fix(dashboard): tell a day with no entry from a day that laid no eggs

specs/product/GLOSSARY.md

@@ -112,7 +112,7 @@ does not count (it vacated its day, #82); a Draft does (captured, not yet
 submitted). Depleted and archived flocks appear only when they do have an
 entry today. Beside it the dashboard shows the last 14 days from the
 **Production report** (submitted days only, yesterday back — a fixed window,
-unlike Reports) with the report's own **Hen-day %** for the last 7 complete
+unlike Reports) with the report's own **Hen-day %** for the last 7 finished
 days against the 7 before, and stock as one stacked bar by grade whose total
 is the Stock screen's.
 
@@ -700,11 +700,44 @@ the admin set on this screen, so clearing it would silently widen the page from
 one record to the whole farm. The empty state stays a plain sentence with no
 action (#655).
 
-**Hen-day % (#91)** — eggs collected ÷ hen-days × 100 (spec §19.3). A
-hen-day is one bird alive for one day; the day's bird count comes from the
-bird ledger (placements + movements). The production report shows it per
-day and for the whole period (period eggs ÷ period hen-days — not an
-average of daily percentages).
+**Hen-day % (#91, #780)** — eggs collected ÷ **recorded** hen-days × 100
+(spec §19.3, amended by #780). A hen-day is one bird alive for one day; the
+day's bird count comes from the bird ledger (placements + movements). The
+production report shows it per day and for the whole period (period eggs ÷
+period recorded hen-days — not an average of daily percentages).
+
+The denominator counts only the houses that **recorded** the day (#780). A
+house that filed nothing produced no eggs the report can see, so counting
+its birds as exposure states a lay rate nothing supports: a day nobody
+recorded read as 0%, and a day one house of three missed read as a collapse
+in output. The **Hen-days** column keeps its full meaning — every bird
+alive, recorded or not — and the gap between it and the recorded exposure
+is exactly what is missing. A day nobody recorded has **no** percentage,
+not 0.
+
+A flock files against a date, and the bird ledger decides separately whether
+that flock was live on it and how many birds it held. The two disagree at a
+lifecycle boundary — an entry filed on the day after depletion, a placement
+date corrected forward past entries that already exist, a mistyped mortality
+that removes more birds than were placed. **The rate's numerator is
+restricted to exactly the flocks whose birds are in its denominator**, so a
+flock in any of those states keeps its eggs in the day's egg total and out
+of the rate. Without that restriction the two exclusions compounded and a
+farm laying 80% was reported at 160%.
+
+The report therefore carries **Rated eggs** beside **Recorded hen-days** —
+the percentage's own numerator and denominator — so the figure can be
+reproduced rather than taken on trust. They equal the day's eggs and its
+hen-days on any ordinary day; a gap means a filing and the bird ledger
+disagree about that date.
+
+**Partly recorded day (#780)** — a date where at least one flock filed an
+official entry and at least one did not. Its egg total is a **floor**, not
+the day's output, so the Dashboard's 14-day strip draws it hatched and
+leaves it out of that panel's Peak and Avg; hen-day % still rates it on the
+flocks that did file, so a missing flock lowers what the figure is measured
+over rather than the figure itself. Distinct from a day nobody recorded,
+which has no figure at all.
 
 **Production report (#91)** — per-day official production over a range
 (Draft entries aren't submitted, Voided ones vacated their day — neither

specs/product/specs.md

@@ -2230,12 +2230,28 @@ sum(bird_inventory_movements.quantity_delta)
 
 ```text
 Hen-day production % (single day) =
-eggs collected on date / hen-days on date × 100
+eggs collected on date / RECORDED hen-days on date × 100
 
 Hen-day production % (period) =
-total eggs collected in period / sum(daily hen-days in period) × 100
+total eggs collected in period / sum(daily RECORDED hen-days in period) × 100
 ```
 
+**Recorded hen-days (#780).** The denominator counts only the flocks that filed
+an official entry for that date. The numerator can only ever contain eggs from
+flocks that filed, so counting a silent flock's birds as exposure reports a lay
+rate nothing supports: before #780 a day nobody recorded read as 0% production,
+and a day one flock of three missed read as a collapse in output. A date with no
+official entry has **no** hen-day % at all — not 0.
+
+The numerator is restricted to the same flocks. A flock outside its lifecycle
+window, or one whose removals have over-run its placement, has eggs the report
+shows and birds it does not count; admitting its eggs against a denominator it
+is absent from reported 160% on a farm laying 80%. Both halves of the fraction
+are carried in the payload so it can be reproduced. `hen-days on date` below
+keeps its full meaning (every live laying bird, recorded or not) and is what the
+report's own Hen-days column shows; the gap between the two is exactly what is
+missing. → [`docs/decisions/`](../../docs/decisions/) and issue #780.
+
 `hen-days on date` is the live laying-bird count **on that operational date**, reconstructed
 from the bird ledger, not the current count:
 
