Explained compute: the result and what drove it

The question that follows the number

A committee receives the quarterly report: the portfolio returned 4.12%. Almost always, the next question is the same: why? Was it fixed income, currency, an allocation decision, the timing of a contribution?

Calculating the return takes seconds. Answering that question usually takes days, because the answer lives somewhere else: in an attribution spreadsheet, in an analyst’s head or in a presentation assembled by hand.

What we mean by explained compute

An explained compute service returns two things at once: the result and an analysis of what drove it, at the same moment and from the same data. In the example above, the answer would not be just 4.12%, but something like:

The quarter returned 4.12%. Fixed income added 2.9 points and currency 1.8; the September fall in equities took away 0.6. The 14 August contribution came in before the rebound, which helped the result.

Illustrative example, not based on a real portfolio.

Why it matters that they come together

• Consistency. If the number and its explanation come from different calculations, sooner or later they stop matching. Calculated together, the explanation always adds up to what the result says.
• Speed. The committee’s question is answered in the same report, not at the next meeting.
• Traceability. The explanation is recorded with the same version of the calculation, so it can be reconstructed later.

Where it applies

Returns are the most obvious case, but the same idea works for the day-to-day change in VaR or in an exposure, for cash variances against budget or for reading a set of financial statements. In every case, the number is only half the answer.

This note accompanies the catalogue in preparation for Greyfield Data & Compute, where explained compute is one of the three service families.