Recomendados
Recomendados
Mais conteúdo relacionado
Mais procurados
Mais procurados (20)
Semelhante a Nick Jesteadt: Predictive Attrition Using Survival Analysis
Semelhante a Nick Jesteadt: Predictive Attrition Using Survival Analysis (20)
Mais de Edunomica
Mais de Edunomica (20)
Último
Último (20)
Nick Jesteadt: Predictive Attrition Using Survival Analysis
- 1. Survival Analysis during the Great Resignation ML approach to attrition intervention Nick Jesteadt February 4, 2022
- 2. The information in this document is confidential information of TIBCO Software Inc. and/or its affiliates. Use, duplication, transmission, or republication for any purpose without the prior written consent of TIBCO is expressly prohibited. This document (including, without limitation, any product roadmap or statement of direction data) illustrates the planned testing, release and availability dates for TIBCO products and services. This document is provided for informational purposes only and its contents are subject to change without notice. TIBCO makes no warranties, express or implied, in or relating to this document or any information in it, including, without limitation, that this document, or any information in it, is error-free or meets any conditions of merchantability or fitness for a particular purpose. The material provided is for informational purposes only, and should not be relied on in making a purchasing decision. The information is not a commitment, promise or legal obligation to deliver any material, code, or functionality. The development, release, and timing of any features or functionality described for our products remains at our sole discretion. During the course of this presentation, TIBCO or its representatives may make forward-looking statements regarding future events, TIBCO’s future results or our future financial performance. These statements are based on management’s current expectations. Although we believe that the expectations reflected in the forward-looking statements contained in this presentation are reasonable, these expectations or any such forward-looking statements could prove to be incorrect and actual results or financial performance could differ materially from those stated herein. TIBCO does not undertake to update any forward-looking statement that may be made from time to time or on its behalf. CONFIDENTIALITY & DISCLAIMER © Copyright 2000-2022 TIBCO Software Inc.
- 3. © Copyright 2000-2022 TIBCO Software Inc. Nick Jesteadt njestead@tibco.com Senior Director of People Analytics, TIBCO Software
- 4. © Copyright 2000-2022 TIBCO Software Inc. What is Survival Analysis? Survival analysis analyzes the expected duration of time until an event occurs or the probability that an event occurs within a stated time. Origins in epidemiology. Can be used in an HR application to understand the probability of voluntary attrition (departure from the company being the event).
- 5. Survival versus Attrition © Copyright 2000-2022 TIBCO Software Inc. 15% voluntary attrition • Reactive & Past Oriented • Not Predictive • Benchmarked & Essential 90% 79% 67% 56% 48% 0% 25% 50% 75% 100% 0 1 2 3 4 5 RETENTION PROBABILITY TENURE (YEARS) SURVIVAL FORECAST • Proactive & Future-Focused • Forecast • Uncommon but Strategic
- 6. Predicting the WHEN of Turnover © Copyright 2000-2022 TIBCO Software Inc. 100% 90% 79% 67% 56% 48% 93% 86% 80% 74% 70% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 RETENTION PROBABILITY TENURE (YEARS) SURVIVAL FORECAST Gen Y Gen X Difference of 10% vs 7% risk in the first year of tenure Between years 1 and 3, risk for Gen X is 13% while Gen Y is 23%. We were able to use the ‘risk windows’ of various teams alongside focus groups, exit interviews, and engagement surveys to identify the drivers of that risk.
- 7. Predicting the WHEN of Turnover © Copyright 2000-2022 TIBCO Software Inc. 100% 92% 82% 71% 60% 57% 87% 80% 74% 68% 94% 79% 66% 59% 59% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 RETENTION PROBABILITY TENURE (YEARS) SURVIVAL FORECAST Compa < 0.9 Compa 0.9-1.1 Compa >1.1 Minimal variance in pay through 2 years. Those receiving consistent advancement & market pay had much greater longevity; Those paid well but not advanced and those not paid at market faced similar survival curves. We used insights such as these to determine a more involved advancement and pay philosophy. One that identified those that are stagnant (> 3 years of tenure, < 0.9 compa-ratio, no recent promotion history).
- 8. Forecasts through Simulation © Copyright 2000-2022 TIBCO Software Inc. Employee Tenure Team Stagnant “Survival” Risk in Next Year Ben Block 2 Finance 11% Gabby Green 3.5 Sales Y 19% Mike Mayhue 8 IT Y 7% Lisa Light 0.5 Sales 19% Hanz Herron 1 Sales 15% Heleen Hansel 4 Product Y 9% 125 195 430 130 70 40 10 0 100 200 300 400 500 0 Departures 1 2 3 4 5 6 Probability Distribution of Departures 1000 unique Monte Carlo Simulations
- 9. Interactive Employee Survival Analysis © Copyright 2000-2022 TIBCO Software Inc.
