9 When do people come to Hubbard Hall?

When the same production plays Friday, Saturday, and Sunday, which day do people pick? This chapter leads with that within-run view — the cleanest read on audience day preference, since it holds the show constant — then asks how the answer has changed over time, whether opera and music runs behave the same way, and closes with the coarser all-events view. How runs are reconstructed from Neon events, what is excluded, and how headcount is counted are in the Methodology appendix.

9.1 Within a theater run: Friday, Saturday, or Sunday?

Only theater production runs that played all three of Friday, Saturday, and Sunday — mostly mainstage shows running one to four weekends. Attendance is true headcount, not registrations.

Code

import pandas as pd
import plotly.express as px
from hh.io import read_parquet
from hh.viz import style, BLUE, CATEGORICAL
from hh.analytics.timing import WEEKDAY_ORDER
from hh.analytics.productions import production_performances, full_weekend_runs, time_slot

regs = read_parquet("processed", "registrations_enriched.parquet")
perf = production_performances(regs, minorcats=("theater",))
fw = full_weekend_runs(perf)

day_order = [d for d in WEEKDAY_ORDER if d in set(fw["weekday"])]
day = (
    fw.groupby("weekday")
    .agg(performances=("swept_event_id", "count"), attendees=("attendees", "sum"))
    .reindex(day_order)
)
day["attendees per performance"] = day["attendees"] / day["performances"]
day["share of attendees"] = day["attendees"] / day["attendees"].sum()

n_runs = fw["production_run"].nunique()
yr_span = f"{fw['starts_on'].min():%Y}–{fw['starts_on'].max():%Y}"
sat_share = f"{day.loc['Saturday', 'share of attendees']:.0%}"
per_perf = {
    d: f"{day.loc[d, 'attendees per performance']:.0f}"
    for d in ("Friday", "Saturday", "Sunday")
}

Across the 34 qualifying production runs (2013–2026), 296 performances drew 10,701 attendees. Any weekday rows below Friday–Sunday are the rare midweek performances inside a qualifying run (Winter Carnival 2015 played two Thursdays):

Code

day.style.format(
    {
        "performances": "{:,}",
        "attendees": "{:,}",
        "attendees per performance": "{:.1f}",
        "share of attendees": "{:.1%}",
    }
)

	performances	attendees	attendees per performance	share of attendees
weekday
Thursday	2	22	11.0	0.2%
Friday	80	2,935	36.7	27.4%
Saturday	128	4,531	35.4	42.3%
Sunday	86	3,213	37.4	30.0%

Saturday brings in the most people (42% of attendees) — but only because most runs play Saturday twice (matinee + evening). Per performance the three days are essentially tied: Friday 37, Saturday 35, and Sunday 37 attendees per show.

How each production splits across the three days (runs with fewer than 30 Fri–Sun attendees are left off the chart but appear in the table below):

Code

wk = fw[fw["weekday"].isin(["Friday", "Saturday", "Sunday"])]
run_day = wk.groupby(["production_run", "weekday"], as_index=False)["attendees"].sum()
run_day["share"] = run_day["attendees"] / run_day.groupby("production_run")[
    "attendees"
].transform("sum")
order = fw.groupby("production_run")["starts_on"].min().sort_values().index.tolist()
run_totals = wk.groupby("production_run")["attendees"].sum()
charted = [r for r in order if run_totals[r] >= 30]
DAY_COLORS = {"Friday": CATEGORICAL[0], "Saturday": CATEGORICAL[1], "Sunday": CATEGORICAL[2]}

fig = px.bar(
    run_day[run_day["production_run"].isin(charted)],
    x="share",
    y="production_run",
    color="weekday",
    orientation="h",
    category_orders={
        "production_run": charted[::-1],
        "weekday": ["Friday", "Saturday", "Sunday"],
    },
    color_discrete_map=DAY_COLORS,
    custom_data=["attendees"],
)
fig.update_traces(
    hovertemplate="%{x:.0%} — %{customdata[0]:,} attendees<extra>%{fullData.name}</extra>"
)
fig.update_layout(
    height=740,
    xaxis_tickformat=".0%",
    xaxis_title="Share of run's Fri–Sun attendees",
    yaxis_title="",
)
style(fig).show()

