Season-ticket renewal risk
Season tickets are the largest recurring revenue line a club has. This model scores a simulated book of 2,400 accounts, ranks it by renewal risk, and sizes the retention program: who to call, what it costs, and what it saves.
Season outlook
Retention program
Outreach capacity
400 accounts
How many at-risk accounts reps can personally work before renewal deadline.
Assumed outreach uplift
+8 pts
Renewal-probability lift from a personal touch. In production this comes from a holdout test, not an assumption.
Cost per contact
$40
Loaded rep time plus any save offer (gift, upgrade, payment flexibility).
Model coefficients
Log-odds contributions, shown openly.
Seat utilization (scan rate)
+2.30 × (scan − 50%)
Share of games resold
−1.70 × share
Auto-renew enrolled
+0.90
Tenure (per year, capped)
+0.05 / yr
Club seating
+0.35
Upper bowl
−0.25
Service contacts (each)
−0.12 / contact
Distance from arena
−0.005 / mile
Forecast · average season
85.0%
forecasted renewal rate across 2,400 accounts and $18,512,853 of season-ticket revenue
$2,574,260
Expected revenue at risk
$217,347
Revenue protected
4.9x
Margin return on $16,000 program
Revenue at risk, by model risk decile
Most of the exposure sits in the top deciles — the case for a ranked call list instead of calling everyone.
Decile 1highest risk
$646,218
Decile 2
$496,871
Decile 3
$395,069
Decile 4
$262,558
Decile 5
$217,622
Decile 6
$181,846
Decile 7
$143,998
Decile 8
$104,455
Decile 9
$77,527
Decile 10safest
$48,096
deciles covered by current outreach capacity
The leading indicator: seat utilization
Accounts that stop showing up stop renewing. Falling scan rates are visible months before the renewal decision.
Under 40%
74.5% (809)
40–60%
86.1% (664)
60–80%
92.2% (590)
Over 80%
95.6% (337)
The call list — twelve highest-risk accounts
| Account | Tenure | Section | Scan | Resold | Value | P(renew) | Primary risk driver |
|---|---|---|---|---|---|---|---|
| A-2801 | 2 yr | Upper bowl | 6% | 62% | $6,096 | 28% | resells 62% of games |
| A-1896 | 1 yr | Upper bowl | 18% | 55% | $2,900 | 34% | resells 55% of games |
| A-2457 | 1 yr | Lower bowl | 7% | 48% | $5,855 | 35% | scan rate 7% |
| A-2046 | 7 yr | Upper bowl | 11% | 71% | $5,656 | 35% | resells 71% of games |
| A-3017 | 1 yr | Lower bowl | 6% | 63% | $13,209 | 35% | resells 63% of games |
| A-1696 | 1 yr | Upper bowl | 7% | 57% | $5,857 | 37% | scan rate 7% |
| A-1472 | 3 yr | Upper bowl | 21% | 67% | $2,681 | 37% | resells 67% of games |
| A-2401 | 2 yr | Upper bowl | 16% | 48% | $6,110 | 38% | not on auto-renew |
| A-2113 | 1 yr | Club | 12% | 50% | $25,497 | 38% | not on auto-renew |
| A-1830 | 1 yr | Upper bowl | 5% | 52% | $6,091 | 38% | scan rate 5% |
| A-3137 | 6 yr | Upper bowl | 13% | 55% | $2,734 | 39% | resells 55% of games |
| A-1985 | 1 yr | Upper bowl | 5% | 62% | $3,093 | 40% | resells 62% of games |
Method
Each account’s renewal probability is a logistic model over tenure, seat utilization (scan rate), resale share, auto-renew enrollment, seat tier, service history, and distance, with a season-outlook shift on the intercept since team performance moves renewal intent broadly. Coefficients are displayed in the sidebar; each account’s primary risk driver is its largest negative log-odds contribution, translated to plain language for the rep making the call.
The retention simulator contacts the highest-risk accounts first and applies an assumed uplift, capped so no account exceeds 98% renewal probability. Revenue protected is the sum of uplift × account value over the contacted set; program return is then computed on the contribution margin of that revenue (36%, consistent with the lifetime-value model) rather than the gross figure — crediting full ticket revenue to a phone call would flatter the program. Uplift should be measured with holdout groups rather than assumed; defining those experiments early is how the program proves the ROI it claims.
The account book is simulated (seeded, stable across visits) and scaled loosely to a mid-market NHL club. In production this model would train on several seasons of actual renewal outcomes in Archtics/KORE data, validate on a held-out season, and be recalibrated as scan behavior shifts.