Theory and Application of Survival Analysis

Theory and Application of Survival Analysis is a cornerstone topic for every serious data practitioner. Before you touch a single notebook, the decisions framed here shape which problems are worth solving, how value is measured, and which evidence counts as persuasive.

Why Theory Application Survival Matters

Strategic clarity at the start of a project compounds. A well-scoped problem with the right success metric is worth more than any sophisticated model built against a vague goal.

• Frame business goals as measurable analytical questions.
• Distinguish the data problem from the decision problem.
• Identify the smallest experiment that can falsify your hypothesis.
• Design feedback loops that keep strategy aligned with evidence.

How Theory Application Survival Shows Up in Practice

In a typical project, theory and application of survival analysis is combined with the rest of the Strategy & Foundations toolkit. You rarely use any one technique in isolation; the real skill is knowing which combination fits the problem you are trying to solve, and being able to explain that choice to a non-technical stakeholder.

Use these ideas when scoping a new analytics initiative, prioritising between competing proposals, or writing the first page of a data strategy for your team.

• The Strategic Imperative of Data-driven Decision
• Translating Business Objectives into Analytical Frameworks
• Principles of Critical Reasoning and Analytical
• Role Data Science Corporate Strategy Innovation

Back to the Data Science curriculum →

# Example 1: Five-year ROI scenario comparison -- Theory and Application of Survival Analysis
import numpy as np

scenarios = {
    "conservative": {"cost": 250_000, "annual_return": 0.06},
    "balanced":     {"cost": 250_000, "annual_return": 0.09},
    "aggressive":   {"cost": 250_000, "annual_return": 0.13},
}
years = np.arange(1, 6)

for name, s in scenarios.items():
    future_value = s["cost"] * (1 + s["annual_return"]) ** years
    npv          = future_value - s["cost"]
    payback_year = int(np.argmax(future_value >= s["cost"] * 1.5)) + 1
    print(f"{name:>12}: year-5 FV = ${future_value[-1]:>10,.0f} | "
          f"NPV = ${npv[-1]:>10,.0f} | 1.5x payback ~ year {payback_year}")

# Example 2: A/B test decision summary -- Theory and Application of Survival Analysis
import numpy as np
from scipy import stats

rng = np.random.default_rng(0)
control   = rng.binomial(1, 0.118, 5_200)
treatment = rng.binomial(1, 0.134, 5_200)

p_c, p_t = control.mean(), treatment.mean()
lift     = (p_t - p_c) / p_c
t, p_val = stats.ttest_ind(control, treatment, equal_var=False)

print(f"control rate   : {p_c:.3%}")
print(f"treatment rate : {p_t:.3%}")
print(f"relative lift  : {lift:+.1%}")
print(f"p-value        : {p_val:.4f}")
print("decision       :",
      "ship treatment" if (p_val < 0.05 and lift > 0) else "keep control")

# Example 3: Customer cohort retention matrix -- Theory and Application of Survival Analysis
import numpy as np
import pandas as pd

rng = np.random.default_rng(0)
n_users, n_months = 1_200, 12
signup = rng.integers(0, 6, n_users)                # cohort month 0..5
active = np.zeros((n_users, n_months), dtype=int)
for u in range(n_users):
    life = rng.geometric(p=0.18) + 1
    end  = min(signup[u] + life, n_months)
    active[u, signup[u]:end] = 1

df = pd.DataFrame(active, columns=[f"m{i}" for i in range(n_months)])
df["cohort"] = signup
cohorts = (
    df.groupby("cohort").mean()
      .round(2)
      .rename_axis(index="cohort_month")
)
print(cohorts.iloc[:, :8])

# Example 4: Monte-Carlo what-if for a pricing change -- Theory and Application of Survival Analysis
import numpy as np

rng = np.random.default_rng(0)
n   = 50_000
# Sample plausible inputs from prior beliefs
price_elasticity = rng.normal(-1.1, 0.25, n)       # demand % change per 1% price
price_change     = 0.08                             # +8% list price
baseline_volume  = rng.normal(12_000, 800,  n)
unit_cost        = rng.normal(22.0,   1.5,  n)
old_price        = 40.0

new_volume  = baseline_volume * (1 + price_elasticity * price_change)
old_profit  = (old_price - unit_cost) * baseline_volume
new_profit  = (old_price * (1 + price_change) - unit_cost) * new_volume
uplift      = new_profit - old_profit

print(f"expected uplift : ${uplift.mean():,.0f}")
print(f"5th-95th pct    : ${np.percentile(uplift, 5):,.0f} .. "
      f"${np.percentile(uplift, 95):,.0f}")
print(f"P(uplift > 0)   : {(uplift > 0).mean():.2%}")

# Example 5: Weekly KPI roll-up with pandas -- Theory and Application of Survival Analysis
import pandas as pd
import numpy as np

dates = pd.date_range("2026-01-01", periods=90, freq="D")
rng = np.random.default_rng(42)
df = pd.DataFrame({
    "date":         dates,
    "revenue":      rng.normal(12_000, 1500, 90).round(2),
    "active_users": rng.integers(8_000, 12_000, 90),
    "churned":      rng.integers(10, 60, 90),
})
df["arpu"]       = df["revenue"] / df["active_users"]
df["churn_rate"] = df["churned"] / df["active_users"]

weekly = (
    df.resample("W-MON", on="date")
      .agg(revenue=("revenue", "sum"),
           users=("active_users", "mean"),
           arpu=("arpu", "mean"),
           churn=("churn_rate", "mean"))
      .round(3)
)
print(weekly.tail())