Signal research
Metricssuccess signals defined
Define the signals
Research and interviews identified metrics such as IP presence and funding status for scoring.
Case Study · Competitive Intelligence
Testing a radar for diagnostic startups.
A global diagnostics leader wanted an early warning system to spot promising startups and disruptors in in vitro diagnostics. We designed the blueprint, co-developed a proof of concept model and tested which signals really predict a startup’s fate.
The Client
The client is a global leader in diagnostics, actively developing in vitro diagnostic products. It wanted a data tool to flag promising startups and disruptors early.
The Challenge
Many new entrants.
Spot the real disruptors.
Prescouter was asked to create the blueprint of an early warning system for in vitro diagnostics startups and co-develop a proof of concept model that scores companies on signals such as IP and funding.
-
Signals
Pick the metrics
IP, funding and other indicators.
-
Data
Find the sources
Startup databases as a base dataset.
-
Proof
Test the model
Does the score predict outcomes?
Our Approach
Testing the signals with real outcome data.
Regression analysis
Oddsratios for each metric
Measure each signal
Logistic regressions estimated how each metric linked to acquisition or closure.
ML model
2startup databases tested
Test the model
An ML model combined the scores and was tested against outcomes from two startup databases.
The Outcome
Learning what does not predict success.
Prescouter generated odds ratios for each metric and tested them with descriptive and logistic regressions and an ML model. The key metrics did not prove to be good predictors of company success, with accuracy below 50% on data from two startup databases.
-
Blueprint
System blueprint
Parameters, benchmarks and data sources.
-
Testing
Odds ratios per metric
From regression and machine learning.
-
Finding
Below 50% accuracy
Common metrics proved weak predictors.
Popular startup signals such as funding and patents do not always predict success. Testing them against real outcomes before scaling a tool avoids false confidence.
Let’s Talk
Which signals really predict the next disruptor?
Define the signals, test them against outcomes and build on what holds up. Wherever you’re starting from, we’ll help you get there.
Or email [email protected]