Compounding

Thinking Tools5.1 Defining the Evaluation CriteriaAlso known as Compounding Reasoning

From Jacob Bernoulli's compound interest question (Quaestiones nonnullae de usuris, 1690), made a habit of business thinking by Charles Munger (A Lesson on Elementary, Worldly Wisdom, 1994).

Judges options by their effect on a rate that repeats over many cycles, since small differences in rate grow into large differences in outcome.

A small weekly gain that compounds, against a bigger one-time lift

The curve is the weekly program's conversion minus the one-time project's, week by week; the flat line at zero is where the two are level. The program starts behind by the full lift, crosses the line in week 23, and then pulls away faster every week. Change either rate to see how far the crossing moves, and the horizon to read the gap at any week. Illustrative.

12 percent
0.50 percent
26 weeks
One-time project: conversion at the horizon (today = 100)
112.0
Weekly program: conversion at the horizon (today = 100)
113.8
Week the program passes the project
week 22.7
Program's lead at the horizon, in points
1.8

Program minus project, points (today = 100)Level with the project

−50050100150TodayWeek 13Week 26Week 39Week 52Week 65Week 78Week 91Week 104−50050100150TodayWeek13Week26Week39Week52Week65Week78Week91Week104

An example

The growth team at a subscription software company had one engineer and one designer free for the next six months, and two ways to spend them. The first was a checkout redesign, forecast from a competitor teardown to lift trial-to-paid conversion by 12 percent, built in one push and shipped once. The second was a standing program of small weekly tests on the same funnel, which the team's past tests suggested would compound to about half a percent a week.

The product director wanted the redesign, since 12 percent was the bigger number on the slide. The team lead multiplied it out instead. At half a percent a week the program would pass the redesign in week 23, sit about two points ahead at the six-month review, and be about 18 points ahead after a year, still climbing.

They chose the program and made the redesign's best idea its first test. The director's condition was a monthly check on the realized weekly rate, since the whole case rested on it holding.

Why it works

People forecast in straight lines. A gain per cycle feels like it should add up to the rate times the cycles, so a 1 percent edge over fifty cycles feels like fifty percent at most. Under compounding, each cycle's gain is earned on everything gained before, so the outcome is the rate raised to the number of cycles, and that curve bends upward. Small rate differences open gaps far wider than the rates themselves.

The tool turns a choice between levels into a choice between rates. Once the cycles are counted and the outcomes multiplied out, an option that protects or raises the rate usually beats one that offers a bigger single step, provided the horizon is long enough and the rate holds. Finding the crossing week, and checking that the rate is real, is the whole analysis.

The thinking behind it

Jacob Bernoulli Jacob Bernoulli, the Swiss mathematician whose Ars Conjectandi (1713) founded the mathematical theory of probability, worked out a question about compound interest in 1683 and published it in Quaestiones nonnullae de usuris (1690), in Acta Eruditorum. Asking what a sum grows to when interest is added more and more often, he found that the result levels off at a limit, the constant now written e.

Charles T. Munger Charles T. Munger, vice chairman of Berkshire Hathaway, made compounding a habit of business thinking in his talk A Lesson on Elementary, Worldly Wisdom As It Relates to Investment Management and Business (1994), given at the USC Business School and later collected in Poor Charlie's Almanack (2005). His advice was to carry a latticework of models from many disciplines, and the mathematics of compound interest is one of the elementary ones. These are an investor's maxims from talks, not research.

How to run it

  1. Count the cycles. Name the unit that repeats (a week of tests, a year of reinvested profit, a hiring cycle) and how many of them the decision will run over.
  2. Estimate each option's rate per cycle. Use your own track record where one exists, and write down how sure you are of it.
  3. Separate one-time steps from rates. A step lifts the level once; a rate applies again every cycle, so put each on the right line.
  4. Multiply the outcomes out. Raise each rate to the number of cycles and set the results side by side at more than one horizon.
  5. Find the crossing point. Note the cycle at which the rate option passes the step, and check that the horizon reaches past it.
  6. Test what breaks the rate. List what would stop the rate from holding (fatigue, saturation, staff turnover) and how you would see it early.
  7. Track the realized rate. Measure the actual rate each period and rerun the comparison when it drifts.

Where it goes wrong

The rate is assumed, never measured. A team projects 1 percent a week from two good months and then never checks; spot it when the forecast curve is still on the slide a year later with no actuals beside it. Measure the realized rate every period and redraw.

The horizon is shorter than the crossing. Compounding wins in the long run, but a program funded for two quarters that needs three to pass the step loses. Find the crossing point before arguing for the rate.

Growth runs into a ceiling. No conversion rate climbs forever, and a market saturates. When the gains per cycle start shrinking, model the ceiling and compare the options again under it.

When to use something else

  • Every Lap Comes Home Ahead. When the gains come from a loop that feeds itself, and the question is how to keep the loop turning.
  • Opportunity Cost Reasoning. When the choice is a one-off use of money or people, with no rate repeating over many cycles.
  • Limits to Growth. When a compounding gain has started to slow and the job is to find the constraint holding it back.

Where it sits in SOLVE eX

Go to the source

  • Quaestiones nonnullae de usuris, Acta Eruditorum, 1690
  • A Lesson on Elementary, Worldly Wisdom As It Relates to Investment Management and Business, USC Business School talk, 1994

Back to the catalog