Ground-Truth Triangulation
Combine statistical data with field conversations to detect and explain change
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 97%
Ground-Truth Triangulation starts with statistically sound data, then supplements it with structured conversations among people experiencing conditions firsthand. The data establishes scale and guards against overgeneralizing from one vivid story, while field evidence compensates for publication delays, revisions, and missing context. Repeated observations can reveal a turning point before it appears in official numbers, such as a recovery in spending or a retreat in demand. Conversations also explain why a metric is moving, helping decision-makers distinguish temporary implementation effects from durable changes. The output is not an anecdote-driven conclusion but a richer interpretation of the baseline, continually tested against diverse businesses, locations, and subsequent data releases.
Origin
Tom Barkin described using five days a week of business conversations across the Richmond Fed district to complement official economic data on Masters in Business.
Core principles
- 01Use broad data as the primary baseline
- 02Treat anecdotes as tests, not replacements, for data
- 03Use current field observations to detect turning points
- 04Investigate the mechanisms behind headline numbers
- 05Sample contrasting locations and business types
How to run it
- 1
Set the statistical baseline
Begin with the largest, most rigorous dataset available. Record what it says, when it was collected, and whether it is likely to be revised.
Pro tip Separate observed facts from forecasts before adding qualitative evidence.
Watch out Do not dismiss representative data merely because one interview contradicts it.
- 2
Map the blind spots
Identify where delayed reporting, aggregate averages, or missing context could conceal current changes. Turn those gaps into specific interview questions.
Pro tip Focus on turning points and causal explanations, where lagging data is weakest.
Watch out Avoid vague conversations that cannot test a defined hypothesis.
- 3
Sample the field broadly
Speak with people across different locations, organization sizes, and market positions. Ask comparable questions about demand, staffing, prices, and planned actions.
Pro tip Visit contrasting environments such as large cities, small towns, and different political regions.
Watch out Convenient contacts can create a systematically biased sample.
- 4
Test rather than replace
Compare recurring field signals with the statistical baseline. Treat anecdotes as evidence for interpreting or anticipating the data, not as permission to discard it.
Pro tip Give more weight to the same mechanism appearing independently across several sources.
Watch out A memorable story is not evidence of prevalence.
- 5
Explain the mechanism
Describe why the observed behavior should produce the measured or anticipated outcome. Specify timing effects, incentives, and operational constraints.
Pro tip Ask what decision a business has made and what would cause it to reverse that decision.
Watch out Correlation without a credible mechanism can produce a false turning-point call.
- 6
Validate over time
Compare the field-based interpretation with later data releases and revisions. Refine the questions and source mix based on where the assessment was accurate or wrong.
Pro tip Maintain a lightweight log of signals, expected timing, and eventual outcomes.
Watch out Do not quietly forget failed predictions.
In the wild
In May 2020, Barkin heard from a western Virginia real estate developer that newly reopened malls across the Tennessee border were packed. The observation did not replace official spending data, but it offered an early behavioral signal that reopening could unleash substantial pent-up demand before delayed statistics captured it.
→ The conversation provided an early indicator of the coming spending surge.
Furniture manufacturers reported that the pandemic-era household-furnishings boom was receding before the change became visible in aggregate data. Repeated industry conversations supplied both an earlier signal and an explanation rooted in normalizing household demand.
→ The field evidence identified a sector turning point ahead of official reporting.
Common mistakes
Letting one story override the sample
A single company's experience may be unusual. Require repetition across independent sources before changing the broader assessment.
Collecting only confirming anecdotes
Questions and source selection can reinforce an existing view. Deliberately seek businesses and locations likely to experience different conditions.
Reporting signals without mechanisms
An observation becomes more useful when it explains why behavior is changing and when that change should appear in the data.
Is it for you?
Best for
It is best for policymakers, executives, investors, and researchers making decisions in fast-changing environments.
Not ideal for
It is not ideal when field contacts are unrepresentative or when reliable real-time data already captures the relevant mechanism.
From the transcript
“the data is better than the anecdotes because you know it's a bigger sample.”
“But if you just count on the data, you're going to miss turning points and you're going to miss explanations.”
“I do try to understand what the data is telling me by testing, you know, with the conversations we're having in the field”
From the episode
Richmond Federal Reserve President and CEO Tom Barkin
Tom Barkin