QM 600 Prescriptive Analytics · Walsh College · Summer 2025 · team of 3

Remote work and productivity

A team study with Eswar Srinivas and Sukesh Anamaneni: do remote workers score higher on productivity and well-being, which work habits track productivity, and what a higher remote share would predict for self-reported productivity.

PythonpandasSciPyWelch t-testOLS regressionExcel scenario model
SituationPost-COVID, remote versus office arguments ran on anecdotes.
TaskTest what survey data actually shows about productivity and well-being.
ActionAnalyzed NSW remote-work surveys (3,019 responses) plus a 1,000-worker dataset with t-tests and regression.
ResultRemote scored higher (p < 0.0001) but explained little variance (R2 = 0.02), so the call was structured hybrid.

Live demo

Three tools from the project, recomputed in your browser. Change a control and the test, chart and prediction update.

1 · Remote vs in-office: Welch t-test

The row-level file behind this comparison is not in the project files, so the test is recomputed live from the group sizes, means and standard deviations recorded in my workbook (sheet Descriptive_RWP).

2 · Explore the 1,000-worker practice dataset

This file has no remote/in-office column, so compare the groups it does have. Pick any numeric column and filter by industry or age band.

3 · Remote share scenario (NSW Remote Working Survey)

Predictions use the per-year intercept and slope from my NSW_Regressions sheet. Outcome: respondents' self-reported productivity when working remotely compared with on-site (percent). With R² ≈ 0.02, remote share explains only about 2% of the differences between respondents, so treat the line as a small average tendency, not a forecast for any one team.

Results

Both charts are computed live from the 1,000-worker practice dataset.

Correlation with productivity score

Pearson r, all 1,000 workers. Task completion, calendar use, focus time and late-task ratio track the score almost perfectly, which fits a score generated from those columns.

Mean productivity by industry

Workers who use AI-assisted planning score higher in every industry in this dataset.

How it works

  1. Summarised productivity and well-being for remote and in-office employees (n = 490 and 510) and tested the gaps with Welch's t-test, which does not assume equal variances.
  2. Profiled a 1,000-worker practice dataset: a correlation matrix of work-pattern metrics against productivity score and an OLS model with 12 predictors.
  3. Regressed self-reported productivity change on the share of time worked remotely, separately for the 2020 and 2021 NSW Remote Working Survey waves.
  4. Built a scenario sheet that turns a chosen remote share into a predicted productivity change using those regression lines.
  5. Wrote prescriptive recommendations on remote policy, flagging that the survey relationship is statistically significant but small.

Data

  • remote_worker_productivity_1000.csv: A public-style practice dataset that appears to be synthetic:
  • remote_work_full_analysis_with_charts.xlsx: my analysis workbook (sheets Descriptive_RWP, T_Test_RWP, NSW2020_Summary, NSW2021_Summary, NSW_Regressions, Worker_Correlation, OLS_Remote_Worker, Scenario_Model). The remote vs in-office test and the NSW regression constants come from here.
  • NSW Remote Working Survey (New South Wales, 2020 and 2021 waves; 1,507 and 1,512 responses per my report): source of the remote share regressions.

First 10 rows of the 1,000-worker file (selected columns):