- 10. Building your own dataset © Copyright 2000-2022 TIBCO Software Inc. Record # Tenure or Time in Role Term or Role Change Date Event … Demographic Cut #1 Demographic Cut #2 Etc… 12345 5.87 0 Sales Male 12346 4.02 May ’21 1 Sales Male 12347 4.88 0 Finance Female 12348 1.40 Dec ’19 1 HR Male … 23456 3.10 0 Sales Male 23457 2.75 Nov ‘20 1 IT Female Considerations: exclude involuntary terminations; conduct a separate survival analysis for those from acquisitions or significant hiring events; adjust tenure window to meet the needs and norms of your business
- 11. Kaplan-Meier Model © Copyright 2000-2022 TIBCO Software Inc. Tenure (Years) Number at Risk (Headcount) Voluntary Departures Survival Probability t Nt Dt 0 20 1 0.5 20 1 1*((20-1)/20) = 0.95 1 19 0.95 1.5 19 2 0.95*((19-2)/19) = 0.85 2 17 0.85 2.5 17 3 0.85*((17-3)/17) = 0.70 3 14 0.70 … Survival Probability: St+1 = St*((Nt+1 – Dt+1)/Nt+1)
- 12. Kaplan-Meier Model © Copyright 2000-2022 TIBCO Software Inc. 100% 95% 95% 85% 85% 70% 70% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 0.5 1 1.5 2 2.5 3 RETENTION PROBABILITY TENURE (YEARS) SURVIVAL FORECAST
- 13. © Copyright 2000-2022 TIBCO Software Inc. Applications in Economic Labor Markets
- 14. The Great Resignation & Economic Uncertainty © Copyright 2000-2022 TIBCO Software Inc. Poor Economy Growth Economy • Less Job Scarcity • Market is more risk-friendly • More ample opportunities • Increase in pay • More Job Scarcity • Market is more risk-averse • Less ample opportunities • Decrease/ Stabilization in pay Brexit Trump Election Start of US-China Trade Tensions Covid Political Unrest The Great Resignation
- 15. It’s All Relative © Copyright 2000-2022 TIBCO Software Inc. x ½ x 2 More Uncertain – more likely to stay Less Uncertain – less likely to stay
- 16. Impact on Survival Forecasts © Copyright 2000-2022 TIBCO Software Inc. 70% 58% 49% 45% 76% 51% 35% 19% 97% 90% 82% 74% 68% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 1 2 3 4 5 RETENTION PROBABILITY TENURE (YEARS) SURVIVAL FORECAST * Lower and higher uncertainty was determined by looking at the total distribution of current (or term date) uncertainty to hire data uncertainty (as ratios) and using positive and negative standard deviations from the mean ratio during the timeframe of the analysis. Lower denotes negative std. dev. upper denotes positive std. dev. Voluntary Attrition Average Tenure 7.1% 4.5% vs 1.2 2.5 vs Hires since 2014
- 17. © Copyright 2000-2022 TIBCO Software Inc. Questions?
- 18. Thank you © Copyright 2000-2022 TIBCO Software Inc.
Notas do Editor
- Introduction to my background and to TIBCO (2 mins)
- Before going into HR application, overview of survival analysis – statistics and health research – loaded roster of events.
- Why survival? Attrition is reflective and while able to build hotbeds and heatmaps, it is hard to predict the upcoming WHERE and WHEN. Depends on mix of tenure and unique demographic interplay of each group. Shows the risk at each tenure interval – different for engineers versus retail for example
- Here is a real example. This shows the attrition curve variance for Gen Y and Gen X. Myth around Millennial departure rates. Don’t really materialize until 2 years in. How did we use this?
- Another case using compa-ratio and how much paid relative to market. High and low touted similar issues and gave us a new strategy for developing talent and pay
- We use this to do monte-carlo simulations and get a better probability distribution of our expectations. Here is an oversimple example using a company of 6 people. Using their tenure & survival curve for the next year, I can forecast a most likely expectation of 2 departures. I can use this to budget and forecast I can also intervene and change the survival outcomes. If I increase my compensation budget for the year and can boost everybody, I can change the distribution. If I offer a promotion to 2 of the 3 stagnant population, I can reduce the upcoming risk and mitigate the window.
- Now I’ll turn to a demo of the tool that we use. This shows how useful it can be to jump from demographic group to demographic group and the forecasted risk window. I can leverage any of demographic groups
- You can build your own dataset with a historical record of your tenure and attrition history where retention is an event of 1 and turnover is an event of 0. You can bring in any mix of variables to support your analysis.
- The model itself seems intimidating but it isn’t. The codification is also pretty simple and can be done in Excel. It can also be done with greater functionality in tools such as R, Stata, SPSS, Matlab or what you saw demonstrated in TIBCO Spotfire. If you constantly feed new data in from a growing and integrated dataset, the probability estimates are also always evolving (the benefit of going beyond an Excel solution).
- Here is a real example. This shows the attrition curve variance for Gen Y and Gen X. Myth around Millennial departure rates. Don’t really materialize until 2 years in. How did we use this?
- A critical element is often the external forces of the labor market. Being able to see how economic confidence shapes attrition likelihood does not mean you can mitigate it all but it can help with better forecasting and budgeting AND it can help identify what factors may push against growth economy forces toward turnover
- This experience is shaped most importantly by when somebody joins your organization and their experience and relative perspective