Activity Capture - Documentation and User Guide

1. Introduction to Activity Capture

Historically, gathering activity data across an organization presented the greatest challenge and time investment in carbon accounting. Traditional workflows depended heavily on extensive spreadsheets, constant email exchanges among colleagues, and manual data entry into software, which heightened the probability of human errors.

To address this issue, Position Green developed Activity Capture, driven by a vision to eliminate manual data collection. The project was launched to solve a fundamental question: how to leverage and capture existing activity data automatically without manual intervention.

Invoices emerged as the most optimal resource since they already reside within corporate systems and specify precisely what was bought and how much was purchased. By extracting this information, Activity Capture significantly enhances data accuracy within the GHG inventory. This improvement is particularly impactful for category 3.1 Purchased Goods and Services, a segment that has traditionally depended on imprecise, spend-based approximations.

Here is how Activity Capture works:

Invoice or receipt line items interpreted using AI

Identified activity data

Matching of activity data with emission factor

Data validation by user

Added to correct scope categories in the GHG inventory

Ultimately, Activity Capture enables organizations to transition from broad spend-based estimates to granular, activity-based emissions data by capturing line-item details from invoices and matching purchases with relevant emission factors.

2. When to Use Activity Capture

Organizations experiencing a shortage of actionable data regarding their purchased goods will find Activity Capture highly beneficial. While its primary application targets Scope 3, Category 1, the solution also effectively scales to support Scope 1, Scope 2, and additional Scope 3 categories. It delivers maximum value to businesses engaged in product- or material-heavy procurement. Optimal outcomes are typically achieved when dealing with commodities that feature a significant volume of Environmental Product Declarations (EPDs), such as IT hardware, construction materials, and office furniture.

Conversely, the platform offers limited suitability for service-heavy procurement fields such as maintenance and consulting. A further constraint arises from the quality of the invoices themselves; missing or vague product descriptions yield sub-optimal results from Activity Capture. To mitigate this limitation, companies can proactively collaborate with their suppliers to enhance and refine their invoicing structures.

Suitable use cases:

• Enriching Scope 3.1 Purchased goods and services

• Product-heavy procurement

• Procurement within EPD-rich sectors

• Invoices with clear line item descriptions

Less suitable use cases

• Vague or missing invoice descriptions

• Service spend with no clear unit or product reference

To the right, is a fictitious invoice to exemplify the ideal information needed for optimal Activity Capture results:

• Line item description / name of product

• Quantity

• Unit

• Invoice date

• Price (can be useful for extrapolations)

• Currency

• Supplier name

• Invoice number

• Product code, if available

3. Prerequisites 

To be able to use Activity Capture, there are a few prerequisites to be aware of.

Willingness to share invoices

Activity Capture might not be the most appropriate choice for organizations that regard their invoice details as highly confidential. In such cases, one viable alternative is to scrub or sanitize the invoice data prior to sending it to Position Green, thereby removing any sensitive elements. Additionally, if only a partial collection of invoices is submitted, the customer assumes full responsibility for ensuring the overall completeness of the final results.

Ability to ingest invoices

There are three primary methods available for ingesting invoices into our platform:

1. Send data to our API: Suited for organizations utilizing data warehouses, data lakes, or centralized invoice management systems. This option gives customers direct control over which specific invoices are uploaded, though they remain responsible for ensuring data completeness.

2. Supported ERP Integrations: Position Green can directly extract files if you use a compatible ERP integration that supports Activity Capture. Please note that some ERP providers restrict invoice file access, preventing ingestion. If an integration for your specific ERP or invoice management system is not yet available, we can scope a dedicated setup project.

3. Manual PDF Uploads: Ideal for businesses without a supported integration or those preferring not to establish automated transfers. This approach requires manually exporting your invoices from your ERP or handling system and uploading them directly to the Position Green platform.

For additional details regarding API and integration methods, please refer to our integration guidance.

4. Reading and Matching Invoice Lines to Emission Factors

Activity Capture utilizes AI to automatically parse and interpret the individual line items within ingested invoices. Following this initial extraction process, the system systematically reviews each line item to match it with a corresponding emission factor.

To ensure accuracy and consistency, this matching process occurs automatically before displaying any results to the user. The system operates sequentially, evaluation line item by line item, and applies a predefined data quality hierarchy to determine the most appropriate emission factor match.

Line item parsing and interpretation

Our system processes invoices using an AI model tasked with reading and converting each individual line item into a product name. The accuracy of this line item extraction relies on the capabilities of the chosen AI model. Typically, more advanced models provide superior interpretation compared to lower-cost alternatives, leading to more accurate pairing with emission factors. 

Data quality hierarchy

Activity Capture initially prioritizes matching invoice line items with Environmental Product Declarations (EPDs). Sourced from EcoPlatform, an international database, these manufacturer-generated emission factors are highly credible and product-specific.

