Performance attribution analysis in plain words: ask why a portfolio lagged, and get an answer with every sentence tied to its attribution cell
Why a portfolio lagged, answered the same day, each sentence tied to its attribution cell.
See one case, screen by screen ↓Performance attribution analysis explains why a portfolio's return differed from its benchmark, by splitting the gap into effects such as allocation, stock selection, interaction and currency. Performance Attribution Analysis adds seven agents that take a question such as “why did we lag?”, read the official attribution, the portfolio manager's notes, the trades and the issuers' own filings, and draft a plain answer in which every sentence lights the attribution cells it relies on. The PM reviews the reasons, a performance analyst approves with a reason code and signs, and only the client team sends.
AI in asset management: a buyer's guide to agentic AI for investment managers →
Updated
Performance attribution analysis tells you how much of a portfolio's gap to its benchmark came from allocation, stock selection, interaction and currency. It does not tell a consultant why. Every quarter consultants, pension boards, wealth clients and the firm's own sales teams ask why a strategy beat or lagged, and each answer becomes a request to the performance team: find the right attribution report and period, email the portfolio manager for the reason, write a draft, check every number against the official returns, then hand it to the client team to send.
The risk sits in the shortcuts. A preliminary figure goes out and is later restated, and the correction lands with every consultant who got it. Gross is shown without net. A driver is explained with a reason the PM never gave. A holding is named before it is public. Two people give the same consultant two different stories. And the same commentary sent to a consultant database's 41 recipients is no longer a conversation — under the SEC Marketing Rule it is treated as an advertisement.
Estimated split of about 2.5 analyst hours per answer by hand and about 0.75 hours to review a cited, checked draft, on the sample firm's volume of about 1,100 questions a quarter.
Seven agents read the official attribution, the PM's notes and the trades, draft and check every sentence; the PM reviews, a performance analyst signs, the client team sends.
Thursday, October 15, 16:42 ET. James Whitcombe of Hartwell Consulting Group writes on behalf of the Lakeshore County Employees' Retirement Fund, which has $1.24 billion in Quillfield's Global Equity strategy: “Ahead of the Lakeshore board on November 4: why did Global Equity lag the benchmark by 142bp in Q3, and is Marcus still holding Tessaro?” Here is what happens next, screen by screen, in the working solution.
Anika signed the quarter's official returns on October 9: 141 of 148 strategies are final. Her Today screen opens with an ask box and what the agents did overnight — 11 new questions logged, 9 drafted, every draft bound to the official Q3 attribution. Hartwell's question is first in Needs you: Marcus Hale reviewed the reasons at 08:12, 7 of 7 checks passed, due Monday at 16:42.
“Answers use only official, signed-off returns and documents you are entitled to see · every figure is cited.”
The Intake agent logged the email as PQ-26Q3-0418 with its CRM case and a tier-A clock of two business days. The Attribution reader read the official attribution ATT-GEQ-26Q3-v2, period Final. The Context reader read 14 Global Equity PM notes and found the reasons: Marcus Hale's July 22 note, his September 9 note, two Tessaro trades in the blotter and Tessaro's August 12 announcement, in German. The Narrator drafted six sentences, 142 words, for a consultant.
The Context reader quotes the PM. It never turns his notes into a new reason.
On the left, the answer: Global Equity returned 1.84% before fees (1.65% after) against 3.26% for the Norvale Global All Cap Index, so it lagged by 142bp before fees and 161bp after. On the right, the official Brinson-Fachler grid by sector. When Anika points at “Most of the gap was stock selection in Health Care, which cost 96bp”, row 14 lights up — portfolio return −6.39%, benchmark 2.18%, selection −96bp — and the rest dims.
“Hover a sentence to see the cells it relies on.”
Citation 3 opens Tessaro Health's ad hoc announcement of August 12 in German, with the English beside it, sentence by sentence: organic sales growth now expected at 2–4%, previously 7–9%, after an FDA warning letter on the Lörrach plant. The share price line shows CHF 88.40 on June 30 and CHF 60.82 on September 30, −31.2%. The same panel lists every source behind the answer: official returns, the attribution report, both PM notes, the trade blotter and the FX report.
The Consistency agent recomputed every figure: 14 checked, 0 differences. The effects add to the total — allocation −39, selection −94, interaction −1, currency −8, −142bp. It added the after-fee figures because the draft showed before-fee figures, and matched the Q3 factsheet, two earlier answers and the benchmark in Lakeshore's investment management agreement. The Disclosure agent found the September 30 Tessaro weight is public only from October 30, but Hartwell has a holdings confidentiality agreement on file, signed March 3, 2025, with a 15-day lag.
“Gross paired with net · Net-of-fee figures added (1.65%, −161bp) · equal prominence.”
