Speak to an Expert
This methodology sets out how Integrum ESG’s Golden Retriever uses our proprietary AI models to find, extract and cite bespoke data points for any list of public or private companies, from the sources a client’s team would normally research by hand. Every data point comes with its source citation and a confidence level, is reviewed in our Assurance Portal and exports to Excel in one step.
Purpose
Why Golden Retriever exists
Research and data teams spend a large share of their time collecting data by hand: opening reports, websites and databases company by company, finding the right figure and recording where it came from. Golden Retriever automates that collection for any set of data points a team defines, so analysts can spend their time on analysis and verification instead.
It applies the same proprietary, ESG-trained AI models that extract and cite company disclosures for the Integrum ESG Platform, where every metric is traceable to its original source. Golden Retriever turns those models on a client’s own companies, sources and data points.
Method
How Golden Retriever works
- Define the project The client provides three lists: the companies to research, public or private; the sources its team would normally use for manual research, including public and password-protected sites; and the data points it needs.
- Ingest and extract Our AI models gather the relevant sources for each company and process, organise and curate them into a structured knowledge base. Each data point is then extracted alongside a citation to the source it came from.
- Assure Every extract is surfaced in our Assurance Portal with its source material and a confidence level, so the client’s analysts can review the results and make any changes directly.
- Export Selecting Complete generates an Excel file with every data point, whether or not it has been amended.
Assurance
Assurance and confidence levels
Our systems currently surface data points and citations with accuracy of more than 90%. Every extract appears in the Assurance Portal with:
The result
The value or answer extracted for each company and data point.
The source
The material the result was extracted from, so it can be checked in context.
A confidence level
Assigned to each extract, flagging straight away the results our systems consider may need further human review.
Direct editing
A simple interface in which the client’s analysts can correct or complete any result before export.
Confidence levels let a team focus its review where it matters, rather than checking every data point by hand. The approach mirrors the Glass Box behind every score on the Integrum ESG Platform, described in our Company ESG Scores and Glass Box Methodology.
Worked example
Worked example
A team needs 25 data points for each of 200 companies, a mix of listed and private, drawn from company reports, company websites and a subscription database it already uses.
- Input: the list of 200 companies, the three source types and the 25 data points.
- Output: 5,000 extracts (200 × 25), each with its result, its source citation and a confidence level.
- Review: the team works through the extracts flagged for human review in the Assurance Portal, amending any that need it.
- Delivery: one click on Complete produces the full data set in Excel.
Uses
Who uses Golden Retriever
Asset managers and owners
Collect data points their ESG data providers do not cover, for their own research, screening or client reporting.
Consultancies and advisers
Run large-scale data collection for client projects at lower cost and in less time.
Private markets
Gather data on private companies from the sources a team already relies on. For data submitted directly by portfolio companies, see our Private Markets ESG Data Capture Methodology.
Research and data teams
Replace repetitive manual collection with a reviewed, cited data set, freeing analysts for deeper analysis and verification.
FAQs
Frequently asked questions
What is Integrum ESG’s Golden Retriever?
Golden Retriever is a data extraction solution that uses Integrum ESG’s proprietary AI models to find, extract and cite bespoke data points for a client’s chosen companies, from the sources the client identifies.
What does a client need to provide?
Three lists: the companies to research, the sources its team would normally use for manual research and the data points it needs.
Can it collect data on private companies?
Yes. Golden Retriever works on public and private companies, using the public or password-protected sources the client identifies.
How accurate is it?
Our systems currently surface data points and citations with accuracy of more than 90%. Every extract carries a confidence level, and those that may need human review are flagged in the Assurance Portal.
Can analysts change the results?
Yes. Analysts can amend any result directly in the Assurance Portal before export.
How is the data delivered?
Selecting Complete in the Assurance Portal generates an Excel file with every data point and its source.
Scope your data project
Speak to our team about the companies, sources and data points you need, and see how Golden Retriever would collect them.