Client confidential

Identifying information have been pixelated
Data communication system for a confidential medical product. The data rendered involved a complex set of variables and performs predictive statistics, based on live and historical measurement. 

This is an analog prototype of how such a system would behave when fed live data.

Starting point


  • page for each subject 
  • graph for each metric, based on time 
  • same visual representation for each metric 
  • vertical slicing by timestamp



Opportunities


  • overlaying metrics to understand anomalies 
  • drawing cross–subject conclusions 
  • looking beyond event–specific peaks in metrics


Step 1

Initial Ideas


  • moving away from time based, 2D graphs 
  • experimenting with overlapping cards 
  • developing the concept of a data card, and the subsequent user “signature”


Possible Premises


  • each point in time, for each subject, for each metric gets a card
  • cards can be thematically linked
  • each metric gets a shape assigned to it, depending on the number of variables it constitutes 
  • a collection of linked cards could be a subject’s signature

Graphic representation of thematically grouped graphs:

  • 2 vars : 2 lines, normal graph 
  • 3 vars : triangle 
  • 4 vars : rectangle 
  • results can be a point, line or an area (shape)

Mapping Axis


  • can we simplify our graph schemas and collate similar metrics? 
  • our unique timestamp approach (for each card) creates an opportunity for simpler visuals

Drawing Logic


  • points are drawn at the intersection of axis, or with the edge of the shape, whatever is first 
  • even in a triangle the results could be rendered as a shape if there is no intersection point 
  • calculating intersection points is done from 
top to bottom 
  • an axis can only be used once to render a point

Next Steps:


  • choose an option for our graph schema
  • decide on the fidelity of spectrum analysis 
  • consider noise within the realistic range 
  • prototype a full signature for a positive, noisy and negative states

Reiterating the approach


  • We’re listening to our subject’s metrics based on realistic values and their signature 
  • we don’t know what we’re looking for, nor when will 
it come 
  • we know that a negative reaction will be multi–dimensional and “loud” 
  • that allows us to focus on core ranges, and listen to large anomalies


Step 2

Assumptions


  • Each timestamp gets a data card
  • time moves in the z axis 
  • a collection of overlapped cards makes the user signature for that timestamp

Rules


  • starts at 12 o'clock
  • moves clockwise
  • when lines meet they create an orange dot, and their line disappear
  • orphan lines create a dot when they hit the bounding box
  • 2 dots create a line, 3 a triangle and so forth

Initial Sketches




Further Refinements 


A mock-up made with
stationary data