BESTINVER is the independent fund and retirement management company belonging to ACCIONA Group. With over 35 years of experience and 6 billion euros in assets under management, we work with the aim of generating long-term returns for our more than 46,000 investors.

Currently, BESTINVER has a customer knowledge and business intelligence model, although we are constantly searching for more innovative models that allow us to continue increasing investor satisfaction and improving their experience through a relationship model more tailored to their needs

Investor relationship model

Today, our investor relationship model is based on providing support throughout their lifecycle. To achieve this, we have an Investor Relations team responsible for staying in contact with them and keeping them informed about the progress of their investments.

Additionally, we have a service matrix that defines the minimum number of contacts an investor will receive annually based on their characteristics.

After each interaction, the Investor Relations team records the most relevant information discussed during the conversation with the investor in our Investor Management tool (CRM). This ensures that we have a record of the entire history of conversations held with the investor.

BESTINVER aims to increase customer satisfaction levels and enhance the quality of service provided, taking into account investor needs based on our interactions with them. This will primarily help us to:

• Improve satisfaction levels by adapting the service matrix based on investor needs (e.g., increasing or decreasing the service level if requested during conversations).
• Optimize the investor relationship model, leading to better utilization of our Investor Relations team by making more pertinent and relevant calls.
• Detect anomalies and discrepancies in information collection to increase quality and comply with standards regarding the recording of marketing information when applicable.

At BESTINVER, we have a customer management tool, CRM Dynamics 365, where we store each interaction our Investor Relations team has with our investors. The information from these interactions contains unstructured text data that is currently not being fully utilized for defining the investor relationship model.

To address this challenge, we will work with the written notes taken by the Investor Relations team after each conversation held in a call or meeting with the investor. This will also include CRM activities such as tasks that may be of interest.

Additionally, we will consider the already collected and categorized information, such as the outcomes of past actions.

BESTINVER presents this challenge to identify end-to-end tools and solutions that facilitate the processing and analysis of unstructured information through the application of natural language processing (NLP) techniques. The proposed solutions should align with the following scopes:

• Provide a pre-developed “as a service” product for the treatment of unstructured texts, incorporating predefined NLP algorithms such as key topic extraction, term frequency analysis, named entity recognition, sentiment analysis, etc.
• Collaborate with BESTINVER’s Technology department and ACCIONA’s Data Protection team to validate the conceptualization and proper development of the proof of concept: security, architecture, information ingestion, etc. Both departments will be part of the decision-making committee for selecting the startup.
• Have experience in implementing this product with a developed and tested methodology in the financial/insurance services sector.
• Proactively identify unidentified use cases or new approaches to information analysis that may be of interest.
• Provide a solution that offers a knowledge base/taxonomy tailored to BESTINVER’s business and needs, allowing for precise categorization and classification of these interactions/investors. For this classification, consideration should be given not only to the processed interaction information but also to the client’s historical information.
• Minimize custom developments and the need for significant subsequent technological maintenance. BESTINVER, however, considers establishing a future collaboration model to ensure basic maintenance of the solution.
• The intellectual property of the proposed solution will belong to BESTINVER, taking into account the startup’s contribution to the solution and respecting the intellectual property and possible usage license of the startup’s own software. This will be regulated in the Participation Agreement to be signed with the winning company of the challenge.

How we want to solve it 

We want to approach the resolution of the problem from the angle of analyzing unstructured information through natural language processing (NLP) with the aim of extracting relevant information from these interactions. To tackle this challenge, we believe it’s crucial to apply a previously defined and tested methodology by the startup. The proposed solution will include the following:

• Indicators with identified concepts: It will transform unstructured information into structured data in the form of indicators based on pre-identified concepts. Starting with a list of interactions for a specific investor, ideally giving more weight to recent interactions, we aim to generate standardized indicators based on concepts we believe to be relevant:
  • Investor satisfaction status: Investor satisfied/dissatisfied with returns/concerns about the market, etc.
  • Quality of service provided and frequency of contact: We need to either reinforce or decrease contact with a specific investor, for example, if they understand the philosophy of the long-term product they have and are content.
  • Investor experience: Investor with a high level of experience/no expertise.
• Analysis of term relevance: In addition to identified indicators and concepts of interest, applying algorithms to surface important and frequent terms, potentially leading to new indicators (e.g., TF-IDF or similar).
• Exploratory environment: The solution must address at least two use cases to determine its scalability in a potential production deployment.
• We believe a possible solution to the problem could involve having a database of unstructured information for storing interactions and indexing, such as ElasticSearch or a similar tool, for subsequent analysis and exploration. This would make key and relevant concepts easily accessible through searches in the indexed information. However, any other approach that addresses the problem will be equally valid and considered of interest.