Customer input to improve your operation

Processo decisionale basato sui dati per le aziende

Ted Kwartler

Data Dude

Examine customer survey data

$$

• Data collection: qualitative interviews, online reviews, transactional data
• Use model to explain customer survey data

Survey inputs as a model

• Q1: How satisfied are you overall [1-5]?
• Q2: How do you rate the quality of the product/service [1-5]?
• Q3: How do you rate the product/service options [1-5]?
• Q4: Do you agree that the offering is fairly priced [1-5]?

	Q1 (overall satisfaction)	Q2	Q3	Q4
Customer-1	1	5	5	5
Customer-2	1	4	5	5
Customer-3	0	1	3	1
Customer-N	1	1	4	4
...	...	...	...	...

Survey inputs as a model

	Q1 (overall satisfaction)	Q2	Q3	Q4
Customer-1	1	5	5	5
Customer-2	1	4	5	5
Customer-3	0	1	3	1
Customer-N	1	1	4	4
...	...	...	...	...

$$

• Target variable
  • Q1
• Explanatory variables
  • Q2, Q3, Q4

Explanatory models from customer data

	Q1 (overall satisfaction)	Q2	Q3	Q4
Customer-1	1	5	5	5
Customer-2	1	4	5	5
Customer-3	0	1	3	1
Customer-N	1	1	4	4
...	...	...	...	...

$$

Logistic regression model

$f(\text{overall satisfaction}) = \beta_1 * Q2 + \beta_2 * Q3 + \beta_3 * Q4$

Sum of betas to understand impact

Logistic regression model

$f(\text{overall satisfaction}) = \beta_1 * Q2 + \beta_2 * Q3 + \beta_3 * Q4$

Model output

$f(\text{overall satisfaction}) = 0.25 * Q2 + 0.25 * Q3 + 1 * Q4$

Sum of betas

	Beta	Sum of beta	Proportion
Q2	.25	0.6	.42
Q3	.25	0.6	.42
Q4	.1	0.6	.16

Adding context with frequency

$$

	Beta	Sum of beta	Proportion
Q2	.25	0.6	.42
Q3	.25	0.6	.42
Q4	.1	0.6	.16

Adding context with frequency

Adding the context of how often the organization does well in a category

	Beta	Sum of beta	Proportion	Frequency of a high score
Q2	.25	0.6	.42	.8
Q3	.25	0.6	.42	.35
Q4	.1	0.6	.16	.15

Off to it!

Processo decisionale basato sui dati per le aziende