Blog Post: Market Analytics and Trends Analysis in ToyLabs

Market Trends and Social Feedback Analysis Methodology in steps

The ToyLabs methodology “Market Trends and Social Feedback Analysis” refers to the identification of market trends as these may emerge through an analysis of the online activity of users (i.e. prospective customers) as well as the efficient management and analysis of the feedback that arises for a specific company through the user mentions of the brand or its products/services in the web.

The ToyLabs methodology “Market Trends and Social Feedback Analysis” refers to the identification of market trends as these may emerge through an analysis of the online activity of users (i.e. prospective customers) as well as the efficient management and analysis of the feedback that arises for a specific company through the user mentions of the brand or its products/services in the web.

Therefore, two different flows can be recognised and should be handled by this methodology and these are as follows:

  1. One flow for feedback management and analysis of a selected brand/product/competitor/etc.
  2. One flow for trend detection and analysis for a specific domain of interest (e.g. dolls, puzzles, etc.).

Broader data sources and larger stakeholder pools are thus being considered as relevant with the toy industry and the design of new toys and are being integrated into the proposed methodology in an attempt to on the one hand gain insights into the expected adoption of a new toy design by a somewhat uninvolved crowd, i.e. a diverse crowd that was not specifically questioned to provide opinions on specific ideas, as is the case for the End Customers engaged in the ToyLabs Open Innovation platform, and on the other hand early identify any issues or customers disapproval/dissatisfaction for their current products.

Broader data sources and larger stakeholder pools are thus being considered as relevant with the toy industry and the design of new toys

In that view, six core steps are being identified (as seen in the figure above), that have a direct correlation with the steps and processes identified from the landscape analysis, as well as one preparatory step, step 0, that is a step performed outside the platform among the involved stakeholders to configure the analysis to follow to their specific needs and their industry’s specific characteristics.

Step 0: Preparatory step

As already mentioned, this step is responsible for analysing the needs of the toy manufacturer in question and customise/adapt the analysis that follows according to the industry/product type he represents

Step 1: Data Collection

Having made all the necessary configurations and decisions on the appropriate key-phrases to initiate the search, the sources to consider etc., this step is responsible for data retrieval and storage. In this context, data collection is based on the settings of the previous step where identified search keywords, known user accounts of the domain and preselected blogs are determining the data to be retrieved.

Step 2: Data pre-processing step

The purpose of this step is to homogenise the diverse input types coming from different social media platforms, blogs, portals etc. giving as output all gathered data in the form of tuples that carry the following attributes:

  • The core textual part of each main piece of data
  • The stakeholder, i.e. the opinion holder
  • The time that each piece of data was published
  • The source of the retrieved data (e.g. Facebook, Twitter)
  • The importance indicator: This is an additional indicator specifically employed to measure the text importance and influential power of the given text extracted from source-specific information that will be excluded from the next processing step.

Step 3: Data processing step

Data processing step is responsible for converting the text that expresses an opinion, an idea and consecutively a potential trend in an appropriate representation for the analysis that will follow in the next steps. Towards this end, linguistic processing has to be employed, where a number of NLP techniques should be applied in order to transform unstructured text into vectors.

Step 4a: 1st-level analysis/ Topic Detection step

Topic detection, together with the trend detection, which is the next step, are the two most important steps in a trend detection methodology. They can either be seen separately and be implemented one after the other or they can be implemented in parallel as one step, and this is the case in many cases where topics that are directly traced are popular topics, i.e. trends (e.g. grouping of popular keywords into themes).

Step 4b: 2nd-level analysis/ Trend detection step

For the trend detection step in the context of ToyLabs, only the two out of the three methodological flows identified and described in the literature analysis that preceded will be examined, namely either the co-existence of the topic and the trend detection step or their treatment separately where topic detection precedes. The first flow where topic detection step is surpassed will not be considered here since its processes are more complicated without providing proven better results.

Step 5: 3rd-level analysis/ Sentiment analysis step

The last step of the methodology deals with the application of sentiment analysis on the selected trending topics. Sentiment analysis is not common in trend analysis approaches. However, in the context of ToyLabs, a differentiation from the standard approaches has been decided, providing an innovative approach, where the selected trending topics are also analysed regarding the sentiment they encapsulate, to identify the ones that are being discussed in a positive way (while also capturing the intensity of the sentiment). In this manner, this step allows to detect the trends that are worth considering for potential future toy designs. In cases where the methodology presented is being utilised to acquire feedback for a brand and its product, this step is even more a substantial one, in the sense that concluding to an aggregated sentiment for a “topic” that will represent a specific product, a brand or anything else which the toy manufacturer in question wants to monitor will actually provide the manufacturer with an easily digestible insight on his customers’ perception about the selected subject.

Conclusion

The opportunity is being given to the system’s user to discover new, previously hidden insights both about trends in the toy industry and their customers’ feeling on their products and brand to denature both information into new, innovative, customer-centric toy designs.

Therefore, the current methodology targets directly a toy manufacturing design team and it is important that actionable insights are being given to them. Generic conclusions about crowd satisfaction or dissatisfaction or indications about general trending topics in the form of plain, unconnected words may be non-exploitable by them, or “painful” to transform them into useful knowledge. For that reason, specific attention should be given on ways to offer combined insights into the data that were retrieved, processed, and deduced during the previous steps.