The Indusa Blog

Share Button

Survival duration analysis is all about timing. Companies need to know when to strike while the iron is hot, but even if you might be able to successfully sell products (or services) to your customers right now, there’s plenty of questions you need be asking in order to stay relevant.

  • Will this short-lived this success last?
  • How long will it be before the current product we’re offering becomes obsolete?
  • At what point in time will our system completely break down?
  • How long will it be before my current marketing campaign’s message starts to become irrelevant?

Nothing lasts forever, and when in the business world, what’s most critical for you to undertake is a lifecycle analysis – either at the customer level, equipment level, or campaign level – to understand how your product, process, equipment or your customers’ needs change over time – how long until it transforms, decays, or dies.


Survival Duration Analysis: What Is It & Why Is It Important?

That’s where survival duration analysis comes to play. The Cornell Statistical Consulting Unit, Cornell University defines survival analysis as ‘a set of methods for analyzing data where the outcome variable is the time until the occurrence of an event of interest’. It comprises a set of techniques that can help you understand the relationship between time and events – and provide insight into how things change (machine operations, product sales, or market dynamics) over a period of time. It also helps you understand what would make such events occur at any given time.

The point and usefulness of the survival duration analytical approach is to directly model the passage and effect of time. There are two major components of a survival model:

  • The shape of the duration curve itself
  • The end-of-time event

Sometimes, the most interesting part of the analysis is the shape of the curve; at other times, the only thing we really care about is the length of time until ‘the end’. The shape of the curve utilizes predictive analytics.  The amount of the length of time is, by its nature, a forecast prediction.


Left and Right Censoring

An important aspect in survival duration analysis is censoring – while observing any activity, there are ideally two types of observations. 1) The event occurred, and we were able to measure when it occurred or 2) the event did NOT occur during the time it was observed. The end of any process, product, or equipment isn’t necessarily death (or total decay or a crash); it generally is just the date when you stop observing the change over time – technically known as right censoring, regardless of the type of the event. Therefore, survival duration analysis is less about causal factors other than those that are influenced by time, or that time influences. Often, a lot is not known about the event that starts the conditions leading to the change-over-time (not enough reliable initial data) leading to effectively missing data at the beginning of the analysis.

4 Key Measurements of Survival Duration Analysis

A significant aspect of survival duration analysis is the choice of time scale: you could either calculate survival since entry into the study, a particular event, or birth. However, survival duration analysis is different from time series forecasting because we are modeling change over time, whereas using time series assumes relative stability over time, even though there is volatility and shocks in a time series. Survival/duration analysis is something you use explicitly due to lack of stability over time, and the only thing that ‘looks’ stable is the censoring event(s) at the beginning and end. Survival duration analysis can be used to measure a host of different aspects in the business world today:

1. Churn Rate

Survival duration analysis is a great tool to calculate churn rate – of employees, products, and marketing campaigns. Say, for example, you want to calculate the churn rate of your app – i.e. the ratio of people who currently use your app vs. the people who initially downloaded it. With survival analysis, you can not only predict or understand when your customers will quit the app, but also get insight into when or how the probability of quitting changes over time – depending upon a variety of factors such as when those users joined, what demographics they belong to, and which device they use.

2. Equipment Failure Analysis

Survival duration analysis can be successfully applied to calculate the rate of equipment failure. Also known as failure time analysis, it builds models to indicate the time it takes for machines (or electronic components) to break down. By collecting and analyzing data, you can determine the exact time when a piece of equipment will fail and the cause of the failure (often with the goal of determining corrective actions). By understanding the how, why, what, and when of a failure, you can take necessary steps to prevent a reoccurrence, and improve the efficiency of your manufacturing process.

3. Demand Forecasting

In today’s highly dynamic businesses, keeping track of customer needs and demands is vital for excellence. The models typically used for demand forecasting give levels at points in time, but rarely give the effect of time on the process or object being studied. The amount of information we have on the dynamics of other factors versus time is an important determinant over which models are chosen and how we can interpret the outcome of the analysis. Using survival duration analysis, for instance, you can forecast demand (across production planning, inventory management, future capacity requirements, or in making decisions on whether to enter a new market) based on past events and current trends.

4. Accurately Calculating Growth (or Decay)

Whether or not we depict change over time as ‘stair steps’ or a smooth curve depends on what we are talking about and how much we care about intermediate steps versus the endpoint. Survival duration analysis’ modeling approach can be inverted; so instead of exponential decay, change can be calculated as exponential growth (and would portray something like saturation analysis). So, survival duration analysis is analogous because you would still care about the shape of the curve and the final event, but the change-over-time is different because the situation is different (growth, not decay). With survival duration analysis, you can not only predict the probability of a certain event to occur but also predict when it will occur.

Final Thoughts

Use the power of this type of analysis to leverage the various models, get the necessary heads up for special attention, and successfully retain existing customers, acquire new ones, manage your campaigns and machinery, and improve your business outcomes.

Looking to make more intelligent business decisions? Learn how Indusa can help you every step of the way.

New Call-to-action

Related Articles:

Manoj Nair
About the Author – Neha Kumar

Neha Kumar is a digital media evangelist and marketing professional. She overlooks Indusa’s content management, social media, online events, email marketing, blogs, digital campaigns, lead generation and inside sales activities on a broader scale.


Contibuting Author: Malavika Nityanandam