Trust Tax Index Pilot

[hola-asya]

Why?

Trust at work is one of the most consistently studied predictors of team performance in organisational psychology. Meta-analyses confirm that intrateam trust correlates positively with team performance (De Jong et al., 2016; ρ = .30), with the relationship strengthening under high task interdependence — a condition that characterises QA work by design. QA engineers sit at the intersection of every team boundary: they report defects to developers, flag risks to product owners, and raise quality concerns to leadership. Every one of those interactions is a trust transaction.

Yet despite the strength of the academic evidence, most organisations have no practical, data-anchored way to measure what low trust actually costs them in QA terms. This study addresses that gap by focusing on two observable, extractable outcomes that are directly shaped by trust dynamics: Defect Escape Rate — the proportion of defects that reach production unreported, driven by the suppression of bad news in low-trust environments — and Handoff Ping-Pong Rate — the frequency with which tickets bounce between developers and QA rather than being resolved through direct collaboration.

This is a pilot research proposal following #26

Expected results

We expect to answer the following research question:
Do software teams with higher team-level trust exhibit lower "transaction costs" — operationalised as measurable friction in everyday work tools — and better quality delivery outcomes, when controlling for relevant team and contextual factors?

We expect teams with higher trust scores will show:

1. Lower Defect Escape Rate — the ratio of production defects to total defects found. In low-trust environments, people suppress bad news and avoid reporting issues that might reflect poorly on them, allowing defects to escape to production.
2. Lower Handoff Ping-Pong Rate — fewer tickets bouncing 3+ times between roles, and a higher proportion resolved through direct collaboration rather than prolonged async comment chains.

Collaboration

Before the main study, we need 1–2 teams willing to participate in a pilot. The pilot will allow us to:

• Test the trust survey instrument and assess response quality
• Inspect the distribution, stability, and outlier profile of friction metrics over the prior 6 months
• Validate the distinction between trust-driven friction and legitimate process controls
• Refine metric definitions (especially Handoff Ping-Pong resolution classification)

What participation involves for a pilot team:

• Completing a short trust survey (~10 items, ~10 minutes)
• Granting read access to anonymised Jira/GitHub/Slack aggregate data for the past 6 months (@sharovatov need your input here)
• A brief interview or async Q&A to contextualise the data

All data will be handled at the team level. No individual identification will be published.
If your team meets the admission criteria and is open to contributing to practitioner-facing organisational research, please comment in this thread or reach out directly.

Target artifact

A Trust Tax Index — a practical method for teams and leaders to measure friction and convert it into hours and cost Heuristics to distinguish defensive controls from structural process practices + a conference talk

Background and additional information

Method of research

Sample: 30 teams across one or more organisations, subject to the following admission criteria:

Stable membership: ≥80% of members working together for at least 6 months
Team size: 3–15 members
Consistent metric collection in the same toolchain for ≥6 months
Available historical data

Trust Measurement: Mixed-referent items (combining "I" and "We" framing) using the Jarvenpaa et al. scale, identified as one of the strongest performers in a meta-analytical comparison of 46 trust measures (Feitosa et al., 2020; ρ = .39). The measurement will operate at both individual and team levels.
Friction Metrics: Extracted from Jira, GitHub, and Slack, then converted to hours to build a Trust Tax Index.
Controls: Team size, tenure, dependency load, work type, test automation coverage, co-location/timezone overlap, Jira workflow design, and cultural composition (individualistic vs. collectivistic context).
Analysis: Correlation and regression. Given the sample size, results will be interpreted with appropriate acknowledgment of statistical power limitations.

References

De Jong, B., Dirks, K., & Gillespie, N. (2016). Journal of Applied Psychology, 101(8), 1134–1150.
Feitosa, J., Grossman, R., Kramer, W. S., & Salas, E. (2020). Journal of Organizational Behavior, 41(5), 479–501.
Jarvenpaa, S. L., et al. — Trust scale, as reviewed in Feitosa et al. (2020).
Edmondson, A. C. (2004). Psychological safety, trust, and learning in organizations.
Bromiley, P. & Cummings, L. L. — Trust and Transaction Cost Economics.

[hola-asya]

next steps:

• @sharovatov to review the pilot proposal with a special attention to everything related to metrics and control variables
• @hola-asya to come up with survey design
• @sharovatov to suggest platfroms to aggregate the data
• @hola-asya @sharovatov to suggest teams to run a pilot. @anupamck @maryiat your involvement will be kindly appreciated at this stage too