How long do executive project dependencies remain unresolved? A cohort study
A decision-grade method for measuring dependency age, escalation, and resolution across leadership-sponsored projects.
Question and decision context
How long do recorded dependencies on leadership-sponsored projects remain open, and which ownership or escalation conditions accompany resolution? The study supports a bounded decision: whether to revise dependency intake, ownership, escalation timing, or project coordination capacity. It is not an employee scorecard and does not begin with a preferred staffing answer. Write the question, eligible systems, observation period, decision deadline, and authority before selecting records. The unit is one dependency entered in an authorized project register, which keeps anecdotes from replacing an auditable population. [1][2]
Population and inclusion rules
Freeze a consecutive cohort and include every eligible dependency entered in an authorized project register, including cancelled, incomplete, duplicated, corrected, and open cases. Record exclusions with reasons and an approving owner. Collect dependency, project, requesting owner, providing owner, opened time, needed-by date, risk, escalation, resolution, and downstream change. Name the authoritative system when timestamps conflict. GAO treats data reliability as fitness for intended use, so assess completeness, accuracy, and consistency for this precise question rather than declaring a whole platform reliable. [1]
Measures defined before review
Primary measures are eligible count, disposition count, elapsed calendar and working time, median, full range, and share completed before a stated useful date. Report open cases separately at cutoff. Missing times are not zero duration. Secondary measures may include clarification, reassignment, escalation, correction, and reopening. Define each term first and show counts beside percentages so small denominators never imply false certainty. [1]
Source lineage and quality checks
Create a field-level lineage table naming source system, record owner, extraction time, transformation, and limitation. Preserve source identifiers privately. Test ordinary, missing, long-duration, corrected, and exception cases against source records. A second authorized reviewer should independently classify a purposive sample. Record disagreement, adjudication, and changed rules, then rerun affected measures instead of silently editing the final table. [1][3]
Privacy and access
Apply purpose limitation and data minimization. Most analysis needs coded operational fields, not message bodies, attachments, participant notes, or privileged content. Assign a custodian, analysis owner, access list, and deletion date; log exports; and separate raw records from public copy. Support staff may collect and classify only within granted authority. Consequential executive, legal, employment, and security decisions stay with qualified owners. [4][3]
Descriptive analysis plan
Start with a flow statement covering eligible, excluded, analyzed, open, and incomplete cases. Present distributions and case categories rather than one average. A median should appear with range and denominator. Compare only groups defined before outcome review, such as complete versus incomplete intake or escalation within versus after an approved window. Label comparisons descriptive because complexity, authority, seasonality, and recording discipline can explain differences. [1]
Sensitivity and missing data
Repeat summaries using calendar versus working time, first versus final disposition, reopened cases separated versus included, and two defensible cutoff rules. If the decision changes, disclose that sensitivity prominently. Describe missingness by field, source, cohort, and case class. Do not invent precise event times from nearby activity. Bound what is knowable and show how many records support each result. Missingness can signal a control problem but not unrecorded behavior. [1]
Interpretation boundaries
The defensible result describes authorized records for the stated population, systems, definitions, and period. It can identify recurring exception classes, old queues, weak ownership evidence, or correction patterns. It cannot rank people, diagnose motivation, establish causation, predict a universal staffing ratio, or show that speed produced a better executive decision. Registers may omit informal dependencies, and speed cannot establish coordinator performance because complexity, authority, and constraints differ. Separate observed facts, analysis, operational inference, and remaining uncertainty. [1][5]
Controlled pilot design
If a material pattern survives quality and sensitivity checks, choose one reversible change tied to the decision. State the owner, authorized actions, start and stop dates, safeguards, success measures, and harm measures. Preserve a baseline and minimize simultaneous changes. Specify what support may prepare, route, remind, schedule, or escalate and what needs executive approval. BLS and O*NET task descriptions provide context, not permission expansion. [5][2][3]
