← Back to Blog

Annual Review for Founders: The Read-Out of a Record You Kept

The Compass Team

November 18, 2026

In 2014 a business-process outsourcer called Wipro trained a batch of new customer-service agents. Same lessons, same tests. One group got 15 minutes at the end of each day to write about what they had learned.

That group scored 23% higher on the final assessment. Fifteen minutes on top of identical training, which is the entire evidence base for the annual review.

Giada Di Stefano, Francesca Gino, Gary Pisano and Bradley Staats ran the version in a lab, with brain teasers and a clock, and got 18%.

Then they offered people a choice: reflect, or take another round of practice. Most chose practice. The ones who reflected did better.

Peter Drucker got there in 1999, in an essay he called Managing Oneself. His method was feedback analysis: whenever you make a key decision or take a key action, write down what you expect will happen.

Nine or 12 months later, compare the actual results with your expectations. He had been practicing it for 15 to 20 years and said it still surprised him.

He also traced the habit back to an otherwise obscure fourteenth-century German theologian, picked up independently 150 years later by John Calvin and Ignatius of Loyola, each of whom built it into the practice of his followers. A one-page ritual that outlived the printing press.

Here is the part founders get wrong. The review reads from a record, and memory is a poor instrument for it in a specific, measurable way.

Teresa Amabile and Steven Kramer collected nearly 12,000 daily diary entries from 238 people on 26 project teams at 7 companies, then compared the best days to the worst.

The most common trigger of a best day was progress on the work, present on 76% of them. On the worst days, setbacks showed up 67% of the time and progress on only 25%.

The asymmetry is the finding. Setbacks hit two to three times harder than progress lifts. And 28% of small events had a major impact on how people felt about their work, which is the closest thing in the data to a lever.

Put those together and you get the founder's December problem. Your memory of the year is a pile of setbacks plus a vague sense that things moved. The small wins that actually carried the year are exactly the events that leave no trace.

So the review needs four lines, and three of them come from somewhere you already wrote.

The expectations: every key call with the prediction you recorded at the time, and how it turned out.

The wins: the small ones, dated, with the number that moved.

The misses: the call you would make differently, judged on what you knew then.

The rule: one sentence about how you will decide next year, written while the mistake is still expensive.

Read all four in one sitting. Then write the rule, because the rule is the only line that changes anything after January.

The honest limits are worth keeping in view. Wipro's customer-satisfaction edge was gone after a month, so a review is a practice rather than a permanent upgrade.

Amabile and Kramer call their analysis correlation, not proof of cause. Drucker's method takes two or three years before the pattern of your own strengths is clear.

The review earns its keep anyway. It turns a year into knowledge you can act on. Memory turns it into a mood.

Which is why the November work is dull. Go find what you wrote. If the record is thin, that is the finding, and next year's review starts today.

Compass keeps that record as you go, so December is a read-back: https://usecompass.app/?utm_source=blog&utm_medium=organic_social&utm_campaign=annual-review&utm_content=cmp_20261116_frameworks_annual-review_001

Share this article