Power BI vs. Cuddle — NLP

Sahil Verma
Crux Intelligence
Published in
5 min readOct 26, 2019

Overview

NLP is a hot cake in the market right now, everybody wants to get their hands on it! Used the right way, NLP can be a very powerful tool for users to get things done quickly and efficiently.

NOTE: This comparison is done using the latest versions of both platforms as of Oct 2019.

Methodology

We compared both platforms across different parameters which are considered to be the building blocks of NLP for an enterprise BI platform.

The process was carried out by ingesting the same dataset in both Power BI and Cuddle to make the comparison fair, and documenting the results across each of the selected parameters for NLP assessment for :

  1. Disambiguation
  2. Custom Time Range
  3. Date Range
  4. Top / Bottom Questions
  5. Conditional Questions
  6. Target Achievement

Let’s find out how Power BI holds up against Cuddle’s NLP capability.

Comparison

1. Disambiguation

Question: “Daily sales this year”

Cuddle Interpretation

“Daily sales this year”

  1. Metric name was disambiguated to “Unit Sales” based on Cuddle’s learnings from user interactions
  2. User had the ability to pick any other measure which had sales in it using a dropdown button
  3. Appropriate visualization presented in the form of a line chart
  4. Users also have the ability to change the time period (this year, last quarter, this month, etc) and interval (daily, weekly, etc) from the visualization widget itself.

Power BI Interpretation

“Daily sales this year”

  1. Measure name had to be entered exactly as it is.
    Power BI did NOT use any kind of keyword disambiguation.
  2. When the measure name was entered exactly as it was in the dataset, the desired line chart was presented.
  3. Users did NOT have the ability to change interval the time period (this year, last quarter, this month, etc) and interval (daily, weekly, etc) from the visualization widget.

Disambiguation: Power BI 👎 | Cuddle 👍

2. Custom Time Range

Cuddle Interpretation

“Sales starting 10th week of 2019”

  • Cuddle interpreted the question correctly and presented an appropriate answer.

Power BI Interpretation

“Sales starting 10th week of 2019”

  • Power BI could only interpret Sales and 2019, and therefore did not return the expected answer.

Custom Time Range: Power BI 👎 | Cuddle 👍

3. Date Range

Cuddle Interpretation

“Dollar Sales from Jan 1 to Feb 15 in 2019 by brand”

  • Interpreted measure, attribute and time period in the question correctly.

Power BI Interpretation

“Dollar Sales from Jan 1 to Feb 15 in 2019 by brand”

  • Interpreted measure, attribute and time period in the question correctly.

Date Range: Power BI 👍 | Cuddle 👍

4. Top / Bottom Questions

Cuddle Interpretation

“Who is the best Sales Rep”

  • Cuddle interprets “best” as Top 1.
  • Cuddle presented an answer which has top sales reps for different Metrics.
    Example, `Top Sales rep in terms of Unit Sales`, `Top sales rep in terms of Dollar Sales Achievement` and so on..

Power BI Interpretation

Who is the best Sales Rep”

  • Power BI could not interpret the question when we used the word “best”.
  • But Power BI understands Top 1, Top 5, Bottom 3, and so on..

Top / Bottom: Power BI 👍 | Cuddle 👍

5. Conditional Questions

Cuddle Interpretation

“Cities with <98% achievement in 2018 and 2019”

  1. Cuddle interprets the conditional part of the question.
  2. Presents all cities for 2018 & 2019 which have achievement score above 98%.

Power BI Interpretation

“Cities with <98% achievement in 2018 and 2019”

  • Power BI could not interpret the question.

Conditional Questions: Power BI 👎 | Cuddle 👍

6. Target Achievement

Cuddle Interpretation

“Sales vs target by reps”

  1. Cuddle disambiguates “sales” to Dollar Sales
  2. It interprets the relationship between Dollar Sales and Dollar Sales Target to present an appropriate target visualization.
  3. A user also has the ability to view the target visualization in the form of a trend line.

Power BI Interpretation

“Sales vs target by reps”

  • Power BI presents a scatter plot for the query which is hard to understand and doesn’t make sense for a user in this case.

Target Achievement: Power BI 👎 | Cuddle 👍

Conclusion

This is how the results shape up for each of the selected parameters:

Results

Cuddle supports all major foundational aspects of NLP in analytics while Power BI still has a lot to catch up on!

Want to know more about Cuddle?

Reach out to us via email — sales@cuddle.ai

Visit our website — Cuddle.ai/ask

--

--