Survival Curve | Zyeon UI

status="ready" — three onboarding cohorts, 180-day window. Every vertical tick is a censored account (still on trial when we last saw it); the shaded band is the 95% interval, and the numbers under the axis are the risk set each estimate is computed from.

Trial-to-paid retention

Kaplan–Meier, 180-day window · event = cancelled before converting

Self-serve260 accounts · 129 churn events · 131 censored · median 158 days
Guided onboarding240 accounts · 98 churn events · 142 censored · median not reached
Concierge setup118 accounts · 34 churn events · 84 censored · median not reached

Trial-to-paid retention: Kaplan–Meier survival curves for 3 groups over 180 days. Censored accounts are marked with a tick and stay in the estimate until they leave the risk set; they never count as churn events. Self-serve: 260 accounts, 129 churn events, 131 censored, 47.6% churn-free at 180 days, median 158 days. Guided onboarding: 240 accounts, 98 churn events, 142 censored, 57.7% churn-free at 180 days, median not reached. Concierge setup: 118 accounts, 34 churn events, 84 censored, 70.3% churn-free at 180 days, median not reached.
Group	Time (days)	Estimated churn-free	95% confidence interval	At risk	Cumulative churn events	Cumulative censored
Self-serve	0	99.6%	97.3% to 99.9%	260	1	0
Self-serve	50	81.2%	75.9% to 85.5%	202	48	10
Self-serve	100	65%	58.7% to 70.5%	158	88	14
Self-serve	150	52.8%	46.3% to 58.8%	124	117	19
Self-serve	200	beyond follow-up	beyond follow-up	0	129	131
Guided onboarding	0	100%	100% to 100%	240	0	0
Guided onboarding	50	86.9%	82% to 90.6%	206	31	4
Guided onboarding	100	75.8%	69.8% to 80.7%	172	57	11
Guided onboarding	150	64.3%	57.8% to 70.1%	146	83	11
Guided onboarding	200	beyond follow-up	beyond follow-up	0	98	142
Concierge setup	0	100%	100% to 100%	118	0	0
Concierge setup	50	93.1%	86.7% to 96.5%	107	8	3
Concierge setup	100	83.5%	75.3% to 89.1%	95	19	4
Concierge setup	150	76.4%	67.5% to 83.2%	87	27	4
Concierge setup	200	beyond follow-up	beyond follow-up	0	34	84

Censored · still churn-free when last observed, so the curve holds95% confidence

Single cohort, band={0.99}, showAtRisk={false} — the at-risk numbers move into the screen-reader table instead of disappearing.

Time to first hardware failure

One fleet, 90-day window · censored units were still healthy at the cut-off

Edge fleet v396 units · 41 failure events · 55 censored · median not reached

Time to first hardware failure: Kaplan–Meier survival curve over 90 days. Censored units are marked with a tick and stay in the estimate until they leave the risk set; they never count as failure events. Edge fleet v3: 96 units, 41 failure events, 55 censored, 56.3% failure-free at 90 days, median not reached.
Group	Time (days)	Estimated failure-free	99% confidence interval	At risk	Cumulative failure events	Cumulative censored
Edge fleet v3	0	100%	100% to 100%	96	0	0
Edge fleet v3	25	84.2%	71.5% to 91.5%	79	15	2
Edge fleet v3	50	70.3%	56.3% to 80.6%	66	28	2
Edge fleet v3	75	62.8%	48.6% to 74.1%	58	35	3
Edge fleet v3	100	beyond follow-up	beyond follow-up	0	41	55

Censored · still failure-free when last observed, so the curve holds99% confidence

Degenerate inputs: a cohort with zero subjects, a cohort where every observation is censored (curve holds at 100%, median “not reached”), a cohort wiped out by one event (S = 0, interval collapses), and a single subject censored at day 0.

Degenerate cohorts

Same contract, four inputs that a naive estimator divides by zero on

Never recruited (n = 0)no subjects
All censored (no events)8 subjects · 0 failure events · 8 censored · median not reached
Wiped out at week 84 subjects · 3 failure events · 1 censored · median 8 weeks
One subject, censored at entry1 subject · 0 failure events · 1 censored · median not reached

