The only alpha that compounds

This week I exported nineteen months of broker statements, handed them to Finn, and asked a question I had been avoiding. Was any of my trading actually worth doing?

The answer came back the same afternoon, and one line of it has stayed with me since.

My ideas beat every alternative I could have chosen. My unfinished exits gave the

profit back.

This post is about the method behind that sentence. It has a name, decision

attribution, a two-decade institutional history, and until recently a price tag that

kept it inside large funds. I want to explain why I think it is the most underrated

practice in investing, why almost nobody does it, and what it looks like when the cost

of doing it collapses to roughly zero. The worked example is my own money, with the

numbers left in.

The edge nobody can arbitrage away

Most edges in markets decay. An information edge decays fastest, because information

spreads. An analytical edge decays as tools improve, and the current wave of AI

research tools, Finn included, is commoditizing it at speed. Whatever a good analyst

could produce for you last year, a subscription produces for everyone this year.

One input to your returns is exempt from this race: the quality of your own decisions.

Nobody else can arbitrage your self-knowledge, and improvements to it do not get

competed away. A leak you find and fix in your own process stays fixed. That is the

sense in which learning from your own behavior is the only alpha that compounds. It is

not that other edges do not exist. It is that this one belongs to you permanently once

you earn it.

The size of the prize is documented. Morningstar's long-running Mind the Gap study

estimates that investors earn roughly 1.2 percentage points a year less than the funds

they invest in, mostly through the timing and size of their own transactions, which

over a decade is about fifteen percent of total returns. Academics argue about how much

of that gap is truly avoidable, and the honest reading is that the number is an

estimate rather than a law. Vanguard's Advisor's Alpha framework points the same

direction from the other side: it values behavioral coaching, an advisor talking a

client out of their own worst instincts, at up to one and a half to two percentage

points a year, the largest single component of what advisors are worth. In other words,

the industry already prices the behavior wedge at roughly the size of a good manager's

entire alpha. It just delivers the fix through expensive humans.

And the errors are not random. They are patterned, which is what makes them

addressable. You cannot fix bad luck. You can fix a habit, if someone shows it to you

with evidence.

The method exists, and it works at fund level

The practice with the best pedigree here is the decision journal. Daniel Kahneman, asked

in an interview what single thing an investor could do to improve their performance,

said to buy a cheap notebook and start keeping track: write down what you decided, why,

what you expected, and what alternatives you weighed, at the time, then revisit it

against what happened. The point is honest feedback about what you were actually

thinking, rather than what hindsight later tells you that you thought. Michael Mauboussin

has recommended the same for years. Michael Ervolini, who ran Cabot Research to analyze

fund managers' decision skill, built a whole book around the observation that conventional

tools give investors feedback about outcomes and almost none about the decisions that

produced them.

At the institutional level the practice grew into a product category. Essentia

Analytics has sold decision attribution to professional portfolio managers since 2013.

Their methodology, published in the peer-reviewed Journal of Investing, scores a manager's

skill across seven distinct decision types, stock picking, entry timing, sizing, scaling

in, size adjusting, scaling out, and exit timing, and their service pairs the analytics

with nudges and with a dedicated former fund manager who coaches you through your own

patterns. The results they publish are not small: one client case reports a sixty-eight

basis point annual improvement, another a five hundred plus basis point turnaround driven

largely by better trimming behavior. Morningstar has adopted their decision-skill score

into Morningstar Direct. A second firm, Alpha Theory, has spent twenty years on one

narrow slice of the same idea, forcing funds to make their sizing rules explicit, on the

argument that managers who never write down upside, downside and conviction leave a large

share of their potential returns unclaimed.

None of these firms has any connection to Finn, and nothing here is their endorsement.

I cite them because they are the proof that the method clears institutional scrutiny,

allocator money, and peer review. The method works. The catch has always been the

delivery: data integration projects, human coaches, fund-level fees. The peer-reviewed

version of their benchmark measured ninety active equity managers over three years. After

a decade of validation, that is the ceiling of the delivery model, and it says nothing

about the demand.

Why almost nobody does it

Below fund level, the practice dies on contact with reality, and the reason is well

understood by the one adjacent industry that made journaling work: trading journals for

day traders. Their own marketing states the law plainly. If logging a trade takes more

than about thirty seconds, you will eventually stop. Day-trading journals survived

because they auto-import from brokers and because their users generate daily material.

A long-term investor has neither property. You make a handful of real decisions a year,

so the habit never forms. Worse, the moments that most deserve a journal entry, the

panicked exit, the conviction add, the freeze during a drawdown, are precisely the

moments you are least likely to write one. The public record of attempts is a

graveyard. Investors describe hundred-page Word documents they can no longer search,

reasoning scattered across spreadsheets, emails and bookmarks, systems rebuilt every

January and abandoned by March. This is not a retail failing. A survey by Bipsync, a

research-management vendor, found that around three quarters of professional funds

still run their research process on shared drives, email and consumer note apps.

