free playbook · 1100 words · ~6 min read
I opened my broker app one Monday in 2026 and the pie chart said I was "diversified across 7 sectors." It lied to me, and not because the app was buggy. The app was working as designed.
The design lies because:
I built a tool to fix this for me (linked below). But the audit itself is more important than the tool. Here it is.
Share count is meaningless. 100 shares of NVDA is a very different position from 100 shares of T.
| ticker | shares | price | $ exposure | % of port |
|---|---|---|---|---|
| AAPL | 50 | 220 | 11,000 | 14% |
| MSFT | 30 | 410 | 12,300 | 15% |
| NVDA | 20 | 850 | 17,000 | 21% |
| ... | ... |
This is the only view that matters as a starting point. Broker apps default to share-count view; you have to click to get the dollar view. They default this way on purpose.
Instead of sector, group by what actually drives the price:
Re-add up the dollar exposure under these buckets.
The gap between "GICS pie" and "factor pie" is the lie size.
In my own case: GICS said 28% tech. Factor pie said 47% AI-infra mega-cap (because I was double-counting AMZN as "consumer," META as "comms," GOOGL as "comms"). The lie size was 19 percentage points.
If your largest factor bucket is > 30% of the book, you are not diversified. You have a thematic bet wearing a diversified costume.
Acceptable for: "I know I am 50% AI-infra, that is my thesis, I sized it deliberately."
Not acceptable for: "My broker says I am 28% tech and I believed it."
The distinction is knowing it vs being told. The audit puts knowing back in your hands.
This is the one most people skip and the one that matters most.
You don't need a quant lab. You need 30 days of daily closes per ticker (free from yfinance, polygon free tier, or portfolio-pulse which computes this server-side).
Compute pairwise daily-return correlation between your top 10 positions.
Any pair with correlation > 0.75 is functionally one position.
Worked example from my own book:
Once you merge correlated pairs, my "21 positions" collapsed to 9 factor bets. The broker pie still said 21.
This is the question to ask of every portfolio: "what did this exact book do in March 2020, in October 2022, in April 2025?"
If your top 5 holdings + their correlated cousins all drew down 35-40% in March 2020 together, that is your real drawdown band. Not the asset-allocation textbook number.
The broker app shows you "year-to-date return." It does not show you "worst 60-day drawdown across the last 5 years." That is the number you actually need.
(This is the radar chart portfolio-pulse draws — a one-page view of: concentration risk, sector drift vs S&P 500, correlated-pair count, and historical drawdown band. All server-side. Nothing stored.)
Market drift is real. A position you sized at 8% can be 14% in three months without you adding a share, just because it moved.
Put it on the calendar. Sunday 9pm. Paste positions. Look at the radar. One screenshot, one minute. If any factor bucket has crept past your personal cap, you trim Monday.
The discipline is not the audit. The discipline is doing it weekly so the answer cannot drift past you.
If this is useful, post your own redacted radar / factor-pie on X and tag me (@lekt8_). I collect the ones that surprised people the most and write them up. The redacted screenshots are the proof that broker pies hide reality at scale.
Run the audit on your own portfolio →Not financial advice. This guide is a structural framework for thinking about portfolio risk. It is not a recommendation to buy, hold, or sell any specific security. Your tax situation, time horizon, and personal risk tolerance are not visible to me or to any tool linked here.