ADR-0046: benchmark comparison — a benchmark is a price series the portfolio’s own flows are replayed into

• Status: Accepted — owner sign-off is the merge of PR #780, the Sprint 11 planning PR (ADR-0026 step 1 as amended on that PR: the merge is the signature).
• Date: 2026-09-07
• Answers: FR-9 (PRD 2026-06-12, released from the advanced-reports gate on 2026-08-12 as scope-ladder level (b)), OQ-3 (the quote source), and the three design questions on #572. The asks it defers are listed in §6, per ADR-0043.

Context

The product’s founding question, in the owner’s words, is “war der Aufwand das wert?” — was the effort worth it. The two figures the performance surface already carries do not answer it. TTWROR (ADR-0010) answers “how did the investments do, as if the money had always been there”; the money-weighted set (ADR-0034) answers “what did my money earn”. Neither says compared to what — to having put the same money, on the same days, into a broad index, or to having left it on a savings account. Portfolio Performance shows both comparisons and its users expect them.

The prerequisites named on #572 are shipped: #545 (trade-price re-pricing no longer inflates long-period TTWROR, closed 2026-07-25) and #568 (the flow classifier of ADR-0034 §1, parameterized by scope so that the benchmark comparison can reuse it — the ADR says so in as many words). The shared chart component renders two series with a two-series tooltip and a percent-rebased mode; the quote sync already fetches any security’s history without a new provider. What was missing was a decision on three questions, and FR-9 carries two more from the PRD. This record answers them.

Decision

1. Two benchmark kinds behind one interface

A benchmark is anything that yields a daily price in the base currency for every day of the walk.

• Fixed-rate baseline — a constant annual rate compounding daily from a base of 1 (the savings-account alternative; “fixed-rate first” is the PRD’s own ordering under OQ-3). In v1 this is also how inflation is expressed: an operator types the rate. No table, no provider, no persistence beyond the request.
• Series benchmark — a security in the catalog carrying a new boolean is_benchmark (additive migration), whose quotes arrive through the existing quote sync. An index is proxied by an ETF, gold by an ETC; that is the quote-source decision of OQ-3: no new source. A benchmark security is never offered in booking forms and is excluded from the data-quality checks that nag about unheld securities; it may also be held (the ETF an operator owns can be their benchmark) — then it is simply both.

2. Two comparisons, both Portfolio Performance’s

• Bought once — the benchmark rebased to the period start, drawn next to the portfolio’s cumulative TTWROR on the same chart. Flow-neutral, so it is the honest companion of TTWROR: it answers “did my selection beat the index”.
• Savings plan (flow-matched) — every external flow of the portfolio, as ADR-0034 §1 classifies it at the same scope, is invested into the benchmark at that day’s price (a fixed rate: at par). The synthetic series has an end value; the comparison figure is the real end value minus the synthetic end value in the base currency, beside the two IRRs computed by ADR-0034’s XIRR on identical flows. This is the number that answers the founding question.

Rules the savings plan inherits rather than invents: a flow on a day without a benchmark close uses the most recent close on or before that day, exactly as the daily walk prices positions; a flow dated before the benchmark’s first quote is excluded and named, and the comparison states the window it covers (the Sprint 8 D-1 shape, applied uniformly). The synthetic portfolio is frictionless — no fees, no taxes — and the explanation line says so; the bias runs against the real portfolio, which is the conservative direction.

3. Scope and period

Portfolio-wide and per view — the classifier is scope-parameterized, and a strategy view’s “was it worth it” is a different question from the household’s. The period is the existing period picker; both comparisons use the same window as the figures they sit next to.

4. Surfaces, agent first

• API: the comparison is a read under /api/v1 on the two performance scopes (/portfolios/:id/performance/benchmark, /views/:id/performance/benchmark) taking benchmark= (security:<id> or rate:<decimal>) and period=, returning both comparisons with every financial value as a string and a computation_basis naming the window, the exclusions and the frictionless assumption. Benchmark securities are listed through the existing securities read with an is_benchmark filter. The contract manifest (#752) gains the entry in the same commit.
• MCP: the twin tools, schemas with the decimals as strings, per ADR-0002.
• UI: an overlay of at most two benchmarks on the Wealth performance chart, and a comparison block next to TTWROR/IRR carrying the one-line “what this answers” explanation EXPERIENCE.md requires of every figure. The selection is explicit per request in v1 (a query parameter the page remembers in the session); a stored per-view default is deferred until the choice proves stable enough to deserve a column.

5. What the engine does not do

Nothing is persisted: the synthetic series is derived on read and memoized under the one mechanism ADR-0039 provides, with the same lifetime as the walk it depends on. There is no benchmark blend, no per-security benchmark, and no attempt to model what a real savings plan would have cost.

6. The asks, answered and deferred (ADR-0043)

Consequences

• Risk-tier attention label (ADR-0036): money math on the performance surface. Its own commit group, exact-Decimal fixtures, and three pinned identities: a 0 % fixed-rate savings plan ends at exactly the net invested capital of ADR-0034; a benchmark that is the portfolio’s single holding with the same flows yields a zero end-value delta and equal IRRs; a flow before the first quote is excluded and named.
• The FR Coverage Map’s FR-9 row moves from “future” to the issue numbers the Sprint 11 plan files; E10 in the Tracker Index gains its first shipped item.
• The comparison is a read model over data the ledger already has. It introduces no write path, no order, no transmission of anything — the scope ladder’s level (b) and nothing beyond it.