From Emissions History to Future Scenarios
Knowing what emissions had been did not automatically tell teams what they were becoming.
As Cognite’s sole North American product designer, I worked on industrial-data experiences where equipment, operations and historical information needed to support real decisions. One client project explored emissions tracking and forecasting using more than a decade of history.
The design challenge was connecting patterns, current conditions, anomalies and possible futures without implying more certainty than the evidence could support.
CLIENT WORK · DESIGN EXPLORATION / PROTOTYPING
Past, present and future were different kinds of evidence.
- Historical Measured
- What actually happened?Emissions and operating conditions recorded over time.
- Current Measured / recent
- What is happening now?Recent measurements, deviations and operating context.
- Forecast Model-derived
- What may happen next?A trajectory suggested by a model under stated assumptions.
- Simulation Changed assumptions
- What if conditions change?A deliberately constructed scenario, not an expected outcome.
Generalized analytical model—not recovered client UI, a dataset or a validated forecasting method. Labels and line treatments distinguish evidence types without relying on color alone.
Measured, forecast and simulated values should never look interchangeable.
More history did not necessarily create more understanding.
Ten-plus years offered rich analytical potential, but also enormous time ranges, changing equipment and processes, recurring patterns, anomalies and uneven information.
The question was not just what the whole chart looked like. It was which part of history mattered to the condition being investigated now.
Emissions needed operating context.
Asset state, production activity, maintenance, process changes, baselines and reporting periods could help users investigate a change.
Contextualized data was valuable because people could examine relationships—not merely view more sources together.
Observe → Compare → Forecast → Explore
- ObserveRead actual historical measurements.
- CompareChoose a relevant baseline or operating period.
- ForecastInspect a model-derived trajectory and its assumptions.
- ExploreChange an assumption deliberately and compare a scenario.
A forecast answers “what may happen.” A simulation asks “what if.”
Time is part of the information architecture.
- DecadeStructural change
- YearPatterns and shifts
- Month / periodUnusual behavior
- EventSurrounding evidence
Reconstructed navigation model: trend → period → event → evidence, rather than every historical resolution at once.
Move from a change to a question worth investigating.
- Notice an unexpected value or trend.
- Compare with history: unusual, seasonal, recurring or structural?
- Narrow the relevant period.
- Inspect asset, process and operating context.
- Distinguish an isolated anomaly from a persistent shift.
- Review the forecast and its assumptions, where available.
- Explore a clearly labeled changed-assumption scenario.
- Carry the evidence into a human operational decision.
Generalized workflow reconstruction, not a claim that every step was implemented. Correlation may suggest where to investigate; it does not establish cause.
Simulation as a thinking tool
CONCEPT · SCENARIO EXPLORATION
Baseline
Inspect the current conditions and assumptions used by the model.
Changed scenario
Change an assumption, recalculate where technically supported, compare the result and investigate the difference.
The value is testing sensitivity and relationships, not producing another authoritative-looking number. A scenario remains conditional on its data and assumptions.
No client dataset, forecast formula or emissions simulator is presented as operationally validated here.
Show what we knew, estimated and were testing.
- Keep measurements distinct. Historical observations are evidence, not forecasts.
- Separate forecast from scenario. A model trajectory and a user-changed assumption have different meanings.
- Move between overview and evidence. Start with a trend; inspect the event when needed.
- Keep context around anomalies. Connect the number to conditions without asserting causality.
- Avoid false precision. A precise display is not proof of a certain model.
Evidence → decision → tradeoff
Evidence: more than a decade of emissions history created both analytical opportunity and complexity.
Decision: distinguish observation, comparison, forecast and scenario exploration.
Tradeoff: explicit evidence types add interface complexity.
Response: layer disclosure around the operational question, with deeper analytical evidence available when needed.
A precise number is not automatically a certain number.
The analytical model stays the source of truth.
Original exploration
Customer / domain problem → industrial-data investigation → workflow and analytical framing → design exploration / prototyping → customer and engineering collaboration.
This lifecycle framing will be tightened against recovered artifacts. It does not imply verified production implementation.
How I would prototype it today
Model the historical dataset → use AI to explore hypotheses → code analytical states and scenario behavior → test understanding, not just layout.
A proposed current approach, not a claim that an AI forecasting system or this interactive reconstruction has already been built.
Reality check · what design could and could not improve
What design could improve
Historical navigation, comparison, anomaly visibility, relationship exploration, scenario understanding and communication across customer, product and engineering.
What it could not guarantee
Model accuracy, complete history, sensor quality, causal relationships, future operations or the validity of a simulation’s assumptions.
The interface could make uncertainty legible. It could not remove it from the underlying system.
CLIENT EXPLORATION / PROTOTYPING · No measured post-launch emissions reduction, compliance result or forecast-accuracy outcome is claimed.
From historical data to decision exploration.
- Establish what was measured.
- Find the relevant comparison.
- Add operating context.
- Separate observation from inference.
- Inspect the forecast carefully.
- Make changed assumptions explicit.
- Return to what the evidence supports.
Useful anywhere people reason from historical evidence toward an uncertain future.