What data adds to investment research

Quantitative work is useful in investment research when it answers a question that would otherwise be guessed. It is wasted when it dresses up a conclusion that was already fixed.

Questions data answers well

How concentrated is revenue across customers. How has pricing moved relative to volume. What does the cohort behaviour look like two years after acquisition. How does the company’s working capital compare with its peers. These are factual and can be settled.

Questions data answers badly

Whether a management team will execute. Whether a market will still exist in its current form in five years. Whether a competitor will respond. These remain judgements, and presenting them with a number attached makes them look more certain than they are.

Reproducibility is the standard

Work that another member cannot rerun is not evidence. Sources are documented, data pulled from filings rather than aggregators where possible, and assumptions written into the model rather than buried in a cell.

• State the question before building the model

• Keep raw data and transformations separate

• Note the date and source of every series

• Show the result that would contradict the thesis

The data analytics desk makes assumptions explicit and gives other members a clear basis for reviewing the analysis.