Assess with Rubrics. A Successful Proposal within the PBL
1. Revisión de la literatura
The thesis researches how to create marketing personas with machine learning. Once the clusters of user behaviour data are created with machine learning, new algorithm needs to create marketing personas from the clusters. The persona creation is divided into two (see Section 7.5) where the algorithm needs to reconstruct the data into generic and deep analysis presentations.
The Generic Marketing Persona has a template that has general text written to it and some visualizations (see Section 7.5.1). Each of the placeholders are pre-determined and la-belled for each filter used in the Program Finder. The algorithm’s responsibility is to fill
the visualization and blanks within the text. The placeholders are filled with the data that comes from the clusters created.
The Data Behind the Persona is more simplified when referenced with the Generic Mar-keting Persona algorithms, since it only needs to visualize the filter weights in a xy-coor-dinator as bar charts. Each vector tree is divided into its own coorxy-coor-dinator and data is transformed into bar charts.
The process for creating marketing personas starts by importing the data from the cluster to the algorithm. When the algorithm receives the data, it starts traversing the vector trees and gathering the data. While traversing the vector trees, the algorithm creates vector variables based on the data found. The vector variables have three sets of data:
• name of the filter
• weight of the filter
• level of the filter
The variable created cannot include only the name of the filter because the generic per-sona uses only the first three levels of the vector structure and the bar charts of the Data Behind the Persona need to contain the information of the weight.
While traversing the vector tree, the algorithm places the data to the generic persona’s template and to the bar charts of the Data Behind the Persona. This way the algorithm doesn’t need to first pass through the vector structure and then go through the data and place them to the needed positions, hence increasing the process time.
For the Generic Marketing Persona, the algorithm is used for the first two levels of each vector tree. The process starts by checking whether both levels are necessary to be written to the template. The decision is based on how close the weight of the first and second levels are, in other words, how similar two user trends are to each other. If the first level dominates the weight of the second level, it can be concluded that for the Generic Mar-keting Persona, it isn’t necessary to mix the presentation of the marMar-keting persona by including multiple filter options to the template. The idea for the Generic Marketing Per-sona is to be easily understandable and fast to adopt. If too many filter options are added, it does not fulfil these criteria because of its complexity. If though the first and second level are close to each other with a ratio of 60/40, it can be concluded that they are both important information to be included to the template since no clear differentiation can be made from which is the dominant user behaviour trend.
The algorithm continues by matching the name of the data imported to the label that is in the Generic Marketing Persona template and adds the name of the filter to the template.
When all the vector trees are traversed, the template should contain the general text that existed at the start of the process and be enriched with the text provided by the algorithm (for example, see Section 7.5.1).
For the Data Behind the Persona, the algorithm uses the name and the weight of the filter. It creates a xy-coordinate for each vector tree inside the cluster and start remodel-ling the data to a bar chart presentation. The x-coordinate is used to present all the branches of the vector tree and the y-coordinate for the weight of the branches. When the algorithm has traversed through the vector trees, the presentation should include bar charts for each filter.
The pseudo code for the data to marketing persona is presented below.
Algorithm Generic Persona Template Check Vector Variable (Name)
Check Generic Persona Template (Placeholder)
If Vector Variable (Name) Equals Generic Persona Template (Placeholder) Add Vector Variable (Name) to Generic Persona Template (Placeholder) Algorithm Generic Persona
If Vector Variable is Level 1 or Level 2
If Level 1 / Level 2 is 60/40 or closer to 50/50
Add Level 1 and Level 2 to Generic Persona Template If Level 1 / Level 2 is 61/39 or further to 100/0
Add Level 1 to Generic Persona Template Algorithm Data Behind the Persona
For Each Vector Tree Create xy-coordinate Traverse Vector Tree For Each Branch
Create Bar Chart Algorithm Data to Marketing Persona
Import Cluster Data
Split Cluster to Vector Trees Traverse Vector Trees
For Each Branch
Create Vector Variable Activate Generic Persona
Activate Data Behind the Persona
Marketing Personas for Educational Program Finder
As is explained in Section 7.4.2, each program Aalto EE offers has its own cluster and the marketing persona is created based on the vector trees inside the cluster. If multiple clusters are found of the same program, multiple marketing personas are created. Since
the marketing personas need to be real time to fit the needs of today’s markets, they should evolve each time data is added to a cluster. This way we can ensure that the marketing and sales using the personas have updated data to support their decision-making. The marketing personas are divided into two, Generic Marketing Persona and Data Behind the Persona.
Generic Marketing Persona (see Section 7.5.1) is a simplified version that has a clear structure and human-like attributes for the readers to understand and embrace the infor-mation within it. Generic Marketing Persona can be used by sales and marketing to un-derstand the target groups. The persona can additionally be presented for a broader crowd because of its simplicity and visualizations.
Data Behind the Persona (see Section 7.5.2) is used for a deeper analysis of the pro-gram’s possible target marketing group. It is a presentation of each filter in its own bar chart to represent the division between the filter options and a broader presentation of what filters have the users used. Data Behind the Persona can be used to, for instance support a new marketing strategy or re-target which customer segment to target.