A four-state PCA biplot that draws observation scores and variable loading arrows on one equal-aspect plane, puts explained variance in the axis titles, and states the arrow scale factor instead of hiding it.
Preview in your theme
Loading preview…
This component is part of Pro
The install command, prompt and concept map are included with a Pro license — one purchase unlocks every Pro component, forever.
status="ready" · 96 workspaces on the plane PC1 × PC2, which carry 68% of the variance between them. Free sits left, Enterprise right, so PC1 is the usual size component; PC2 splits the workspaces run by scripts (API calls, integrations point down) from the ones run by people (sessions, exports point up). Pin any mark — point or arrow — and the selection appears under the card.
Workspace feature usage · PCA
96 workspaces · 7 variables · PC1 + PC2 carry 68% of the variance · furthest workspace McKenzie Group, 3.56 from the centre
Free(34)
Team(28)
Business(22)
Enterprise(12)
variable loading
Point at a mark, or tab into the cloud and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
Arrows are drawn at 3.23× their loadings so both layers fit one plane; the ticks stay in PC1 units. One factor for every arrow, so directions and relative lengths are exactly as computed — only the absolute length is arbitrary.
Workspace feature usage · PCA— variable loadings. An arrow's direction is what it means: two variables pointing the same way move together, at right angles they are unrelated, opposite ways they trade off. Length is how well this plane represents that variable.
Variable
PC1
PC2
Length
Direction
Seats
0.88
0.21
0.9
right
Storage used
0.71
0.34
0.79
up and right
Sessions / seat
0.55
0.61
0.82
up and right
Exports / week
0.34
0.74
0.81
up and right
Tickets / seat
-0.48
0.52
0.71
up and left
Integrations
0.62
-0.55
0.83
down and right
API calls / week
0.79
-0.42
0.89
down and right
Workspace feature usage · PCA — the 6 workspaces furthest from the centre, which is the mean of the projection. The remaining 90 are reachable by tabbing into the cloud and using the arrow keys.
Workspace
Group
PC1
PC2
Distance from centre
McKenzie Group
Enterprise
3.56
0.1
3.56
Blick Inc
Enterprise
3.5
-0.56
3.54
Rogahn, Satterfield and Armstrong
Free
-3.36
-0.01
3.36
Kling - Tremblay
Enterprise
3.26
0.41
3.28
Streich LLC
Business
2.84
1.33
3.13
Ryan, Krajcik and Kiehn
Enterprise
2.77
-0.95
2.92
onSelect → nothing pinned
status="empty" · a valid contract with no projection in it, worded so it cannot be mistaken for a failed fetch — and saying out loud that this chart plots a PCA, it does not compute one.
Workspace feature usage · PCA
Observation scores and variable loadings on one plane
Nothing to project
A biplot needs a projection that already exists: one pair of component scores per observation, one pair of loadings per variable. It cannot take a raw feature table or a correlation matrix and work the components out on your behalf.
status="loading" · a skeleton cloud and four skeleton arrows, aria-hidden, honouring prefers-reduced-motion, with no timer involved.
Workspace feature usage · PCA
Observation scores and variable loadings on one plane
Loading the projection
status="error" · the fetch failed. Try again only appears when onRetry is passed, and clicking it moves focus to the card and flips it to ready.
Workspace feature usage · PCA
Observation scores and variable loadings on one plane
Couldn't load this projection
The data source didn't respond.
Awkward data · two benches arrived with a NaN coordinate and one variable did too; all three are dropped and counted on the card rather than drawn at nowhere. Humidity loads on neither component, so it has no arrow and sits at the centre — a reading, not a gap. One bench points at a group the payload never declared and gets its own legend entry instead of being counted as Calibrated. Two labels are far too long for the plot and elide, with the full text in the tooltip and the data table. PC2 carries 0.6% of the variance, so the card says this plane is very nearly a line.
Sensor bench calibration · PCA
6 benches · 4 variables · PC1 + PC2 carry 72% of the variance · furthest bench Bench 19, 2.31 from the centre · 3 dropped for want of two usable coordinates
Calibrated(5)
Unassigned(1)
variable loading
Point at a mark, or tab into the cloud and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
2 benches and 1 variable arrived without two usable coordinates and are not on the plot. Every figure on this card counts only what is drawn.
