IN SUMMARY
Life cycle assessment used to be an analytical exercise: commission a study, get a result, answer a question. Now organisations embed LCA into procurement, product design, compliance and reporting, which means it has to function as infrastructure: reproducible, explainable and able to withstand scrutiny. When that happens, outcomes depend less on the impact method and more on whether the underlying data reflects how products are actually made and sourced. This paper sets out why a deliberate primary data strategy, focused where it changes decisions, is what separates defensible results from a false sense of rigour.
Impact is concentrated. Around 20% of processes typically drive the majority of a product's footprint, so investing in primary data everywhere adds cost without leverage; targeting the few upstream, energy-intensive or highly variable processes is where results actually change.
Stable results can mislead. When a model's numbers don't move year on year, that often reflects unchanged assumptions and ageing data rather than a robust, accurate picture of reality.
Regulatory hotspots are a floor, not a strategy. Predefined hotspots reflect where impacts tend to sit across a market, not within a specific supply chain, so true hotspots shift when operating conditions differ.
From analysis to infrastructure
For most of its history, LCA was a tool for answering a specific question with a one-off study, where bespoke assumptions and one-off datasets were tolerable. That no longer fits how organisations use it. Regulation, customer requirements, investor scrutiny and internal decision-making now demand product-level data applied consistently across portfolios and updated over time. Infrastructure cannot tolerate the inconsistency an analytical study could; it must be reproducible and able to withstand audit. As LCA becomes infrastructure, the centre of gravity shifts away from calculation, since standards and impact methods are a stable foundation, and toward the quality, specificity and traceability of the underlying data, especially for the upstream processes that dominate impacts.
Why generic and modelled data mislead decisions
Secondary and AI-generated datasets are not wrong; teams need them to build complete models at scale. The problem is using them to support decisions that hinge on differences between processes, suppliers or operating conditions. By design, these datasets smooth variability and average performance, which makes them stable but poorly suited to how a specific site or supply chain actually behaves. Three failure modes recur: hotspots get mis-ranked because the data embeds geographies, technologies and conditions that may not match reality; false confidence builds because clean, consistent numbers look precise even when assumptions are never revisited; and meaningful differences between suppliers get hidden by averaging, which matters most where regulatory thresholds determine market access. Introducing process-specific data at the right points breaks these patterns: hotspots shift, risk profiles change, and the differences between options become visible.