Section Two Analytics Assignment
Section Two
Section two of this assignment has required me use a GameStop product review dataset. The requirement was to complete a topic modeling analysis of this data-set. This is to be done using the LDA algorithm under the Gibbs sampling method. After the visualisations are completed I will then assess the quality of these topics both visually and numerically. After all of this I will then give analytically backed recommendations to GameStop based off of our results.
Topic 1
Our first graph showcases Topic 1. Topic one seems to be based around different television specs and terms, with “tv” being the largest bar. However, the real insight from this is the next three bars. The second largest bar is “picture”, this tells us that people are interested in the “picture” of a tv. The phrase”picture of a tv” would usually be used by people who aren’t entirely tech savvy to refer to how clear/HD a television is.
The third largest bar is “price”, which tells us that people are extremely interested in the price of tv’s. This would be useful to the company as they can change their marketing efforts to work around drop drops/discounts as consumers are very impacted by price.
The fourth largest bar is “issue”, this tells us that consumers run into issues with their televisions quite often. The company could use this information to offer additional assistance, warranties and guarantees to consumers when purchasing televisions
Topic 2
Topic 2 seems to be based around consumers that have recently purchased and were satisfied with their experience or purchase. However, there is not enough detail for us to make any clear or useful observations.
Topic 3
Topic 3 seems to be based around different types of batteries on offer. Not only this but it seems to have some interesting information about the use cases the consumers have for these batteries. For example we can see that “flashlight” shows up towards the bottom of the bar-chart. Although on first thought we might presume that because its lower on the visualization that it might not be useful, this is not the case. The company now has an opportunity to push the use-case of specific types of batteries towards consumers. Example being a sign in store stating”Great for Flashlights”.
Otherwise this graph gives us a lot of insight into what consumers are looking for in their batteries. They’re searching for long lasting, powerful, “alkaline” “batteries”. It also tells us that consumers consider brand on their customer purchase journey.
Topic 4
Topic 4 seems to be based around what consumers are looking for in a video game. This is interesting data for both the store but also for the game developers. It means that the store can categorize their shelving units by what the most popular want the consumers have. Example being that Game Stop might give video games that are categorized as “fun” and have an open “world” setting the most shelf space.
Topic 5
Topic 5 seems to be based around consumer opinion of different video games. However, there is not enough detail for us to make any clear or useful observations.
Topic 6
Topic 6 seems to be based around the controllers from the various video game consoles. We can see from this graph that people seem to be quite the fan of the “xbox” “controller”, with no real negative feelings towards this. This would tell GameStop what controller might sell best, it would also be incredibly useful to Xbox’s competitors such as PlayStation, knowing that the public seem to rate their competitors controller more.
Topic 7
Topic 7 seems to be based around the video game “Zelda”. It tells us a lot about how the consumer feels about the video game. Most notably, consumers seem to talk about the “gameplay” and the “past” a lot. This would tell us that consumers feel nostalgic while playing Zelda, thinking about the older versions of the video game they would’ve played when they were younger. Something that might start this nostalgia is the continued use of the same “music” and the continuation of the character “Link” within the video game.
Topic 8
Topic 8 initially seems to be based around the video game franchise “Fallout”. This is further backed up by the fourth and fifth largest bars “3” and “4” as they were the most popular “Fallout” video games over the years. However, this visualization does not have enough information to make any clear observations.
Topic 9
Topic 9 seems to be based around what people value about their video games. From the graph we can tell that people are interested in the “story”, telling us that a video game needs to captivate the user. They also want good “characters” in the video game, something or someone they can connect with. This is interesting for the company as they can categorize these variables and wants that the consumer has and bucket video games together based on these variables. And then use these buckets of video game titles to market to specific groups of consumers that share these values.
Topic 10
Topic 10 seems be to roughly based on time. However, there is not enough detail for us to make any clear or useful observations.
Topic 11
Topic 11 seems to be based around headsets/headphones. The graph below tells us what consumers are looking for when it comes to headsets/headphones. There are several bars that standout to me, “quality”, “hours” and “comfortable”. This tells us that consumers want a long battery life on their headsets/headphones. We can presume that if they want a long battery life, they want to wear them for long periods of time so they would want them to be “comfortable”. Finally the word “quality” stands out, and this is to be expected as every consumer wants value for money especially when it comes to “quality” and headphones.
Topic 12
Topic 12 seems to be based around the Nintendo Switch Console. This graph gives us the interpretation that the product is enjoyed by “families” as it’s “easy” to use. It seems to be a highly recommended product. This tells us that the Nintendo Switch is performing very well in terms of customer satisfaction and presumably sales with some consumers claiming to “love” the product.
Topic 13
Topic 13 seems to be based around monitors. Expensive monitors are commonly purchased by video game users who’s needs are based around high definition, quality and value. This graph gives us some standout insights into monitors. Firstly, the brand “Samsung” stands out as it is clearly the best brand that these consumers have found. We then look at what consumers are using these monitors for. With “gaming” being the clear use case, however we can see that some consumers are also using these monitors for “movies”. Both of these use cases seem like activities that would demand quality, this is shown as “quality” is the sixth largest bar in this visualization.
