Unstructured data is complex and mixed. You cannot map it to a predefined structure, like a data table or a relational database. But wait…what exactly is unstructured data?
Over 80% of business data is unstructured. It doesn’t fit in a classical relational data store. Unstructured data comes from call transcripts and knowledge articles. It comes from Word and PDF documents and all kinds of video, audio, and text files, webpages, medical records, social media, and survey responses. Transcripts of chat conversations are an example. They are a treasure trove of market intelligence if analyzed well.
What is the process of bringing this data into Salesforce?
- A new unstructured data pipeline in Data Cloud enables customers to select the unstructured data they want to bring in – such as help articles sitting in a knowledge management service, or call transcripts in an S3 bucket – and make those usable across the Einstein 1 Platform. When ingesting text documents, that data can be chunked into smaller fragments to ensure more precise operations against that data. And to make this usable in operations like semantic search, it can then be transformed and stored in the new Data Cloud Vector Database.
- The Data Cloud Vector Database combines structured and unstructured data seamlessly and transforms it into a numerical representation called a vector embedding.
What is a vector database and why is it different from other databases?
Vector databases are purpose-built to store massive arrays of numeric representations of data – including unstructured data. Once stored and indexed (made available for search) this vectorized data can be used for fast query and semantic search-based retrieval using advanced similarity algorithms.
And vector embedding is a numerical representation of data. It’s called an embedding because this often lengthy sequence of numbers is designed to encode the unique nature of that data so that it can be “embedded” in dimensional vector space. Once embedded, we can use various machine learning algorithms. They will help us understand how related different embeddings are based on their “closeness” in that space. Then, we can find the ones most similar to, for example, a question we are trying to answer.
What is Semantic Search?
Semantic search relies on the meaning or intent of a query. This goes beyond simply matching specific words in user queries (which we call “keyword search”). Semantic search transforms your analysis by presenting a holistic view of how stakeholders discuss any given topic, even if they used entirely different words.
Now that we have covered the basics, how can we analyze data in Tableau? We’ll look at structured and unstructured data.
Case Analytics as an example of how unstructured and structured data together offer powerful querying capabilities and visualization
Let’s take a hypothetical example. It’s a company, CrystalBlue, that maintains private homes and offices. It has operations throughout the US. CrystalBlue leverages Salesforce Service, Salesforce Data Cloud and Tableau to connect better with their customers.
With the visual power of Tableau, CrystalBlue can see the geographical distribution of cases in the past year related to “water damage”. Entering the term “water damage” in the search prompt will display all cases related to “water leakage,” “pipe damage,” “broken pipe,” or any other semantically related search term in the result list.
Why is this important?
We humans express ourselves in various ways using different words. But, we usually mean the same thing. As case names and titles are created by customers of CrystalBlue or their customer service agents, the only way to meaningfully analyze this data is by semantically structuring it.
The intensity of the color in the Tableau visualization above shows the semantic relevance of the cases related to the search term entered, in other words, the darker the color, the more semantically similar the cases are.
What happens under the hood?
After ingesting the Service Cloud cases and transcripts into Data Cloud, CrystalBlue can index them. It will then chunk and vectorize the data using Salesforce’s search index. CrystalBlue can then connect Tableau to Data Cloud using the built-in Salesforce Data Cloud connector. All Data Cloud objects are available as tables right away, enabling you to query and relate any unstructured and structured data in a blended data model.
What is the value for CrystalBlue?
CrystalBlue can now leverage this blended data model to uncover new similarities and trends. Each query now casts a wider net and returns a richer set of results so CrystalBlue can improve their service offerings and customer satisfaction like never before!
- Discover which postal codes are most affected by “water damage” and any related issues. Then, investigate any outliers or patterns to explore the root cause of similar cases (extreme weather, old pipes in buildings, etc.).
- Check if customers have repeatedly raised similar issues this year. If they have, reach out to them. Understanding their needs can help prevent future problems.
- Get insights into the duration of case resolution for similar cases. Use these insights to speed up case resolutions and educate the field service agents.
What other use cases leveraging unstructured and structured data provide insights customers cannot get easily otherwise?
Knowledge Articles Analytics
Service agents rely heavily on knowledge articles to share with customers to solve issues. Knowledge bases grow over time as new products and services evolve and others get deprecated. Companies try to keep their knowledge databases up-to-date. They want them to be relevant to speed up case resolution.
Let’s take a hypothetical customer Energy4You, a solar energy company.
Use case: Discover duplicates
- Find all knowledge articles around “low battery consumption” found in the description of the article. Plot on a graph to show views and similarity score to detect duplicates.
Action: Clean up the knowledge database. Remove duplicates and keep the most recent articles.
Resume Screening and Matching
Help recruiters sift through vast amounts of resumes and process applications faster.
Let’s take a hypothetical company AceRecruits, a global provider of recruitment services.
Use case: Match the best available talent to open positions in the most efficient way possible
- Analyze job applicants’ resumes for skills, education, experience, and achievements.
Action: Compare and rank the applicants based on their qualifications. Use semantic search. Consider their suitability for the position based on individual attributes (e.g. location, job experience). Match them with the job’s requirements and preferences. Identify top 10 individuals and invite them for an interview.
Contracts Analytics
In the pharmaceutical industry, drug development can be very costly and is often outsourced. Once a drug has been developed, companies like Sanofi, Roche, AstraZeneca, will often contract the drug manufacturing company for the commercial production. Because of the regulations inherent in the industry, the monitoring of compliance with the FDA regulatory requirements might be outsourced.
Let’s consider a hypothetical customer, Cure4You. It’s a global pharmaceutical company that works with contracts.
Use case: Regulatory Compliance
- Find all drug manufacturing contractors who have a FDA regulatory compliance clause in the contracts.
Action: Create a segment from Tableau to Data Cloud for these contractors and initiate a communication.
Customer Survey Analytics
Every company strives to make data-driven choices. They use data to find their product’s strengths and weaknesses. They also find the satisfaction and loyalty of customers. And they find areas for improvement or innovation.
Let’s consider a hypothetical customer, an airline called “Flight4You”.
Use case: Improve CSAT
- Find all extra services “Flight4You” offers. These include wheelchair support, extra legroom, and onboard entertainment. Find which ones customers value most and least.
Action: Make investment decisions to invest only in the most valued services. Disinvest in all poorly valued services.
Summary
Using Tableau and Salesforce Data Cloud, customers can unlock unprecedented value from their trapped data. They can find hidden insights and trends through semantic search to improve service and boost customer satisfaction like never before. Here are the four steps of the process:
- Ingest Any Data [Un(structured)]
- Chunk + Score via Data Cloud
- Connect to Data Cloud from Tableau
- Query and Visualize Your Data
Start your journey now with Data Cloud and Tableau:
- Salesforce Data Cloud
- Data Sources in Data Cloud
- Data Cloud and Tableau
- New Vector Database in Salesforce Data Cloud
- Unstructured Data Capabilities for Data Cloud
Written by
