Style Sense - Consistency, forecasting and market dynamics

14 March 2024

benchmarkcostforecastGatekeeperGatekeepersInvestmentsMarket Trends

Expert: Andy Evans, Fund Manager, Value and Jonathan Yow, Lead Data Scientist, Schroders Facilitator: Paul Kearney, Director, Asset Risk Consultants (UK) Limited


  1. Becoming better investors involves analysing historical buy/pass recommendations, benchmarking forecast accuracy versus consensus, identifying biases like the 'cookie problem' of underestimating costs, building tools to apply base rates and quantify PM preferences, evaluating value traps 


Schroders' Value Investment Team highlighted their process for reviewing past investment decisions to improve future forecasting accuracy.

Overall, the analysis led to more humility, adjusting models for cash costs, focusing forecast directionality over precision, emphasising risk scoring, and seeking reader-defined value.

Analysing historical buy/pass decisions:
The Value Team has an archive of over 3000 buy/pass recommendations since 2013 capturing key valuation metrics, ratios, thesis factors and actual outcomes. Initial analysis reviewed forecast accuracy for sales, profits, cash flows against realized figures and showed the team was broadly directionally right but with a wide range of outcomes reflecting the challenge of predicting the future.

Benchmarking against consensus forecasts:
Comparing the team's 3-5 year forecasts versus consensus from the time of initial valuation showed their median accuracy was better especially further out in time, though still off in absolute terms. However more accuracy did not clearly translate into better subsequent stock returns.

Learning from past errors:
Detailed after-action reviews assessing each decision revealed biases like the 'cookie problem' of underestimating costs. This led to adjusting models for cash costs and focusing more on IRRs and value traps.

Building tools to improve decisions:
New tools were built applying base rates to anchor expectations and machine learning to quantify PM preferences. Though not yet used for portfolio construction, these provide warning flags for potential biases and help screen for opportunities.

Assessing value traps
Looking holistically at buy/pass recommendations and subsequent returns, the process exhibited some efficacy in avoiding 'value traps' - stocks falling over 20%. But further work is needed to evaluate if holdings were exited prematurely.

Becoming better investors:
The overall analysis led to more humility in forecasting, adjusting valuation models, emphasizing reader-defined value versus original ideas, and seeking to improve via evidence-based reviews - though markets involve many intertwined variables. Continued iterations and data gathering are planned.

Key takeaways:

  • Identity key value words for specific journal or reader community codes
  • Aim for reader-defined value not original or new knowledge