Topic 14
Topic 14 seems to be based around the Pokemon games. This graph gives us a lot of insight into the consumers thoughts about the Pokemon franchise as there are so many Pokemon games. For example this graph tells us that consumers are talking most about Pokemon “black” and Pokemon “white”, with Pokemon “black” being the more popular game. This graph also tells us that consumers are really enjoying the “Pokemon” games with “graphics” and “awesome” showing up on the visualization.
Quality
Looking at the quality of the topics produced by the LDA algorithm we can make comment on several items.
If we look at topic_size we can see that the numbers are relatively balanced, meaning that the algorithm split the tokens as evenly as it could. However, topic 3 stands out as the lowest topic_size however it has the largest mean_token_lenght. Which would tell us that this has the lowest number of tokens, however the tokens that it does have are substantially more detailed and longer than the others. While topic 8 is the largest topic_size, with the lowest mean_token_lenght. Telling us it has the most tokens, however there is very little detail. If we look back at the graph from Topic 8 we can see that it was hard to make any observations due to the lack of detail. While Topic 3 had a lot of observations due to the level of detail involved.
When looking at topic_coherence, the closer the number is to 0 the better the quality is as topic_coherence refers to the overlapping of data. We can see that Topic 3 has the best topic coherence with -113.6211. While Topic 10 has the worst topic coherence with -206.2124. This makes sense as we had mentioned previously due to the topic size and mean topic length, Topic 3 was of high quality, this is further backed up by topic coherence. While in Topic 10, there was no real quality to make any observation.
Topic 3’s quality is further shown through topic exclusivity as it has produced a 9.994323. When thinking of topic exclusivity, the higher the number the better quality the top bar on the visualization is, making it easier to interpret the topic. While Topic 9 is the lowest quality in terms of topic exclusivity, this is seen in the visualization of Topic 9 as the phrase “games” is the number one spot. This term is very open to interpretation as it is not very specific making it very difficult to interpret.
| topic_num | topic_size | mean_token_length | dist_from_corpus | tf_df_dist | doc_prominence | topic_coherence | topic_exclusivity |
|---|---|---|---|---|---|---|---|
| 1 | 727.4474 | 4.9 | 0.6471285 | 4.334033 | 46 | -165.0276 | 9.913542 |
| 2 | 660.0322 | 4.9 | 0.6609999 | 1.486317 | 18 | -205.4634 | 9.987487 |
| 3 | 553.0684 | 6.7 | 0.6674659 | 8.213726 | 154 | -113.6211 | 9.994323 |
| 4 | 633.3016 | 5.3 | 0.6278735 | 12.131437 | 8 | -175.5083 | 9.920989 |
| 5 | 662.2048 | 5.7 | 0.6126989 | 12.168554 | 13 | -144.3217 | 9.772076 |
| 6 | 703.1038 | 5.3 | 0.6360770 | 2.623424 | 29 | -183.4483 | 9.878324 |
| 7 | 702.2969 | 5.0 | 0.5984157 | 12.507642 | 29 | -158.7826 | 9.818651 |
| 8 | 740.2119 | 4.0 | 0.6295982 | 3.345680 | 14 | -199.3155 | 9.936862 |
| 9 | 708.4366 | 6.0 | 0.6198190 | 12.497857 | 19 | -158.0331 | 9.467051 |
| 10 | 715.6828 | 4.2 | 0.6398966 | 2.001048 | 12 | -206.2124 | 9.974860 |
| 11 | 765.7622 | 5.8 | 0.6481332 | 2.195597 | 38 | -198.2749 | 9.907865 |
| 12 | 657.2549 | 5.8 | 0.6566324 | 2.348319 | 11 | -203.7851 | 9.965962 |
| 13 | 701.0964 | 6.8 | 0.6585347 | 3.337024 | 45 | -168.6369 | 9.934888 |
| 14 | 672.1000 | 6.2 | 0.6083859 | 14.760148 | 69 | -134.5991 | 9.809813 |
Summary
Although I have given my interpretation of each specific visualization, in order to conclude and summarize I’m going to bullet point actions I would suggest to GameStop to take based on the data.
Collaborate with gaming developers
- Leverage the customer feedback about what customers value in terms of video games
Set up a customer service line for televisions
Knowing consumers have issues with televisions, set up a phone line that will attempt to help troubleshoot any issues the consumer is having.
If unfix-able have the customer service line help the consumer go through the warranties of their products
Begin to strategically categorise video games
- Knowing what consumers enjoy about types of video games gives GameStop the opportunity to re-arrange their store to make the customers in store experience better.
Change promotion of the Nintendo Switch
- Leverage GameStops customer insight knowledge of the Nintendo, to run a family first promotion. Use recommendations as “reviews” to promote the product to famalies, not just their kids.