2. What specs dominate
sentiment?
Who are our
target reviewers?
When to release
the product?
Release Strategy
Python
SAS Enterprise
Miner
SAS Enterprise
Guide
SAS Visual
Analytics
SAS JMP Pro
TOOLS USED
1 2 3
Release Strategy
What Who When
3. DATA PREPARATION
Exploratory
Data Analysis
(Data
Manipulation
and Cleaning)
Sentiment
Analysis
(Using TextBlob)
Stratified
Sampling
Topic
Extraction
Python SAS EM/TM SAS EM/TM
(Phone vs Tablet)
Transform
Data for
Research
Questions
Dominant Topic Extraction
Principal Component Analysis
SAS VA Dashboard to drill down
Release Strategy
What Who When
4. Review text
Topics about specs
Dissatisfaction Satisfaction
1. What device specifications
should be present in my next
product offering?
a. What are the specs that lead to satisfaction and
dissatisfaction?
b. What specifications exist in particular product models
that are more/less popular?
c. Are there any trends in the release of
specifications?(predictive)
Companies
Models, Actual Specs and Trends
Filtering
Dominant topics: → Hand picked, Specs related
device_release_year → 2009-2014 (2015 is stopped halfway)
Grouping
Product Type: Phone and Tablet
Selection
Polarity: Dissatisfied, Neutral, Satisfied, Very Satisfied
Release Strategy
What Who When
5. Sentiments Phone Tablets Models, Specs and Comparison
Dissatisfaction
T
O
P
I
C
S
1.+screen,+touch,+work,+touch screen,+big
2.+camera,+front,mp,+quality,+front camera
3.battery,+life,battery life,+backup,+battery
backup
1.+screen,+touch,+work,+touch screen,+big
2.+battery,+charge,+day,+hour,+last
3.+battery,+charge,+day,+hour,+last
B
R
A
N
D
1. Micromax
2. LG
3. Lenovo
1. Lenovo
2. Samsung
3. Apple
Satisfaction
T
O
P
I
C
S
4.+camera,front,mp,+quality,+front camera
5.+memory,+card,internal,+gb,apps
6.battery,+life,battery life,+backup,+battery
backup
4.+battery,+charge,+day,+hour,+last
5.+screen,+touch,+work,+touch screen,+big
6.battery,+life,battery life,+backup,+battery
backup
B
R
A
N
D
1. LG
2. Micromax
3. Huawei
1. Lenovo
2. Samsung
3. Apple
Mixed Reviews
Screens, Camera, and Battery are the most relevant
specifications to focus on for the next product release
Release Strategy
What Who When
6. Brands
Reviewers
Subjectivity / Posted Time
2. Who are the most suitable
reviewers to employ as part of
my test/marketing strategy?
a. Which brands’ reviewers am I interested in?
b. Which reviewers are objective in their assessment? What’s
their occurrence pattern like?
c. What are the specific brands a particular reviewer has
reviewed and what are the models he/she has reviewed?
d. What can be deduced from the nature of the reviews? (is
there a skewed sentiment/polarity; are the actual reviews
informative?)
e. How does the polarity of each brand reviewed vary over
the specified posted time period?
Polarity over time (Grouped
by Brand Name)
Choice of Reviewer
Filtering
Subjectivity (Informational)
Grouping
Brand Name
Selection
Polarity, Brand Name,
Subjectivity, Posted Time
Polarity
Hierarchy
User Name -> Brand Name
->Model Name
Categories
Body, Model (list table) ,
Polarity, Username,
Posted Name
Release Strategy
What Who When
7. Subjectivity Level Satisfaction Level Time Period
0-0.5 All Ranges 01-Jan-2010 – 31 Jul 2015
Invite them for focus group discussions
FINDINGS Tcool* Egyptian* Sam* Zyper25 Raj
Posting Frequency 109 80 69 62 49
Top Reviewed Brand
/ # Reviews
Huawei
92
Lenovo
80
Microsoft
17
Lenovo
62
Microsoft
17
No of models reviewed 5 1 12 1 15
Average Subjectivity 0.14687 0.15637 0.10723 0.18542 0.08192
ASSESSMENT
• They all appear to be relative informative given the average subjectivity.
• We could engage these two to be part of a focus group discussion to test and feedback on products before their release.
• These groups can include both positive and negative reviewers (as long as they show interest in the product
Release Strategy
What Who When
8. 3. When is it best time to
release a new device into a
market?
a. What are the observable trends in sentiment levels for
reviewers?
b. Is there a time when tablets are more popular than phones?
c. Is there a time when big players dominate?