1 variable loads on neither component, so it has no arrow to draw and sits at the centre as a small ring. That is a reading, not a gap: this plane says nothing about that variable.
The vertical component carries 0.6% of the variance, so this plane is very nearly a line: read left-to-right, and treat vertical distance as noise.
Arrows are drawn at 2.01× their loadings so both layers fit one plane; the ticks stay in PC1 units. One factor for every arrow, so directions and relative lengths are exactly as computed — only the absolute length is arbitrary.
Sensor bench calibration · PCA— variable loadings. An arrow's direction is what it means: two variables pointing the same way move together, at right angles they are unrelated, opposite ways they trade off. Length is how well this plane represents that variable.
Variable
PC1
PC2
Length
Direction
Humidity
0
0
0
nowhere — it has no length on this plane
Temperature
0.94
0.03
0.94
right
Settling time after a cold power cycle
0.66
0.05
0.66
right
Drift / hour
-0.87
0.02
0.87
left
Sensor bench calibration · PCA — the 6 benches furthest from the centre, which is the mean of the projection. The remaining 0 are reachable by tabbing into the cloud and using the arrow keys.
Bench
Group
PC1
PC2
Distance from centre
Bench 19
Calibrated
2.31
0.02
2.31
Bench 27 · humidity-controlled enclosure, rack B
Calibrated
-2.04
-0.07
2.04
Bench 11
Calibrated
1.42
0.06
1.42
Bench 12
Calibrated
-1.18
-0.04
1.18
Bench 31
Unassigned
0.62
0.09
0.63
Bench 04
Calibrated
0
0
0
marks="scores" · the same 96 workspaces as a plain PC score plot. No arrows, so nothing is rescaled and the ticks are exactly the component scores — and the card says the seven arrows in this payload are not drawn, rather than letting you assume there were none.
Workspace feature usage · PCA
96 workspaces · PC1 + PC2 carry 68% of the variance · furthest workspace McKenzie Group, 3.56 from the centre
Free(34)
Team(28)
Business(22)
Enterprise(12)
Point at a mark, or tab into the cloud and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
Scores only: the 7 variables in this payload are not drawn, so nothing has been rescaled and the ticks are exactly the component scores.
Workspace feature usage · PCA — the 6 workspaces furthest from the centre, which is the mean of the projection. The remaining 90 are reachable by tabbing into the cloud and using the arrow keys.
Workspace
Group
PC1
PC2
Distance from centre
McKenzie Group
Enterprise
3.56
0.1
3.56
Blick Inc
Enterprise
3.5
-0.56
3.54
Rogahn, Satterfield and Armstrong
Free
-3.36
-0.01
3.36
Kling - Tremblay
Enterprise
3.26
0.41
3.28
Streich LLC
Business
2.84
1.33
3.13
Ryan, Krajcik and Kiehn
Enterprise
2.77
-0.95
2.92
A weak plane, said out loud · these two components carry 49.9% of the variance between them, so the card warns that two teams looking close here may be far apart in the data. Loadings here really are correlations, so showUnitCircle draws the ceiling every arrow must sit inside: Tooling nearly touches it, Meeting load is well short of it and is therefore poorly represented on this plane. Five groups exercise all five shapes.
Employee survey · PCA on the correlation matrix
74 teams · 7 variables · PC1 + PC2 carry 49.9% of the variance · furthest team Heaney, Hyatt and Schmeler, 2.64 from the centre
Engineering(22)
Sales(16)
Support(14)
Design(10)
Ops(12)
variable loading
loading of 1
Point at a mark, or tab into the cloud and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
These two components carry 49.9% of the variance between them: the other 50.1% is off this plane. Two teams that look close here may be far apart in the data.
Arrows are drawn at 2.52× their loadings so both layers fit one plane; the ticks stay in PC1 units. One factor for every arrow, so directions and relative lengths are exactly as computed — only the absolute length is arbitrary.
Employee survey · PCA on the correlation matrix— variable loadings. An arrow's direction is what it means: two variables pointing the same way move together, at right angles they are unrelated, opposite ways they trade off. Length is how well this plane represents that variable.