The same runs in detail — attendees by day, with each run’s dates and performance count:

Code

by_day = (
    fw.pivot_table(
        index="production_run",
        columns="weekday",
        values="attendees",
        aggfunc="sum",
        fill_value=0,
    )
    .reindex(columns=day_order)
)
by_day["Total"] = by_day.sum(axis=1)
meta = fw.groupby("production_run").agg(
    first=("starts_on", "min"), last=("starts_on", "max"), performances=("swept_event_id", "count")
)
detail = meta.join(by_day).reindex(order)
detail["first"] = detail["first"].dt.strftime("%b %d, %Y")
detail["last"] = detail["last"].dt.strftime("%b %d, %Y")
detail.index.name = None
detail = detail.rename(columns={"first": "First", "last": "Last", "performances": "Perfs"})
detail.style.format("{:,}", subset=[c for c in detail.columns if c not in ("First", "Last")])

	First	Last	Perfs	Thursday	Friday	Saturday	Sunday	Total
Ondine (2013)	Nov 15, 2013	Dec 08, 2013	12	0	134	88	120	342
Parallel Lives (2014)	Jan 17, 2014	Feb 02, 2014	9	0	61	52	46	159
King Lear (2014)	Feb 28, 2014	Mar 23, 2014	12	0	107	84	96	287
Of Mice and Men (2014)	Apr 25, 2014	May 18, 2014	12	0	84	104	103	291
The 25th Annual Putnam County Spelling Bee (2014)	Nov 07, 2014	Nov 30, 2014	16	0	190	235	203	628
Winter Carnival of New Work (2015)	Jan 16, 2015	Feb 01, 2015	14	22	75	109	61	267
Tartuffe (2015)	Feb 20, 2015	Mar 15, 2015	16	0	81	168	89	338
An Iliad (2015)	Apr 24, 2015	May 17, 2015	12	0	79	82	59	220
Love's Labour's Lost (2015)	Jul 18, 2015	Jul 26, 2015	4	0	2	1	2	5
The Year of Magical Thinking (2015)	Oct 02, 2015	Oct 04, 2015	4	0	36	52	28	116
Winter Carnival of New Work (2016)	Jan 22, 2016	Jan 31, 2016	8	0	62	120	29	211
The Crucible (2016)	Apr 16, 2016	May 08, 2016	15	0	135	269	115	519
Peter and the Starcatcher (2016)	Oct 08, 2016	Oct 23, 2016	11	0	44	185	73	302
Winter Carnival of New Work (2017)	Jan 21, 2017	Jan 29, 2017	7	0	17	137	48	202
The Book Club Play (2017)	Mar 04, 2017	Mar 19, 2017	11	0	74	283	150	507
The Glass Menagerie (2017)	Apr 22, 2017	May 07, 2017	11	0	63	229	112	404
The Mystery of Edwin Drood (2017)	Nov 17, 2017	Dec 03, 2017	10	0	267	439	160	866
The Tarnation of Russell Colvin (2018)	Jan 19, 2018	Jan 28, 2018	8	0	94	118	80	292
The Velocity of Autumn (2018)	Feb 23, 2018	Mar 11, 2018	12	0	75	254	119	448
A Walk in the Woods (2019)	Jan 25, 2019	Feb 03, 2019	8	0	75	137	79	291
I Am My Own Wife (2019)	Mar 08, 2019	Mar 17, 2019	8	0	44	88	66	198
A Box of Monkeys (2019)	Jun 28, 2019	Jun 30, 2019	3	0	34	73	45	152
Faith Healer (2019)	Nov 15, 2019	Nov 24, 2019	6	0	71	84	53	208
Stupid F*%king Bird (2020)	Jan 24, 2020	Feb 02, 2020	8	0	116	171	86	373
The Susan B. Anthony Project (2020)	Oct 16, 2020	Oct 18, 2020	7	0	67	65	77	209
My Witch: The Margaret Hamilton Stories (2021)	Nov 05, 2021	Nov 14, 2021	6	0	87	94	99	280
Grant's Ghost (2022)	Nov 18, 2022	Nov 20, 2022	3	0	78	90	110	278
Fun Home (2023)	Feb 03, 2023	Feb 12, 2023	8	0	140	305	189	634
What the Constitution Means to Me (2024)	Sep 06, 2024	Sep 15, 2024	6	0	79	65	102	246
Valley Song (2024)	Nov 15, 2024	Nov 24, 2024	5	0	85	22	92	199
Dancing at Lughnasa (2025)	Apr 04, 2025	Apr 13, 2025	6	0	77	111	144	332
Eurydice (2025)	Oct 03, 2025	Oct 12, 2025	6	0	83	59	110	252
Fefu and Her Friends (2026)	May 01, 2026	May 10, 2026	6	0	63	46	97	206
Sunday in the Park with George (2026)	Jun 19, 2026	Jun 28, 2026	6	0	156	112	171	439