Marcus reviewed Hartwell's answer at 08:12 and changed one phrase to his own words: “a one-off” became “contained to one plant”. On another of his strategies, Global Equity Income lost 31bp in September from Consumer Staples selection, mostly Lianhe Foods. No note, minute or trade explained holding it through the warning, so the Context reader's confidence was 0.71, below the 0.85 floor. It asked Marcus instead. Once he replies, his words are quoted in the answer and cited to his reply.
“I found no note, minute or trade that explains holding it through the warning. In one or two sentences, what was your reasoning?”
Anika asks why Global High Yield lagged in Q3. The Consistency agent will not answer yet: the period is preliminary while a price challenge on Varenne Telecom 7.25% 2029 is open with the fund administrator, and it is expected to be final on October 21. She asks which strategies lagged by more than 100bp. The Attribution reader lists four final strategies with their largest effect, each from the official Q3 attribution, and offers to prepare answers for all four.
“I will not answer that yet. Any figure I gave you now could change.”
Anika chooses A1 · Numbers agree with the official returns, and signs as Performance Analyst, Equities. The signature is recorded with the time and the answer version. Until she signs, the client team cannot send: the Send button turns on only after an analyst signs.
“I have checked the answer against the official returns and sign as Anika Sorensen, Performance Analyst, Equities.”
Owen opens the approved answer, reads the email to James Whitcombe — copied to himself and to Elena Vasquez, subject “Global Equity · Q3 2026 attribution (Lakeshore board, Nov 4)” — and sends it. The Delivery agent logs it to CRM case CRM-CS-2610-0412 and writes the evidence pack. Hartwell has a cited answer on Friday morning; the clock allowed until Monday afternoon.
The Delivery agent sends nothing on its own. A person on the client team presses Send.
International Small Cap's Q2 2026 was restated on October 2, after the spin-off of Brevik Marine from Halvard Group was booked late: the excess moved from −118bp to −137bp. The Attribution reader found the three answers already sent to Stanmore Partners, Larkspur Consulting and Grafton Union Pension Trust that quoted the old figure, and drafted corrections. Lucia Moretti approves them before anything goes out.
The same Global Equity answer, seen by five of the people who carry it — what their quarter-end looked like, and what it looks like now.
Seven agents read the official attribution, the PM's notes and the trades, draft and check every sentence; the PM reviews, a performance analyst signs, the client team sends.
Reads emails, CRM cases, consultant-portal questions and meeting notes; identifies the strategy, account, period, benchmark and question; merges duplicates and sets the clock from the asker's service level.
Pulls the official attribution for the exact strategy, period, model and benchmark from the performance measurement system — Brinson-Fachler for equities, carry, duration, curve, spread and selection for fixed income, account level for separate accounts — and finds sent answers a restatement has made wrong.
Reads PM notes, investment committee minutes, the trade blotter and issuer filings in their own language, and ties each decision to the driver it explains.
Drafts the answer for its audience — a consultant, a board or a wealth client, in English and the client's language — binding every sentence to the attribution cells and source passages it uses.
Recomputes every figure from the official returns, pairs gross with net, checks the effects add to the total, and compares the answer with the factsheet, the portfolio commentary, earlier answers and the mandate's benchmark.
Applies the holdings disclosure policy — the public lag and any confidentiality agreement on file — and the advertisement test, and routes an answer going to more than one recipient to Marketing Compliance.
Prepares the email and PDF, logs to the CRM after a person sends, watches the clocks, escalates late requests and writes the evidence pack.
Approves and signs; the client team sends. The agents propose; a named person decides.
Ask about any of the 148 strategies and any closed period. Answers come only from official, signed-off returns, every figure is cited, and a preliminary period is refused with the date it will be final. On a draft, ask back or instruct: “Make it shorter for the board” returns a three-sentence board version with the same citations, and it needs the same sign-off as the full answer.
Why did Global High Yield lag in Q3?
I will not answer that yet. Global High Yield Q3 2026 is preliminary: a price challenge on Varenne Telecom 7.25% 2029 is open with the fund administrator, and the period is expected to be final on Oct 21. Any figure I gave you now could change. I can tell you when it is final, and draft the answer then.
Which strategies lagged by more than 100bp in Q3, and why?
4 final strategies lagged by more than 100bp before fees in Q3 2026: Global Equity ex Fossil Fuels −151bp (no Energy, an exclusion, −56bp); Global Equity −142bp (Health Care selection −96bp, Tessaro Health); US Core Equity −105bp (underweight Utilities −51bp); Global Industrials −105bp (industry allocation −67bp). Each figure is from the official Q3 attribution.
Why did Emerging Markets Debt beat its benchmark in Q3?