If an EPD is unavailable, the system proceeds to match items against average activity-based emission factors from our curated library. Learn more about our emission factor library here (link).

As a final resort when verified activity-based factors are missing, the system defaults to our proprietary EEIO model for spend-based estimations. This is generally applied to service procurement or entries with incomplete data. Read more about the EEIO model here.

Confidence scores

When matching line items with emission factors, the system assigns a confidence score to each pairing based on two criteria:

1. Certainty of the interpreted name: Have we correctly identified the exact item or service purchased?

2. Certainty of the matched emission factor: Have we paired the item with the most appropriate emission factor?

Based on these evaluations, the final results are categorized and color-coded as high confidence matches (green), uncertain matches (yellow), or unlikely matches (red).

5. User validation and improving results

After data ingestion and processing via Activity Capture, users can begin validating the results. Validation can be performed either in the “Data” tab immediately following ingestion or inside specific collections within the “GHG emissions” view.

Recommended Data Validation Workflow

To streamline this process, we advise focusing first on high-uncertainty data. Below is a structured workflow designed to optimize validation efficiency:

1. Isolate low-quality alignments by filtering for unlikely and uncertain matches.

2. Target high-impact entries by sorting material rows by costs and emissions from highest to lowest.

3. Examine the original source invoices linked to individual line items.

4. Execute bulk adjustments to emission factors by grouping entries, searching for alternate EPDs or average activity-based factors, or utilizing the EEIO spend-based calculation model.

5. Modify scope category placements in bulk across multiple selections.

6. Exclude irrelevant or undesired lines from inclusion in your final inventory.

7. Flag specific rows for subsequent follow-up (on roadmap)

8. Attach explanatory comments to document any assumptions made for individual entries.

9. Reverse any accidental adjustments using the undo function (on roadmap)

10. Approve lines either individually or in bulk.

11. Dispatch validated lines directly to designated collections from the validation view.

Reuse of previous matches

The system saves user modifications, including adjustments to emission factors or scope categories. Consequently, during subsequent data ingestions, identical or similar records are automatically mapped to the user-defined locations. To provide clear visibility, the system highlights these remembered adjustments in blue.

Analyzing results

After completing the validation process, users can analyze various insights generated by Activity Capture, including:

• Identifying products or suppliers with high impact

• Examining recommendations for improvement and overall match rates

• Assessing the decarbonization potential of specific purchases (on roadmap)

• Gaining insights into pricing (on roadmap)

• Exporting data to Excel

6. Traceability and audibility

As line items are ingested into the platform, the original invoice is automatically attached as supporting documentation for both the interpreted product name and the designated emission factor. Users can easily review the emission factor documentation simply by clicking on the matched factor itself. Additionally, a comprehensive White Paper detailing the full methodology behind the EEIO spend-based emission calculation is available. You can access it via this link.

7. Potential for improvement and future development

As Activity Capture is a newly introduced capability at Position Green, ongoing enhancements are being made to maximize its performance. Through our EEIO framework, we leverage spend-based approximations from day one to guarantee nearly 100% data coverage. Currently, the rate of successfully matching invoice items to activity-based metrics ranges between 30% and 50%, varying with the depth of details provided in the invoices. We are actively exploring and developing several key initiatives to advance this capability further.

Optimizing Invoice Data Quality

Enhancing the granularity and completeness of invoice data, as detailed in Section X (link), represents one of our highest-leverage growth areas. We strongly recommend that users collaborate with their suppliers to upgrade the quality of information included in the invoices they receive.

Refining Processing Algorithms

Upgrading the internal logic responsible for scanning line items and aligning them with emission profiles can boost match performance by an estimated 15-20%. Furthermore, because the system learns from historical manual inputs, the underlying algorithm continuously adjusts and sharpens based on user feedback.

Evaluating Alternative AI Architectures

Performance metrics shift significantly based on the specific AI architectures deployed. Lower-cost models tend to cast a wider net for activity-based matching but compromise on precision, whereas premium models deliver superior precision across a narrower subset of items. Structuring a balanced approach to model utilization is expected to elevate both match accuracy and overall volume.

Expanding the Emission Factor Library (EFL)

A primary bottleneck stems from the current volume of activity-specific parameters in the EFL. Growing the database - particularly with supplier-specific or product-specific emission factors - will significantly enhance the match rate.

Using AI-Generated Emission Factors

We are investigating techniques for creating emission factors using AI in a way that align with formal validation and verification rules. Currently, these methodologies lack widespread endorsement from formal standards groups like the Science Based Target initiative. Nevertheless, regulatory and auditing frameworks are shifting swiftly, suggesting these techniques could become viable when supported by adequate verification tracks. Implementing this would unlock comprehensive activity-based alignment alongside a detailed data infrastructure.

Expanding the Use Case Beyond Invoices

Although Activity Capture is presently tailored for receipts and invoices, this approach can be applied across various file formats to extract activity information from any type of document.

Appendix 1: Key Concepts and Descriptions