Filtering
Subjectivity ->
(Informational)
Near Release Date reviews
→ Range of release date
and review post date < 3
months (90 days)
Grouping
Polarity (pie chart)
Categories
body, model (list table)
Posted time (line chart)
Release Date (line chart
Selection
Polarity : Dissatisfied, Neutral,
Satisfied, Very Satisfied (pie
chart selection)
Brand_Name (text input)
Subjectivity (slider)
Posted_Time (of
review) (slider)
Hierarchy
month --> Product Type -
>Brand Name → Model Name
Distribution of Review Satisfaction
Validate with Cyclical Model releases
Drill down to product review distribution
Companies which champion release times
Release Strategy
What Who When
9. November is a good time for Apple releases
May and November are the best
times for release
Tablets are more popular during May
and November by a ratio of 3:2
Lenovo, Samsung and Huawei do
well in May, whereas Apple
dominates in November
Release Strategy
What Who When
10. By answering the What, When, and Who questions, we can gain
powerful insights that grant us a competitive edge in our next
release.
Mixed Reviews for Samsung,
Lenovo and Apple (Phones and
Tablets)
Cameras are the cause for
satisfaction/dissatisfaction
among phone users
Battery is the cause for
satisfaction/dissatisfaction
among tablet users
Best times to release is
in May and November
Tablets are more popular during
May and November by a ratio of
3:2
Lenovo, Samsung and Huawei do
well in May, whereas Apple
dominates in November
Users such as Tcool and Egyptian’s
engagement can be traced to
single brands
Top reviewers inspire objective
assessment
Brand polarity over time fluctuates
for top users
Thank you for your time!
Editor's Notes
Good morning everyone, my name is Vincent Tatan and this is Ranon Sim and Smeet Malvania. I’m the leader of the triSAS team and we are going to present to you about our report
Limitation:
Anon, anony, anonymous. Etc
TODO:
Timeline tracker
Principal component Analysis Image
The When part:
Whether we should have planned our release date (will help reconsider to companies diff release strategy)
Plan what release date gives optimum (count reviews, count positive reviews)
Release Strategy: Optimize positive reviews and sales by knowing what specs to target, who to reach out, and when to release our next release.
The followings are the the topics that we would like to cover today. First of all we would to advise the proper release strategy so that companies could get well received reviews in the future. And in order to answer this , we split the objective into 3 manageable questions:
What specs should I release for my next product offering?
Who: who to push new releases to? Or who are the reviewers that we could ask for more feedbacks?
When is the right season to release my product (i.e. when reviews are most well received (review counts, satisfaction counts)?
-> ->
Exploratory data analysis:
Filtering some columns that we find redundant.
Making sure that the month and year released are expressable in date time and quarter format
Afterwards, we also cooked up sentiment analysis which using python TextBlob library a part of python NLP natural language processing library that could determine the polarity and subjectivity of each row of body. We will explain further during our demo later on
Afterwards, we used this filtered data and sentiment anlysis to further analyse the topics using SAS EM and SAS TM. But before this analysis, we sampled the data using stratified sampling method based on the product type. So that we are able to differentiate our analysis into two market: tablets and phones. And we could get samples large enough to represent both product types
Finally, we did last transformation that includes Dominant topic extraction for each row of body and use SAS VA dashboard to drill down so that we could have more comprehensive analysis and insights.
So now, let us answer the “what” for our first question. What are the topics of specs satisfactions and dissatisfactions for both tablet and phone?
And we would like to find out more in the sense of the companies that are being talked the most for each category, the models, and the trends.
And the following at the right side, is also our methodology to come up with the answers. Which will be brought further during the demonstration
So these are the key findings. We actually come up with the top specs topics and the brands according to their sentiments and product type. You could also see the pictures that we used to answer these questions
who are genuinely interested in my product and who provide the most informative assessments of my product/products
This chart is meant to give a quick overview/insights on the behavior of the top 5 most informative reviewers according to posting frequency. For example, Tcool seems to be the most active reviewer, Egyption seems to be quite interested on a single Lenovo brand whilst Sam appears to have quite abit of engagement with multiple Microsoft products. We can of course delve deeper into things like the reviewers’ polarity trend across different products but this will be covered later.
All the aforementioned insights can be summarized in this slide. To understand the process and methodology that led up to these insights, Vincent will now conduct a demonstration using SAS VA !!!!!!!
So these are the conclusions that we received from combining all the insights. We would definitely love to share with you our findings during our poster demonstration. Thank you.