RadarCompare

Compare entities across multiple dimensions

Data Editor

	Color	Entities / Dimensions	Formality of Design	Interface Density	Customizability	Enterprise Orientation	Self-contained Solution	Asset Centricity
		Samsara	4	4	4	5	5	5
		Motive	4	4	4	5	5	4
		Geotab	3	5	5	5	2	4
		Verizon Connect	4	4	3	4	5	3
		Netradyne	3	3	2	3	2	2
		Lytx	3	4	2	3	1	2
		Nauto	2	2	1	3	1	2

VisualizationTweak

Similarity Analysis

	Samsara	Motive	Geotab	Verizon Connect	Netradyne	Lytx	Nauto
Samsara	-	99.8%	97.3%	99.2%	98.4%	95.4%	97.0%
Motive	99.8%	-	97.3%	99.6%	98.7%	95.7%	96.9%
Geotab	97.3%	97.3%	-	95.7%	97.9%	97.9%	97.0%
Verizon Connect	99.2%	99.6%	95.7%	-	98.7%	95.6%	95.9%
Netradyne	98.4%	98.7%	97.9%	98.7%	-	98.8%	98.4%
Lytx	95.4%	95.7%	97.9%	95.6%	98.8%	-	97.7%
Nauto	97.0%	96.9%	97.0%	95.9%	98.4%	97.7%	-

High (≥80%)

Medium (60-79%)

Low (<60%)

How are similarity scores calculated?

RadarCompare calculates similarity between entities using two different methods:

1. Cosine Similarity

Measures the angle between two vectors in multi-dimensional space. It focuses on the pattern of ratings rather than their magnitude.

similarity = (A·B) / (||A|| × ||B||) × 100%

Where A·B is the dot product of vectors A and B, and ||A|| is the magnitude of vector A.

2. Spearman Rank Correlation

Measures the monotonic relationship between two entities by comparing their rank orders. It captures whether entities tend to rank dimensions similarly, regardless of the actual values.

ρ = 1 - (6 × Σd²) / (n × (n² - 1)) × 100%

Where d is the difference between ranks for each dimension, and n is the number of dimensions.

3. Ordinal Distance (Gower)

Calculates the average absolute difference between scores after scaling them to the 1-5 range. It is well-suited for ordinal values.

distance = Σ|A_i - B_i| / (n × 4)

Similarity is expressed as 1 − distance, then converted to a percentage.

Interpretation:

• 80-100%: High similarity - entities have very similar patterns
• 60-79%: Moderate similarity - some common patterns
• 0-59%: Low similarity - different patterns