@@ -2259,8 +2275,8 @@ Rules:
   - Flag hen-day % as **approximate** for `mixed` flocks, and reserve a future
     `female_bird_count` (or sex-split placement movements) to make it exact.
   Do not silently report an exact-looking hen-day % for a mixed flock.
-- Period rate divides total eggs by the sum of daily hen-days, not by an end-of-period
-  headcount.
+- Period rate divides total eggs by the sum of daily **recorded** hen-days, not by an
+  end-of-period headcount and not by the calendar's full exposure (#780).
 - The benchmark-meaningful period anchors to **start of lay** (`flock.expected_start_lay_date`,
   or first laying production date), not an arbitrary calendar range. Comparing week 1 of
   lay against week 20 as one blended rate produces a number that matches no industry

src/Cluckwork.Application/Features/Reports/IReportQueries.cs

@@ -25,16 +25,58 @@ public interface IReportQueries
 // many did we grade" vs "how many can we sell"), they were only ever equal
 // because conditions used to be losses, and merging them would silently move
 // the number the capture screen is validated against.
+// #780 — three fields exist so a consumer can tell what the farm KNOWS from
+// what the farm PRODUCED. Every other figure here is 0 both for a day nobody
+// recorded and for a day that genuinely produced no eggs, so without them the
+// two are indistinguishable, which is what left the Dashboard's 14-day strip
+// asserting a zero it had no evidence for.
+//
+// `RecordedFlocks` — flocks that filed an OFFICIAL entry (Submitted, Locked,
+//   ManagerAdjusted). 0 means nobody recorded the day; it does not mean the
+//   farm produced nothing. Counted from the entries, so a flock that filed
+//   counts whether or not the bird ledger agrees it was live that day.
+// `ExpectedFlocks` — flocks that owed one: placed, not yet depleted or
+//   archived. It and `RecordedFlocks` can legitimately disagree EITHER WAY at a
+//   lifecycle boundary — a flock files on a day the ledger says it had ended, or
+//   a placement date is corrected forward past entries that already exist — so
+//   never assume one bounds the other.
+// `MissingFlocks` — expected flocks with no filing, and the ONLY sound test for
+//   a partly recorded day. Comparing the two counts above is not: they are
+//   counts over different sets, so expected {A, B} against filings {A, C} gives
+//   2 and 2 and hides B entirely, presenting a day short one flock's eggs as
+//   complete and feeding it into the Dashboard's Peak and Avg. `MissingFlocks`
+//   is computed from identities. `ExpectedFlocks - MissingFlocks` is how many
+//   of the flocks that owed a count actually filed one.
+// `RecordedHenDays` — the exposure behind the rate. `HenDays` keeps the
+//   glossary's meaning (one bird alive for one day, over every flock) so the
+//   Reports column still says what it always said; the rate divides by the
+//   subset with evidence behind it. Dividing by `HenDays` is what made an
+//   unrecorded day read as a day of zero production, and did the same to a day
+//   one flock of three missed.
+// `RatedEggs` — `HenDayPct`'s NUMERATOR: the eggs of exactly the flocks whose
+//   birds are in `RecordedHenDays`. Equal to `TotalEggs` on any day whose
+//   filings all come from live flocks with birds, which is every ordinary day.
+//   It is carried rather than left implicit so the rate is reproducible from
+//   the payload; without it, `TotalEggs / RecordedHenDays` silently disagrees
+//   with the percentage beside it on exactly the contradictory days.
 public sealed record ProductionDay(
     DateOnly Date, int TotalEggs, int Cracked, int Dirty, int Discarded,
-    int Sellable, int FromCounts, int Deaths, long HenDays, decimal? HenDayPct);
+    int Sellable, int FromCounts, int Deaths,
+    int RecordedFlocks, int ExpectedFlocks, int MissingFlocks,
+    long HenDays, long RecordedHenDays, int RatedEggs, decimal? HenDayPct);
 
 public sealed record GradeTotal(Guid EggGradeId, string Name, int Quantity);
 
+// `TotalRecordedHenDays` and `TotalRatedEggs` are `PeriodHenDayPct`'s exact
+// denominator and numerator (#780), carried so the period figure is
+// reproducible from the payload rather than being a number nobody can check.
+// `TotalRecordedHenDays` equals `TotalHenDays` on a period every flock
+// recorded, and the gap between them is exactly what is missing.
 public sealed record ProductionReport(
     IReadOnlyList<ProductionDay> Days,
     int TotalEggs, int TotalSellable, int TotalFromCounts, int TotalDeaths,
-    long TotalHenDays, decimal? PeriodHenDayPct,
+    long TotalHenDays, long TotalRecordedHenDays, int TotalRatedEggs,
+    decimal? PeriodHenDayPct,
     IReadOnlyList<GradeTotal> GradeTotals);
 
 // About the period's ORDERS (order date in range): revenue is their confirmed