Degenerate cohorts: Kaplan–Meier survival curves for 4 groups over 20 weeks. Censored subjects are marked with a tick and stay in the estimate until they leave the risk set; they never count as failure events. Never recruited (n = 0): no subjects observed. All censored (no events): 8 subjects, 0 failure events, 8 censored, 100% failure-free at 20 weeks, median not reached. Wiped out at week 8: 4 subjects, 3 failure events, 1 censored, 0% failure-free at 8 weeks, median 8 weeks. One subject, censored at entry: 1 subject, 0 failure events, 1 censored, 100% failure-free at 0 weeks, median not reached.
Group	Time (weeks)	Estimated failure-free	95% confidence interval	At risk	Cumulative failure events	Cumulative censored
Never recruited (n = 0)	0	beyond follow-up	beyond follow-up	0	0	0
Never recruited (n = 0)	5	beyond follow-up	beyond follow-up	0	0	0
Never recruited (n = 0)	10	beyond follow-up	beyond follow-up	0	0	0
Never recruited (n = 0)	15	beyond follow-up	beyond follow-up	0	0	0
Never recruited (n = 0)	20	beyond follow-up	beyond follow-up	0	0	0
All censored (no events)	0	100%	100% to 100%	8	0	0
All censored (no events)	5	100%	100% to 100%	7	0	1
All censored (no events)	10	100%	100% to 100%	4	0	4
All censored (no events)	15	100%	100% to 100%	2	0	7
All censored (no events)	20	100%	100% to 100%	1	0	8
Wiped out at week 8	0	100%	100% to 100%	4	0	0
Wiped out at week 8	5	100%	100% to 100%	3	0	1
Wiped out at week 8	10	beyond follow-up	beyond follow-up	0	3	1
Wiped out at week 8	15	beyond follow-up	beyond follow-up	0	3	1
Wiped out at week 8	20	beyond follow-up	beyond follow-up	0	3	1
One subject, censored at entry	0	100%	100% to 100%	1	0	1
One subject, censored at entry	5	beyond follow-up	beyond follow-up	0	0	1
One subject, censored at entry	10	beyond follow-up	beyond follow-up	0	0	1
One subject, censored at entry	15	beyond follow-up	beyond follow-up	0	0	1
One subject, censored at entry	20	beyond follow-up	beyond follow-up	0	0	1

Censored · still failure-free when last observed, so the curve holds95% confidence

band={false} — bare curves plus censoring marks, for a dense dashboard grid. The marks never go away: a survival curve without them cannot be read.

Trial-to-paid retention

Kaplan–Meier, 180-day window · event = cancelled before converting

Self-serve260 accounts · 129 churn events · 131 censored · median 158 days
Guided onboarding240 accounts · 98 churn events · 142 censored · median not reached
Concierge setup118 accounts · 34 churn events · 84 censored · median not reached

Trial-to-paid retention: Kaplan–Meier survival curves for 3 groups over 180 days. Censored accounts are marked with a tick and stay in the estimate until they leave the risk set; they never count as churn events. Self-serve: 260 accounts, 129 churn events, 131 censored, 47.6% churn-free at 180 days, median 158 days. Guided onboarding: 240 accounts, 98 churn events, 142 censored, 57.7% churn-free at 180 days, median not reached. Concierge setup: 118 accounts, 34 churn events, 84 censored, 70.3% churn-free at 180 days, median not reached.
Group	Time (days)	Estimated churn-free	At risk	Cumulative churn events	Cumulative censored
Self-serve	0	99.6%	260	1	0
Self-serve	50	81.2%	202	48	10
Self-serve	100	65%	158	88	14
Self-serve	150	52.8%	124	117	19
Self-serve	200	beyond follow-up	0	129	131
Guided onboarding	0	100%	240	0	0
Guided onboarding	50	86.9%	206	31	4
Guided onboarding	100	75.8%	172	57	11
Guided onboarding	150	64.3%	146	83	11
Guided onboarding	200	beyond follow-up	0	98	142
Concierge setup	0	100%	118	0	0
Concierge setup	50	93.1%	107	8	3
Concierge setup	100	83.5%	95	19	4
Concierge setup	150	76.4%	87	27	4
Concierge setup	200	beyond follow-up	0	34	84

Censored · still churn-free when last observed, so the curve holds

status="loading" — the skeleton holds the same 240px plot and the at-risk rows, so nothing jumps when the cohort lands

Trial-to-paid retention

status="empty" — a healthy source with nobody enrolled yet (a "ready" payload whose cohorts are all empty renders this same branch)

Trial-to-paid retention

Kaplan–Meier, 180-day window · event = cancelled before converting

No accounts observed yet

The curve appears once follow-up times arrive — one row per subject, each marked an churn event or censored.

status="error" + onRetry — "Try again" swaps this card to the ready payload

Trial-to-paid retention

Kaplan–Meier, 180-day window · event = cancelled before converting

Couldn't load this cohort

The data source didn't respond.