Designing trustworthy interactions with large language models
How can we integrate large language models in design effectively while addressing their limitations and ensuring trust?
In part 1, we shared our vision for design’s role within ESG, and explored how design is used to present ESG scores. In this article, we’ll be diving into ESG reports and metrics.
But first, a quick recap.
We believe that data should help people understand the world around them. In our view, the ESG domain ought to be:
Reports provide in-depth analysis across various ESG issues, over a given year. Some ESG reports exist within annual reports, others are produced as standalone PDFs, and others are presented as micro-sites or pages within a corporate website.
Following the new EU legislation, nearly 50,000 companies will soon be required to compile a comprehensive ESG report on their environmental and social impact. These reports, alongside other publicly available information, are used by scoring companies to inform their assessments and determine an ESG score.
We defined criteria that describe, from our design perspective, a good ESG report. This criteria came from our own experiences analyzing visual material and was informed by specific examples we encountered whilst reviewing ESG content.
Given that reporting topics can vary significantly between industries, we chose to use fashion and apparel as a consistent baseline for comparison. The fashion industry is particularly significant due to its:
We combed through five reports from this industry to understand how each company communicates ESG data, guided by the following questions:
Clarity
Engagement
Format
Accountability
Transparency
Branding
Data Visualization
None of the companies we examined performed well across all categories, and the quality of the visuals and content varied greatly. The reports we reviewed, though all from consumer-facing brands, struggled balancing detailed, transparent and accountable information, and presenting it in a way that is accessible and engaging to the general public. This feels like a missed opportunity, given that a joint study from McKinsey and NielsenIQ in 2023 (albeit for consumer staples) showed consumers shifting their spending toward products with ESG-related claims.
We concluded that ESG reports lack visual-communication guidance, data design expertise and inspiration as to how to connect this data with a wider audience. As such, we chose to reformat our original assessment criteria as a set of guiding principles:
Collectively, we hope these principles will guide others toward our vision of accessible, transparent, and comparable ESG reporting.
Metrics refer to the global issues under each category of ‘Environmental’, ‘Social’ and ‘Governance’, such as resources used, product safety and quality, or tax transparency. This information exists within ESG reports, or through other forms of disclosure.
Each industry weighs ESG issues differently, depending on which have more substantial impact. We revisited the fashion reports we’d previously analyzed, and looked for high impact metrics for this industry (product carbon footprint, labor management, and governing board composition) to understand how they were visualized.
The most consistent data presentation choice was a table. While tables can be useful and consistent ways to present data — especially for comparing values and looking up information — relying solely on this format misses the opportunity to highlight aspects such as trends over time, which are important for reviewing performance against future targets.
When companies did choose other forms, they frequently opted for a bar chart, pie chart or line chart. However, many of the charts we saw fell short of their potential due to factors such as poor chart choice or lack of visual hierarchy.
So how do we create effective data visualizations? We’ll share an example of our thought process when analyzing a specific chart.
There are almost infinite choices available when creating a data visualization. As a basic starting point, we ask ourselves:
We’ll apply this thought process to this small multiples bar chart from LVMH’s 2023 Social and Environmental Responsibility. It displays CO2 emissions over the past three years, broken down into the different business groups.
Chart choice depends on the data itself, the audience, the format and the intent. Every chart has benefits and drawbacks.
A pie chart could be used to illustrate CO2 emissions per business group. However, this form is ineffective for a large number of categories, and doesn’t allow for precise comparisons. For emissions split by scope, a stacked bar chart is ideal, but a line chart may be more effective for showing change over time.
In this example, we’ll continue with LVMH’s small multiples bar chart. It offers flexibility and clarity across the multiple data types, showing both change over time and contribution per business group.
Prioritization is crucial for creating focus, guiding design decisions, and determining what to include, remove, and emphasize. In the original chart, various data types appear to have equal priority, resulting in information competing for attention.
We established these top three priorities:
To maintain the report’s integrity and transparency, we’d suggest using an appendix for additional details and providing access to the data for those who want to explore further.
Even if someone only briefly glances at this chart, the total emissions for 2023 compared to previous years should be immediately clear.
It should be clear that this is a small multiples chart, and we should be able to quickly scan to establish the biggest contributors in 2023.
Scope breakdown values don’t follow the same formatting as the total annual values, and add visual clutter. We would recommend incorporating this information into bars themselves, and being selective about labelling exact values.
In addition, we applied a few tweaks to improve overall comprehension:
This is small thought experiment has been explored in isolation, but all data visualization design choices ought to support the broader narrative of an ESG report. Seemingly small visual adjustments in spacing or color can significantly shape the focus of each visualization and play a crucial role in providing clarity and enhancing the understanding of ESG data.
Ensuring that an ESG report is accessible, transparent, and comparable extends to every data visualization within the report and without adequate care, attention, and expertise, companies risk obscuring key insights or misrepresenting their data.
We’re hopeful that upcoming changes in ESG legislation will empower consumers to better navigate sustainable choices, but there’s still a long way to go before ESG data could be described as accessible, transparent or comparable. The importance of effective data design in achieving this mission cannot be overstated.
While regulatory bodies work to enhance reporting standards and ensure compliance, we as designers see our role as shaping the communication and visualisation standards that bring this data to life. As ESG reporting becoming more widespread and with more data becoming available, we see powerful potential in bringing our expertise to this space.
For an example of our previous work transforming sustainability data, explore the comprehensive investment platform created with Globalance.
We’re passionate about exploring this field further, and welcome the input from those working with sustainability data.
Designing trustworthy interactions with large language models
How can we integrate large language models in design effectively while addressing their limitations and ensuring trust?
The power of data design systems
How to get to better data-driven products at scale faster
AI Driven musical experience for Google
What are the possibilities of AI for creating interactive engagements with music?