Analyzing Booking Times with the Exponential Distribution

Welcome back, Detective. Our investigation into the suspicious booking surge at Aphrodite's Veil continues.

In our last briefing, we established a strong statistical basis for suspicion. By modeling their booking requests as a Poisson process, we determined that receiving 15 bookings in a single hour, when their average is only 5, is a highly improbable event—a 1-in-5000 chance. This gave us our first major clue: the count is anomalous. But a complete investigation requires corroborating evidence. Today, we turn our attention from the number of events to the time between them. If 15 requests were truly packed into 60 minutes, the gaps separating them must have been suspiciously brief.

This lesson will equip you with the tool to analyze these temporal gaps. We will explore the flip side of the Poisson process, the exponential distribution, to calculate the probability of the inter-booking times we would expect to see in such a surge. By analyzing the timing itself, we can build a second, independent line of evidence to either support or challenge the studio's claim.

The Other Side of the Coin: From Counts to Times

As you'll recall from your work with stochastic processes, any process where the count of events in an interval follows a Poisson distribution has a fascinating dual property: the time between those events follows an exponential distribution. They are two ways of describing the same underlying random process.

Poisson Distribution: Answers "How many events per unit of time?" (A discrete count)
Exponential Distribution: Answers "How much time between events?" (A continuous duration)

The following resource from ProbabilityCourse.com formally describes this fundamental relationship.

11.1.2 Basic Concepts of the Poisson Process

This section of the online textbook "Introduction to Probability, Statistics, and Random Processes" directly connects the Poisson process to its interarrival times.

Please read the section titled Interarrival Times. Focus on the core conclusion: if event counts follow a Poisson process with rate \lambda, the time between those events, X_i, is a random variable following an exponential distribution, also with parameter \lambda.

This duality gives us a new angle of attack. We have a rate of arrivals; now we can model the expected waiting time between them.

The Exponential Distribution: A Formal Introduction

The exponential distribution describes the probability of a certain amount of time passing before an event occurs. Its probability density function (PDF) is given by:

f (x) = λ e^{- λ x} for x \geq 0

Here, $x$ is the amount of time, and $λ$ is the rate parameter—the average number of events per unit of time. This is the same rate we used in our Poisson analysis.

The video below from LearnChemE provides a clear distinction between the Poisson and exponential distributions and introduces the key formulas we will need.

The Exponential Distribution

This video, "The Exponential Distribution," effectively contrasts it with the Poisson distribution and walks through the relevant functions.

Watch the first part of the video, from the beginning to the introduction of the formulas. Pay close attention to the explanation of how the exponential distribution "flips the question" from the number of events to the interval between them, and make sure you understand the PDF and CDF formulas presented.

A crucial point to internalize is the meaning of $λ$ and the mean.

$λ$ is the rate (e.g., 5 bookings/hour).
The mean time between events, denoted $μ$ , is its reciprocal: $μ = 1/ λ$ .

For Aphrodite's Veil, the claimed average rate is $λ = 5$ bookings per hour. Therefore, the average time between bookings should be $μ = 1/5$ of an hour, or 12 minutes. This is our baseline.

The shape of the exponential distribution is a rapidly decaying curve, showing that short gaps are much more common than long ones. The steepness of this decay is controlled by $λ$ .

Case Analysis: Were the Gaps Too Short?

To accommodate 15 bookings in one hour (60 minutes), the average time between them would have to be $60/15 = 4$ minutes. This is already suspicious, given the expected average of 12 minutes. But random processes are rarely perfectly average. The real question is: what is the probability of observing a gap of 4 minutes or less by pure chance?

To answer this, we use the cumulative distribution function (CDF) of the exponential distribution, which gives the probability of the waiting time being less than or equal to a certain value $x$ .

P (X \leq x) = 1 - e^{- λ x}

Let's calculate this. First, we must ensure our units are consistent.