Emerging Markets Debt was 61bp ahead of the Norvale EM Sovereign Index before fees in Q3 (48bp after). Bond selection in Latin American corporates added 29bp (Andes Cobre 2031s), and the overweight to frontier African sovereigns added 24bp. It was not duration: being short cost 6bp.
What drove Global Equity year to date?
Global Equity is 97bp behind its benchmark year to date before fees (154bp after): +31bp in Q1, +23bp in Q2 and −142bp in Q3. Year-to-date returns compound, so the gap is not the sum of the quarters (−88bp).
13 screens from the working solution, on its sample data. Pick one to see it large.
The rules every answer is checked against are settings: versioned, on the audit trail, proposed by the administrator or the Head of Performance and approved by a second person.
| Setting | Default | Choose from |
|---|---|---|
| Driver materiality (equity) | 5 bp | Name every effect of at least this size; a fixed income value is set separately |
| Rounding tolerance | 0.5 bp | Figures must tie to the official returns within this |
| Confidence floor for a reason | 0.85 | Below this, the agents ask the PM instead of answering |
| Periods allowed | Final and restated periods | Preliminary and open periods are refused |
| Gross and net | Always | Show after-fee figures whenever before-fee figures appear |
| Holdings named | 30-day public lag · 15-day under agreement | Public holdings, or recipients with an agreement on file |
| Advertisement test | On | More than one recipient → Marketing Compliance |
| Forecasts | Blocked | Never state or imply future returns |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate, from the sample firm's volume · “typical” = published figures (Cerulli via Institutional Investor — consultants have intermediated more than 60% of institutional flows; SEC enforcement actions under Rule 206(4)-1, via Lowenstein Sandler — 17 advisers fined about $2.3 million since November 2022; SEC Marketing Rule, Rule 206(4)-1, and the staff FAQ of March 19, 2025 on attribution shown as a portfolio characteristic). Quillfield Capital and every person, consultant, pension fund, issuer, index and identifier named on this page are fictional — characters and sample data in the working solution.
Performance attribution analysis explains why a portfolio's return differed from its benchmark, by splitting the gap into effects such as allocation, stock selection, interaction and currency. Performance Attribution Analysis adds seven agents that take a question such as “why did we lag?”, read the official attribution, the portfolio manager's notes, the trades and the issuers' own filings, and draft a plain answer in which every sentence lights the attribution cells it relies on. The PM reviews the reasons, a performance analyst approves with a reason code and signs, and only the client team sends.
The allocation effect measures what being overweight or underweight a sector added or cost; the selection effect measures what picking securities within the sector added or cost. In the sample answer, being underweight Energy (1.9% against 4.6%) cost 38bp of allocation, and stock selection in Health Care cost 96bp. Contribution to return is a different measure: what a holding added to the portfolio's own return, without the benchmark.
Yes. It reads the attribution model your performance measurement system already produces — in the working solution, Brinson-Fachler attribution by sector for equities, carry, duration, curve, spread and selection for fixed income, asset allocation and manager selection for multi-asset, and account level for separate accounts. The agents explain the official attribution; they never recalculate the returns.
The Attribution reader takes the largest effects from the official attribution, and the Context reader finds the decision behind each one in the PM's notes, committee minutes, trades and the issuer's filings. The Narrator writes the answer with every sentence bound to its cells, and the PM's reasons quoted, not rewritten. When no note explains a driver, the agents ask the PM instead of guessing.
Yes. Whenever an answer shows gross, before-fee returns, the Consistency agent adds the net, after-fee returns with equal prominence, the pairing the SEC Marketing Rule (Rule 206(4)-1) requires in advertisements. An answer going to more than one recipient is treated as an advertisement and goes to Marketing Compliance before it can be sent.
The Consistency agent refuses to answer from a preliminary or open period and says when it is expected to be final; the client team can send a holding reply. When official returns are restated, the agents find every answer already sent that quoted the old figure and draft corrections for the Head of Performance to approve.
Yes. The agents draft, check and route; they never send, approve or change an official return. The PM reviews the reasons, a performance analyst who did not draft approves with a reason code and signature, the client team sends, and every step is on the audit trail.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your attribution reports, PM notes, past answers and answer rules — and it goes live once every quality gate has passed. Setup connects your systems, maps strategies and benchmarks, and tests the agents on a question already answered by hand; the first week runs side by side with your own answers.
A cited yes, no or how-much for every fund, in under a minute, before the order.
1-min demoModel Portfolio RebalancingEvery model change reaches every account the same day, within each client's own rules.
1-min demoInvestment Policy Statement MonitoringEvery policy limit checked daily, and every exception tied to its clause and its holdings.
1-min demoWe'll run Performance Attribution Analysis on a sample of your own attribution reports, PM notes and past consultant questions.