Saturday’s totals hide one more split: most runs play a 2 pm matinee and an evening show that day. Splitting Saturday performances by time slot:

Code

sat = fw[fw["weekday"] == "Saturday"].assign(slot=lambda d: d["event_name"].map(time_slot))
slot = (
    sat.groupby("slot")
    .agg(performances=("swept_event_id", "count"), attendees=("attendees", "sum"))
    .reindex(["matinee", "evening"])
)
slot["attendees per performance"] = slot["attendees"] / slot["performances"]
mat_rate = f"{slot.loc['matinee', 'attendees per performance']:.0f}"
eve_rate = f"{slot.loc['evening', 'attendees per performance']:.0f}"
slot.style.format(
    {"performances": "{:,}", "attendees": "{:,}", "attendees per performance": "{:.1f}"}
)

	performances	attendees	attendees per performance
slot
matinee	43	1,228	28.6
evening	85	3,303	38.9

Saturday evenings draw 39 attendees per show versus 29 for Saturday matinees — the weakest slot of the weekend. Sunday’s single 2 pm show (37 per show) draws as well as any slot, so the matinee habit is a Sunday habit, not a Saturday one.

9.2 Has the best day changed over time?

The numbers above pool 2013–2026, but the day ranking has not been stable. Still holding the product constant (full-weekend theater runs, attendees per performance so days with two shows don’t get double credit), here is the same question across roughly five-year periods.

Code

PERIODS = ["2013–2015", "2016–2020", "2021–2026"]

def period(year):
    return pd.cut(year, bins=[2010, 2015, 2020, 2026], labels=PERIODS)

run_rate = (
    wk.groupby(["production_run", "weekday"], as_index=False)
    .agg(performances=("attendees", "count"), attendees=("attendees", "sum"))
)
run_rate = run_rate[run_rate["production_run"].isin(charted)].copy()
run_rate["rate"] = run_rate["attendees"] / run_rate["performances"]
run_rate["run_start"] = run_rate["production_run"].map(
    wk.groupby("production_run")["starts_on"].min()
)

recent = run_rate[run_rate["run_start"] >= "2021-01-01"]
best_day = recent.loc[recent.groupby("production_run")["rate"].idxmax(), "weekday"]
n_recent = recent["production_run"].nunique()
n_recent_sun = int((best_day == "Sunday").sum())

wk_periods = wk.assign(period=period(wk["starts_on"].dt.year))
pd_rate = (
    wk_periods.groupby(["period", "weekday"], observed=True)
    .agg(performances=("attendees", "count"), attendees=("attendees", "sum"))
    .reset_index()
)
pd_rate["rate"] = pd_rate["attendees"] / pd_rate["performances"]
rate_of = lambda p, d: pd_rate.set_index(["period", "weekday"])["rate"].loc[(p, d)]

One dot per run and day — each run’s attendees per performance on Fridays, Saturdays, and Sundays:

Code

fig = px.scatter(
    run_rate.sort_values("run_start"),
    x="run_start",
    y="rate",
    color="weekday",
    color_discrete_map=DAY_COLORS,
    category_orders={"weekday": ["Friday", "Saturday", "Sunday"]},
    hover_name="production_run",
    custom_data=["performances", "attendees"],
)
fig.update_traces(
    hovertemplate=(
        "<b>%{hovertext}</b><br>%{y:.0f} attendees/perf "
        "(%{customdata[1]:,} over %{customdata[0]} shows)<extra>%{fullData.name}</extra>"
    )
)
for d in ("Friday", "Saturday", "Sunday"):  # 5-run rolling average as a guide line
    s = run_rate[run_rate["weekday"] == d].sort_values("run_start")
    fig.add_scatter(
        x=s["run_start"],
        y=s["rate"].rolling(5, center=True, min_periods=3).mean(),
        mode="lines",
        line=dict(color=DAY_COLORS[d], width=1.5),
        opacity=0.6,
        showlegend=False,
        hoverinfo="skip",
    )
fig.update_layout(
    xaxis_title="Run opening date", yaxis_title="Attendees per performance", legend_title=""
)
style(fig).show()

Figure 9.2: Attendees per performance by day, one point per full-weekend theater run

Aggregated to periods:

Code

fig = px.bar(
    pd_rate,
    x="period",
    y="rate",
    color="weekday",
    barmode="group",
    color_discrete_map=DAY_COLORS,
    category_orders={"period": PERIODS, "weekday": ["Friday", "Saturday", "Sunday"]},
    custom_data=["performances"],
)
fig.update_traces(
    hovertemplate="%{y:.1f} attendees/perf (%{customdata[0]} shows)<extra>%{fullData.name}</extra>"
)
fig.update_layout(
    xaxis_title="Period", yaxis_title="Attendees per performance", legend_title=""
)
style(fig).show()

Figure 9.3: Attendees per performance by day and period, full-weekend theater runs only

Per-performance attendance has roughly doubled across the periods, and the day ranking has flipped. In 2013–2015 Friday was the strongest day (27 per show vs 24 on Sunday); in 2016–2020 the days were about even; since 2021 Sunday is clearly strongest — 66 per show vs about 50 on Friday and Saturday. Sunday has been the best-attended day in 9 of the 9 full-weekend runs since 2021.

Recent-period caveats: 2021–2026 rests on 9 runs (52 performances), and recent runs rarely schedule Saturday matinees, which used to pull Saturday’s average down — yet Sunday leads anyway.

9.3 Do opera and music runs behave the same way?

Music and opera get the same treatment: map event names to productions, keep runs that played all of Friday, Saturday, and Sunday. Almost all music events are one-off concerts (Music from Salem plays single afternoons, cabarets play one or two nights), so nearly everything that qualifies is opera — the summer festival stagings of 2013–2017, which also played real midweek (Wednesday/Thursday) performances — plus one 2015 musical folktale with a theater-style four-weekend run. No over-time comparison is possible: the qualifying runs all fall in 2013–2017.

Code

mo_fw = full_weekend_runs(production_performances(regs, minorcats=("music", "opera")))

mo_day_order = [d for d in WEEKDAY_ORDER if d in set(mo_fw["weekday"])]
mo_day = (
    mo_fw.groupby("weekday")
    .agg(performances=("swept_event_id", "count"), attendees=("attendees", "sum"))
    .reindex(mo_day_order)
)
mo_day["attendees per performance"] = mo_day["attendees"] / mo_day["performances"]
mo_day["share of attendees"] = mo_day["attendees"] / mo_day["attendees"].sum()
mo_rate = {d: f"{mo_day.loc[d, 'attendees per performance']:.0f}" for d in mo_day_order}
n_mo_runs = mo_fw["production_run"].nunique()

The 9 qualifying runs, chronologically:

