter here: alpha is what the calm score is built from, gamma is what focus is built from.

Sensor contact

Two sensors lost contact at times. A sensor sitting on hair still produces numbers, they just mean nothing, so those 1,338 readings were thrown away rather than averaged in. 90% of the session survived, which is enough to trust the rest.

What this gets right: it is unmistakably an interpreter. Someone arriving from the console sees their own numbers on the left and finally gets told what they mean. Provenance is in the header, permanently. It also serves a GFT engineer, who can check the translation against the source.

What it costs: it is dense, and it asks you to read two columns. Someone who just wants to know how their day went has to do work to get there.

Leads with the finding in a sentence. Every claim carries a control that opens the full chain from raw readings to that number, shown here already open, because the chain is the point.

localhost:5273, Session Debrief

One work session

Reads a recording from your Neurosity Crown and works out what happened in it. Every number below can be opened to show where it came from.

Crown-2B1 13,501 readings 12,163 usable synthetic sample data

You concentrated for 2 hr 3 min of the 7 hr 30 min you recorded, almost all of it before lunch.

Which is a typical Friday for you, 2 minutes off your average across the last ten sessions.

Time concentrating 2 hr 3 min Hide the working

Recorded 13,501 readings, one every 2 seconds

→

Sensor check −1,338 discarded, poor contact

→

Your usual range 0.29-0.58 learned from this session’s own readings

→

Clearly above it 3,285 readings over 0.46

→

Time 2 hr 3 min 3,285 × 2 seconds

Why “clearly above” and not just “above”: a reading that hovers on the line flickers, and counting every flicker would turn one good stretch into forty. The threshold sits half a standard deviation above your median, and a stretch has to hold for three minutes to count at all.

Longest unbroken stretch 09:53-11:53 Show the working

Clearest dip 13:30-14:38 Show the working

Signal worth trusting 90% Show the working

What this gets right: anyone can read the first sentence and stop there. The receipts satisfy a sceptic without taxing everyone else, and the chain teaches the method to whoever opens it.

What it costs: the raw data is still one click away rather than present. Someone arriving straight from the console does not immediately see their own numbers, so it is less obviously a translation of the thing that confused them.

The thing you were reaching for is called a coach mark, a bubble anchored to a real element, with everything else dimmed. You are right that a “How it works” page cannot do this job: it explains the dashboard somewhere the dashboard is not.

localhost:5273, Session Debrief · guided tour

One work session

Reads a recording from your Neurosity Crown and works out what happened in it.

You concentrated for 2 hr 3 min of the 7 hr 30 min you recorded, almost all of it before lunch.

Which is a typical Friday for you, 2 minutes off your average across the last ten sessions.

Time concentrating 2 hr 3 min Show the working

Longest unbroken stretch 09:53-11:53 Show the working

These are not exclusive. B is the shape of the app; A is a screen inside it.

My recommendation: build B as the main screen, and make A a real screen you can reach from it, called something like the raw data, translated. B answers “how did my day go” for the person who wants that. A answers “what am I actually looking at” for the person who came from the console confused. The coach marks introduce both.