There is a second barrier behind the first. Even a perfectly kept diary cannot score

itself. Scoring a decision requires the counterfactual you were actually weighing at

the time, the prices since, and return math that handles money moving in and out. A

notebook holds the reasoning. It cannot mark it to market.

The closest precedent for how this gets solved comes from outside finance. Sleep

diaries existed for decades and almost nobody kept one. Sleep tracking became a

mass-market behavior the moment a device did the logging. People want the outcome,

which is self-knowledge. Almost nobody wants the activity, which is bookkeeping. Any

version of decision attribution that depends on the investor writing things down will

fail for the same reason every previous version failed.

What changes when the analyst is already in the room

Here is the structural shift. When your research happens in dialogue with an analyst

that also sees your book, the record accumulates automatically while it does the job you

hired it for.

Your reasoning is already articulated, because you articulated it while researching.

The alternatives you weighed are on the record, because you weighed them with the

analyst. The trades arrive from broker statements, so nothing needs manual entry. The

at-the-time snapshot, which is the whole scientific value of a decision journal, exists

by default, because the conversation happened at the time. Scoring becomes

deterministic code plus a narrator: pull the prices, compute the paths, and explain

what they mean. And the part no notebook ever offered, enforcement, becomes trivial,

because software does not get bored watching a level you set or a question you left

open.

One honesty note before the example, because it matters. None of this makes anyone

rational. What it does is make the record cheap and the mirror complete. The decisions

remain entirely yours, and so does the discomfort of reading about them.

Nineteen months of my own decisions, scored

The setup took me about ten minutes. I exported the transaction statements from my two

brokers, covering January 2025 through the start of this month, and gave them to Finn.

What came back was a full reconstruction of my portfolio, month by month, position by

position, and then the part I actually wanted: every decision scored against the

alternative I was weighing when I made it.

The reconstruction is worth one paragraph, because it is where this kind of exercise

usually goes wrong. Neither statement records what I held on the last day of 2024, so the

opening book was backed out from settlement amounts divided by the price on each trade

date. The check is that every implied quantity resolved to a whole number of shares at an

execution price within 0.1% of that day's close, which would not happen if a fund had

been misidentified. Both cash balances then tied to the statements to the cent. One

securities total came out 0.28% apart, because the workbook prices everything at exchange

closes and my broker uses its own snapshots. One holding predates the period with no

transaction in either statement, so it is carried forward on trust rather than derived.

The spine of the review is four versions of the same nineteen months. All four start

from my actual holdings, and all four receive exactly the money I deposited and

withdrew, on the same dates. The only thing that differs is what was done in between.

Read the last row first, because it is the good news. My trades, held rather than

unwound, would have beaten every passive alternative, including the global index. The

ideas were right. Then read the first row, which is what my account actually earned. The

six percentage points between those two rows come from two specific decisions: selling

holdings and leaving the proceeds in a deposit account. That gap is worth 5.3% of the

portfolio. Priced a second way, each exit against what that holding alone would be worth

today if I had left it, the same two decisions cost 6.7% of the portfolio, which is more

than three times everything my good calls earned combined. The two figures differ because

the first reruns the whole book and the second prices each exit on its own. I am quoting

both, because using whichever number suited the sentence is exactly the move this exercise

exists to catch.

The ledger underneath makes it concrete. In February 2025 I rotated the core of the

portfolio out of American funds and into European ones, weeks before that became a

consensus trade. Left alone, the new book would have returned 27.7% against 23.4% for

the American basket I sold. As I actually managed it, the new book returned 20.7%, which

is less than the thing I got rid of. The rotation did not fail. The selling that came

afterwards turned a good call into a slightly bad one. The review also added a footnote I

would never have written myself: the American funds I sold did fine too, up 26.8% and

39.3% in euro, so the honest description is that I picked the better horse in a race both

horses ran well, and any story I tell about escaping an expensive market is not

supported by my own data.

In August 2025 I sold a software position and moved the money into a pharmaceutical name.

The software stock has fallen by 51.9% since, and the replacement is roughly flat, so that

one swap preserved more than my two losing purchases of 2026 have lost together. It is

evidence for something the literature says is rare: I do not appear to suffer from loss

aversion. I sell losers cleanly.

Then the same action with the opposite grade. In November 2025 I sold an index tracker

and the money went nowhere. The fund is up 14.3% since, the cash earned under one percent

a year, and the difference between this exit and the good one was never the selling. It was

whether the proceeds had a destination. Measured against the money that was in it, that one

exit has cost 16.3% and the March exits 13.3%.

The sizing row stung the most. My best idea of the whole period, the European defence

entry, got 3.7% of the portfolio. It returned 34.3%, which added about 1.1% to the whole

book. My own written plan puts the active sleeve at roughly 15% of the portfolio; funded at

that level, the same idea would have added about 5.7% instead. My conviction and my position

sizes point in opposite directions, and the review named the pattern in one line: your best

ideas get your smallest cheques.