Code

mo_by_day = mo_fw.pivot_table(
    index="production_run",
    columns="weekday",
    values="attendees",
    aggfunc="sum",
    fill_value=0,
).reindex(columns=mo_day_order)
mo_by_day["Total"] = mo_by_day.sum(axis=1)
mo_meta = mo_fw.groupby("production_run").agg(
    first=("starts_on", "min"), last=("starts_on", "max"), performances=("swept_event_id", "count")
)
mo_order = mo_meta["first"].sort_values().index.tolist()
mo_detail = mo_meta.join(mo_by_day).reindex(mo_order)
mo_detail["first"] = mo_detail["first"].dt.strftime("%b %d, %Y")
mo_detail["last"] = mo_detail["last"].dt.strftime("%b %d, %Y")
mo_detail.index.name = None
mo_detail = mo_detail.rename(columns={"first": "First", "last": "Last", "performances": "Perfs"})
mo_detail.style.format(
    "{:,}", subset=[c for c in mo_detail.columns if c not in ("First", "Last")]
)

	First	Last	Perfs	Wednesday	Thursday	Friday	Saturday	Sunday	Total
The Barber of Seville (opera) (2013)	Aug 16, 2013	Aug 25, 2013	5	0	41	48	100	58	247
Trial by Jury (opera) (2013)	Aug 17, 2013	Aug 24, 2013	4	0	0	24	26	27	77
The Marriage of Figaro (opera) (2014)	Aug 15, 2014	Aug 24, 2014	5	0	61	34	137	75	307
Gianni Schicchi (opera) (2014)	Aug 16, 2014	Aug 22, 2014	3	0	0	95	19	36	150
Rigoletto (opera) (2015)	Aug 14, 2015	Aug 23, 2015	5	0	102	76	177	88	443
The Old Maid and the Thief (opera) (2015)	Aug 15, 2015	Aug 22, 2015	5	6	0	23	51	11	91
Wayward Home - A Musical Folktale (2015)	Nov 06, 2015	Nov 29, 2015	16	0	0	73	93	63	229
Madama Butterfly (opera) (2016)	Aug 12, 2016	Aug 21, 2016	5	0	98	59	180	77	414
The Elixir of Love (opera) (2017)	Aug 11, 2017	Aug 20, 2017	5	101	0	199	60	105	465

And by day of week:

Code

mo_day.style.format(
    {
        "performances": "{:,}",
        "attendees": "{:,}",
        "attendees per performance": "{:.1f}",
        "share of attendees": "{:.1%}",
    }
)

	performances	attendees	attendees per performance	share of attendees
weekday
Wednesday	2	107	53.5	4.4%
Thursday	4	302	75.5	12.5%
Friday	13	631	48.5	26.0%
Saturday	22	843	38.3	34.8%
Sunday	12	540	45.0	22.3%

Opera audiences didn’t privilege the weekend: the midweek opera nights drew the most people per performance (Thursday 76, versus 49 on Friday and 38 on Saturday, the weakest day per show — diluted, as in theater, by Saturday matinees). The sample is small (a few midweek performances, all in the 2013–2017 festival era), so treat the midweek numbers as suggestive rather than precise.

9.4 Which days are busiest overall?

The coarser view: every event of every kind, whether or not it was part of a run. Counts here are registrations (a proxy for headcount), and because most events are scheduled on the day the organizers chose, this mostly reflects what gets booked rather than audience preference — use the within-run sections above for that. It is kept for context.

Code

cat_order = ["theater", "music", "dance", "opera", "class"]
named = regs[regs["event_minorcat"].isin(cat_order)].assign(
    cat=lambda d: d["event_minorcat"]
)
combo = pd.concat([named, regs[regs["event_majorcat"] == "class"].assign(cat="class")])
combo = combo.dropna(subset=["weekday"])
tab = combo.groupby(["cat", "weekday"], observed=True).size().reset_index(name="attendance")
fig = px.bar(
    tab,
    x="weekday",
    y="attendance",
    facet_col="cat",
    category_orders={"cat": cat_order, "weekday": WEEKDAY_ORDER},
    color_discrete_sequence=[BLUE],
)
fig.update_layout(yaxis_title="Registrations")
fig.for_each_xaxis(lambda a: a.update(title=""))
fig.update_traces(hovertemplate="%{y:,.0f}<extra></extra>")
style(fig).show()

Theater peaks Saturday; music peaks Sunday. Classes spread across the week with a Friday peak.

How far people travel — including whether Friday theater audiences travel farther than Sunday’s — is in How far do people travel to Hubbard Hall?.