Variable
PC1
PC2
Length
Direction
Manager support
0.81
0.22
0.84
right
Career growth
0.72
0.44
0.84
up and right
Team belonging
0.66
0.55
0.86
up and right
Workload
-0.64
0.51
0.82
up and left
Meeting load
-0.55
-0.42
0.69
down and left
Tooling
0.31
-0.79
0.85
down
Pay fairness
0.58
-0.63
0.86
down and right
Employee survey · PCA on the correlation matrix — the 6 teams furthest from the centre, which is the mean of the projection. The remaining 68 are reachable by tabbing into the cloud and using the arrow keys.
Team
Group
PC1
PC2
Distance from centre
Heaney, Hyatt and Schmeler
Sales
0.62
-2.57
2.64
Romaguera LLC
Engineering
2.63
0
2.63
Stokes, Hirthe and Boehm
Engineering
1.79
1.91
2.62
Paucek - Pouros
Support
-2.44
-0.91
2.6
Parker - Zboncak
Engineering
1.36
2.17
2.56
Streich-Kub and Sons
Sales
2.1
-1.41
2.53
marks="loadings" · the same survey as a pure loading plot. With no scores to share the plane the arrows keep their own units, so the ticks read in loading units and the dashed circle sits at exactly 1. The 74 teams in the payload are not drawn, and the card says so.
Employee survey · PCA on the correlation matrix
7 variables · PC1 + PC2 carry 49.9% of the variance
variable loading
loading of 1
Point at a mark, or tab into the arrows and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
These two components carry 49.9% of the variance between them: the other 50.1% is off this plane. Two teams that look close here may be far apart in the data.
Loadings only, so the ticks read in loading units and nothing is rescaled. The 74 teams in this payload are not drawn.
Employee survey · PCA on the correlation matrix— variable loadings. An arrow's direction is what it means: two variables pointing the same way move together, at right angles they are unrelated, opposite ways they trade off. Length is how well this plane represents that variable.
Variable
PC1
PC2
Length
Direction
Manager support
0.81
0.22
0.839
right
Career growth
0.72
0.44
0.844
up and right
Team belonging
0.66
0.55
0.859
up and right
Workload
-0.64
0.51
0.818
up and left
Meeting load
-0.55
-0.42
0.692
down and left
Tooling
0.31
-0.79
0.849
down
Pay fairness
0.58
-0.63
0.856
down and right
Density knobs · arrowFill=0.55 pulls every arrow back to just over half the reach of the furthest workspace, which keeps the arrow labels clear of the Enterprise cluster; pointRadius=3 and labelledScores=12 name twelve outliers instead of six. Only the absolute arrow length changed — directions and relative lengths are the same numbers as in the first card, because one factor scales them all.
Workspace feature usage · PCA
96 workspaces · 7 variables · PC1 + PC2 carry 68% of the variance · furthest workspace McKenzie Group, 3.56 from the centre
Free(34)
Team(28)
Business(22)
Enterprise(12)
variable loading
Point at a mark, or tab into the cloud and use the arrow keys, for the row behind it. Click or press Enter to pin one; Escape releases it.
Arrows are drawn at 2.16× their loadings so both layers fit one plane; the ticks stay in PC1 units. One factor for every arrow, so directions and relative lengths are exactly as computed — only the absolute length is arbitrary.
Workspace feature usage · PCA— variable loadings. An arrow's direction is what it means: two variables pointing the same way move together, at right angles they are unrelated, opposite ways they trade off. Length is how well this plane represents that variable.
Variable
PC1
PC2
Length
Direction
Seats
0.88
0.21
0.9
right
Storage used
0.71
0.34
0.79
up and right
Sessions / seat
0.55
0.61
0.82
up and right
Exports / week
0.34
0.74
0.81
up and right
Tickets / seat
-0.48
0.52
0.71
up and left
Integrations
0.62
-0.55
0.83
down and right
API calls / week
0.79
-0.42
0.89
down and right
Workspace feature usage · PCA — the 12 workspaces furthest from the centre, which is the mean of the projection. The remaining 84 are reachable by tabbing into the cloud and using the arrow keys.
Zyeon UI