The cash chart is the one I find hardest to look at. Forty percent of the book in a

deposit account is one of the largest asset-allocation bets in it, and I never decided to

make it. April 2026 priced that: the portfolio rose 2.2% while the world index in euro

rose 8.3%, a gap of 6.2 percentage points in a single month, almost entirely because 44%

of the money was sitting in cash.

Two more findings deserve the daylight. First, in one eight-day stretch this July I

asked Finn four separate times how to deploy the same pool of idle cash. Four reports.

Zero allocations. The review concluded that, for allocation decisions, research was

functioning as a substitute for deciding, which is an uncomfortable thing for a

research product to tell its own customer, and exactly why I trust it more now.

Second, I once set a price floor on a position and wrote it down. In March that position

fell 37% in a month and closed below the floor. Nothing happened. It recovered, which

flatters the outcome and answers nothing about the process. A rule without enforcement is a

mood.

The last difference from every journal I ever attempted is that the review's findings

did not end as prose. Its five open questions became tracked items in my record, each

with an explicit condition that closes it, and three process rules it proposed are

sitting in a queue waiting for my yes or no. A fourth I have already adopted, and it is

the only one that binds Finn rather than me: a second report request on the same open

decision now has to carry a decide-by date, which I choose. Nothing gets to quietly

disappear, which, after nineteen months of things quietly disappearing, is the point.

Where this could be wrong

Every counterfactual above is marked at today's prices, and hindsight is doing real

work. The March exits happened during an oil shock, when Brent had touched $104 and

Hormuz shipping traffic was down 90 to 95%, and selling oil-exposed assets was a

defensible risk decision. A review written six weeks later would have graded them well.

In April my cash position briefly looked like wisdom. The fair criticism in my record is

about the nineteen weeks after the risk passed, when a decision made for a specific

reason outlived the reason with no one deciding to keep it.

Risk-adjusted, the picture also narrows. Forty percent cash lowers volatility, and if

that were my stated objective the gap in the table would partly be the price of

safety. It is not my stated objective. My own written framework, set two weeks before the

review ran, calls for an operational cash buffer of a little under 8% of the book. I am

holding 38.6%, so about 31 percentage points of the portfolio sit outside the plan I wrote

for myself. The only benchmark that is fairly mine is the one I wrote. That is the standard

the review holds me to, and the one I would encourage any investor to be measured against:

your own plan, at your own risk tolerance, with your own words at the time.

There is also a limitation in the report itself, and leaving it out would undercut the

whole argument. Finn runs its output through independent model reviews before it ships.

On this report those reviews failed to complete, and the report says so on its own last

page: it went out only partially validated. The deterministic checks passed, which is why

I trust the arithmetic above, and the tie-outs are reproducible from the workbook. The

prose claims got less scrutiny than they should have. I found one myself while writing

this post, a sentence in my own first draft that quoted the two-exit cost as the gap

between two rows of the table when those are different measurements. That is the kind of

thing the model lens exists to catch, and this time it did not run.

What the record becomes

In month one, this is a research analyst that happens to know your book. By month six

it holds your theses with their version history, your rules with their enforcement

record, and your decisions scored in both directions, the saves as well as the costs.

Somewhere in year two it can answer the only performance question that compounds: am I

actually getting better, in my own basis points, against my own history?

That accumulation is what decision attribution was always for. The fund-level version

proved the method and priced it out of reach. Doing your research beside an analyst that

keeps the record lets the same method accumulate automatically, and the record it builds

is, as far as I can tell, the only thing in investing that gets more valuable simply

because you kept going.

Measurement, not advice. The decisions, including all the ones above, were mine.

If you want the version of this with a stopwatch on it rather than a scorecard, the six-week time log

is the other end of the same argument: one PM's 55 hours of research work, done in 19.5,

published as recorded including an error Finn made.

Sources

• Morningstar, Mind the Gap

  — the 1.2 percentage point investor return gap over the decade to 31 December 2024, and

  about 15% of total fund return.

• Fulkerson, Jordan, Riley and Yan, Bad Timing Does Not Cost Investors 15% of Their Funds' Returns — the academic argument that the gap is smaller than the headline.

• Vanguard, Advisor's Alpha — behavioral coaching valued at about 150 basis points, the largest single component.

• Essentia Analytics, the Behavioral Alpha Benchmark and its methodology paper, published in the peer-reviewed Journal of Investing; the Morningstar Direct alliance; the 68 basis point case and the trimming case.

• Alpha Theory on position sizing and uncaptured alpha.

• Michael A. Ervolini, Managing Equity Portfolios, MIT Press, 2014.

• Daniel Kahneman on the decision journal, as recorded by Farnam Street.

• Bipsync, The Buy-Side Investment Research Process is Professionalizing

  — three quarters of funds surveyed still on shared drives, email and note apps. Bipsync

  sells research-management software, so read it as a vendor's survey.

Essentia Analytics, Morningstar and Alpha Theory have no connection to Finn, and nothing

here is their endorsement. They are cited because they are the evidence that scoring

decisions against stated alternatives clears institutional scrutiny and peer review. The

category is theirs. Finn's implementation is its own.

If you want your own version of the table above, request access.