As the adoption of sustainable investments is on the rise, ESG data and the means to harness it, has become a must-have for ESG investors and index providers. In the quest of gathering quality and comparable data, investors and index providers are seeking integration of machine learning into the data collection process. However, the state of automation is not fully matured in this area due to various limitations. Nevertheless, there are certain possibilities for small automations. Individually these automations can bring small efficiencies however collectively and on a large scale it can lead to substantial savings in costs and improves data quality.
The cost of gathering and analyzing ESG data is a major barrier to using ESG information for investment decision-making. Therefore, the industry is looking out for solutions that can make ESG research more efficient, affordable, and scalable. Many companies are focusing on machine learning to capture the most structured reported numbers. However, as of now, even the ones who are claiming to have developed machine learning, still spend lot of manual time in processing ESG data. Nevertheless, if successfully implemented this could make the research affordable for all.
There are certain challenges in developing automation tools for ESG data:
1. Unstructured/Incomplete data:
Unstructured or incomplete data that are often incomparable across firms, industries, and sectors. Unstructured data include texts, pictures, multimedia content, online reports/presentations, etc. Collating data of such different types without manual work is currently not possible. Moreover, the data is also often found incomplete for automated collection and analysis.
2. Irregular reporting by companies:
Different companies follow different cycles of reporting including annual reporting, biennial reporting, triennial reporting as well as irregular reporting of quantitative data. This makes extraction and storage of data in automated matrix difficult.
3. Complexity of Data:
Data is often not reported in units and metrics required for analysis. This compels some level of manual work in analysis. Therefore, complete automation is difficult. For example, as Health & safety KPI, some companies report LTIFR, some report loss time injuries, some report total injuries, etc. To enable comparison across such companies, certain assumptions and calculations must be made manually.
4. Qualitative Analysis:
A significant part of the analysis is dependent on the qualitative data available. With multiple keywords for a particular aspect, automation becomes very difficult as it would not only involve extracting all the information with the relevant keywords but also sorting through this information to derive a conclusion.
5.Reporting based on different reporting instruments:
There are more than 400 mandatory and voluntary reporting instruments, and automating comparison of ESG performance of companies, that follow different instruments, is difficult.
Nevertheless, as sustainability reporting matures, the ease of collecting and analyzing ESG data will improve. This will also enable sophisticated solutions for automation of collecting and analyzing ESG data. Till we reach there, there are certain possibilities for small automations. Individually these automations can bring small efficiencies however collectively and on a large scale it can lead to substantial savings in costs and improves data quality. For example:
- Automated trend calculation for various sets of quantitative indicators.
- Automated graphical presentation of KPIs.
- Automated data conversion from one unit to another.
- Automated normalization of data to make it comparable over the years and with other companies.
- Automated linking of quantitative data on two different metrics to calculate ratios, percentages etc.
- Automated comparison of company’s performance with policy goals like the SDGs to understand the progress towards them and to identify areas that need focus in order to avoid current/future regulatory risks.
Studies show that the next stage in ESG research process would be to study the causal correlation on how previously distinct strands of ESG data and analysis relate to or reinforce performance in each other. With the above automations in place at this point of time, implementing the emerging stage would also be easier in the future.
At Sustainometric, we focus on understanding clients’ current research processes to make it lean, efficient, cost effective, and scalable. To know more about us, please contact firstname.lastname@example.org