Reporting protocol
Publish the question, period, population, inclusion rules, exclusions, sources, field definitions, quality checks, denominators, missingness, measures, sensitivities, and limitations. Include checked dates and links for authorities. Keep case data private and assess reidentification risk before using examples. Report facts first, analysis second, and recommendations last, with alternative explanations and evidence that would change the recommendation. [1][4]
Operational conclusion
For HireExecutiveTeam's executive support niche, this research is useful only when it supports a bounded operating choice, not a broad productivity claim. The question is whether clearer rules, better evidence, or authorized support can reduce avoidable coordination friction while preserving executive judgment, confidentiality, and accountable ownership. Run a time-bounded baseline and, if warranted, one pilot using unchanged definitions. Report improvements, harms, and deviations. [1][4]
Decision and follow-up
At review, choose to adopt, revise, stop, or extend the pilot and record who decided. Remove access no longer required. A later measurement should use the same population rule, field definitions, cutoff logic, and missing-data treatment; otherwise describe it as a new study rather than a trend. The conclusion remains local to this workflow and period until repeated evidence justifies a broader claim. [1][3]
Classification guide
Before extraction, create a coding guide with positive and negative examples for every status and exception class. Distinguish a request received from one accepted for action, a reminder from an escalation, and a provisional response from final disposition. Version the guide and date each rule change. When a case fits two classes, apply a written precedence rule and retain the secondary class for sensitivity analysis rather than letting reviewers choose whichever label improves the result. [1]
Workload and timing context
Interpret elapsed time alongside arrival volume, concurrent open work, stated priority, working hours, planned absences, and dependencies outside the support function. These contextual fields do not excuse missing ownership, but they prevent a queue measure from becoming a simplistic productivity judgment. Separate active handling time from waiting time only when reliable event records support that distinction. Otherwise report end-to-end elapsed time and explain that the record cannot allocate delay among attention, approval, dependency, and system latency. [1][2]
Adverse effects and stopping rules
Track possible harms during any pilot: inappropriate access, premature routing, missed high-risk exceptions, increased executive review, stakeholder confusion, rework, and displaced workload. Define a named owner and threshold for pausing the change. A faster median is not an improvement if sensitive information reaches the wrong person or consequential choices bypass accountable leaders. Record incidents and near misses in the same review window, while keeping security and privacy investigations in their authorized channels. [4][3]
Replication package
Preserve a privacy-safe replication package containing the protocol, blank extraction schema, data dictionary, coding guide, calculation formulas, validation checklist, and aggregate result tables. Do not include raw confidential records. Another authorized analyst should be able to reproduce totals from the approved analysis dataset and explain every exclusion. Record software versions and manual steps. Reproducibility reduces dependence on one analyst, but it does not repair biased source systems or expand the population to teams that were never observed. [1][4]
Sources
- Assessing Data Reliability, U.S. Government Accountability Office.
- Executive Secretaries and Executive Administrative Assistants, O*NET OnLine.
- Security and Privacy Controls for Information Systems and Organizations, National Institute of Standards and Technology.
- NIST Privacy Framework, National Institute of Standards and Technology.
- Secretaries and Administrative Assistants, U.S. Bureau of Labor Statistics, Occupational Outlook Handbook.
- Executive Secretaries and Executive Administrative Assistants, U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics.
- Business Formation Statistics, U.S. Census Bureau.
- Nonfarm Business Sector: Labor Productivity, Federal Reserve Economic Data.
- Productivity Statistics, OECD Data Explorer.
- World Development Indicators, World Bank DataBank.
- ILOSTAT Labour Statistics, International Labour Organization.
- 2024 Work Trend Index Annual Report, Microsoft and LinkedIn.
- The economic potential of generative AI, McKinsey Global Institute.
- Creating helpful, reliable, people first content, Google Search Central.
- Search Engine Optimization Starter Guide, Google Search Central.
- Dear Manager, You Are Holding Too Many Meetings, Harvard Business Review.