We build the models, benchmarks and positioning frameworks behind early-stage investment and strategy decisions. Every figure arrives with the source it came from and the assumption it rests on.
Four pieces of work, each built to be handed over and rerun by someone else.
A proprietary scoring and sentiment fusion model applied across 50,000+ startup data points. FRED and yFinance feed it directly, so the signals stay tied to macro conditions instead of floating free of them.
A machine learning pipeline built on more than 100 economic indicators, returning startup valuation, revenue forecasts and financial, operational and risk metrics. It runs end to end from a single company input, with no manual step between the query and the output.
A 2x2 positioning matrix paired with weighted feature scoring, applied across 20+ comparable companies. It locates the whitespace worth moving into and shows where a defensibility claim does not hold up.
A benchmarking template built on 15+ precedent transaction comparables, structured to produce multiple ranges that hold up line by line in an early-stage review rather than a single headline number.
Excel and Python builds are handed over whole, with the workbook, the assumptions and the code that produced the output.
Power BI, Tableau or an embedded web view, for work that has to be revisited each month rather than read once.
The reasoning laid out step by step, so a reviewer can follow how the conclusion was reached and challenge it where they disagree.
Noesis Research is an independent research practice working with founders, investors and advisory teams.
We take on the analysis that has to survive a second reading. Valuation ranges that get questioned in a committee. Positioning claims that get tested by a competitor. Forecasts that get compared, a year later, against what actually happened.
Most research of this kind arrives as a conclusion with the workings stripped out. We do the opposite. Sources, assumptions and model structure ship with every engagement, so the work can be checked, rerun and extended without us in the room.