Rate $λ = 5$ bookings per hour.
To work in minutes, we convert the rate: $λ = 5/60 \approx 0.0833$ bookings per minute.
We want to find the probability of a gap being 4 minutes or less: $P (X \leq 4)$ .

Plugging this into the CDF formula:

P (X \leq 4) = 1 - e^{- (5/60) \times 4} = 1 - e^{- 20/60} = 1 - e^{- 1/3}

P (X \leq 4) \approx 1 - 0.7165 = 0.2835

This tells us there is a 28.35% chance that any single gap between bookings is 4 minutes or shorter. This might not seem exceptionally low. However, to fit 15 bookings in an hour, you would need a sequence of 14 such short gaps. The probability of that happening consecutively is exceedingly small, confirming our initial suspicions from a different perspective.

The following resource provides several worked examples of calculating probabilities for various intervals, which will solidify your understanding of the mechanics.

The Exponential Distribution

This chapter from an introductory statistics course covers the fundamentals and provides clear examples of applying the exponential distribution.

Skim the initial introduction to see how the parameter is defined (note they use 'm' for the decay parameter, which is our \lambda). Then, carefully study the examples in the next section, which show how to calculate probabilities for P(a < x < b) and P(x > a) using the CDF. Finally, the Formula Review at the end is a useful quick reference.

The Memoryless Property: A Counter-Intuitive Clue

A defining and often counter-intuitive feature of the exponential distribution is its memoryless property. This property states that the time you have already waited for an event tells you nothing about how much longer you will have to wait. Formally:

P (X > s + t ∣ X > s) = P (X > t)

In the context of our case, it means that if you've been monitoring the booking line for 10 minutes without a request, the probability of waiting at least 3 more minutes is exactly the same as the probability of waiting at least 3 minutes from the very start. The process "forgets" the waiting time that has already passed.

This arises directly from the assumption that events (bookings) are independent. An excellent video from Iain Explains discusses this property and clarifies a common, but flawed, real-world analogy.

What is a Poisson Process?

This section of the "What is a Poisson Process?" video gives a great intuitive explanation of the memoryless property.

Watch the segment from the discussion of the memoryless property. The presenter does a particularly good job of explaining why the common "waiting for a bus" example is misleading and how the property correctly relates back to the core assumption of independent arrivals.

This property is a powerful diagnostic tool. If you were to find that wait times are dependent on past events (e.g., a long lull makes a booking more likely), it would be evidence that the underlying process is not a simple Poisson process, suggesting some other mechanism is at play.

Conclusion

Our second angle of investigation has reinforced our initial findings. By shifting our focus from the count of bookings to the time between them, we've established that the gaps would need to be suspiciously short to accommodate the claimed surge.

Here are the key takeaways for your case file:

The Model: The exponential distribution effectively models the time between independent events occurring at a constant average rate.
The Tool: Using the CDF, $P (X \leq x) = 1 - e^{- λ x}$ , we can calculate the probability of inter-event times falling within any specific range.
The Clue: The average inter-arrival time needed to fit 15 bookings in an hour is 4 minutes, far below the expected 12 minutes. While a single gap this short is not exceptionally rare (28% probability), a consistent pattern of such short gaps is highly improbable.
The Verdict: The temporal analysis corroborates the count analysis. Both point to the conclusion that the booking surge at Aphrodite's Veil was very unlikely to be a product of normal, random client activity.

Preview of the Next Lesson

We have analyzed the booking data as a single stream of events. But where did these bookings come from? Aphrodite's Veil takes bookings via phone, their website, and a third-party app. Is it possible the anomaly is isolated to just one of these channels? A surge from a single source could be a stronger indicator of manipulation, like a bot targeting their website.

In our next session, we will dissect the aggregate data stream. We'll learn techniques to analyze merged and thinned Poisson processes, allowing us to investigate whether the suspicious pattern holds across all channels or is concentrated in a way that reveals the saboteur's method.