Behavioral Research

Focus on observed behaviors in real contexts. Use the Say‑Do Gap Diagnostic to choose methods. See Downloads.

TLDR: Behavioral research in the Behavioral Strategy context focuses on observing and measuring what people actually do (not what they say they’ll do), using mixed methods to validate problems, identify leverageable behaviors, and test interventions.

Term crosswalk for readers of the Overview article (see the Overview)

Older term Where it lives here
Situational Survey Context mapping and Environmental Audit
Behavioral Audit Behavioral Observation, Diaries, Time-motion, Behavior Calendar
Worldview Analysis Belief mapping, Emotional journey mapping
Problem Examination Behavioral Event Interviews focused on problems

Overview

Behavioral research differs fundamentally from traditional market research. While market research asks “What do you want?” and “Would you buy this?”, behavioral research asks “What do you actually do?” and “What prevents you from doing X?”

This guide provides practical methods for conducting behavioral research throughout the DRIVE Framework.

Example (enterprise data workflow): In a platform rollout for analytics teams, the target behavior was “Publish a certified dataset with complete metadata.” Observation and time‑motion studies showed the weakest step was writing complete metadata. The team prototyped a metadata wizard + templates, then field‑tested with 30 analysts. SMF was measured as completion rate and time‑to‑first publish; follow‑on bPMF tracked certified‑publish behavior in 30‑day acquisition cohorts.

Say‑Do Gap Diagnostic

Context stability Consequence salience Observability Method
High High High Log analysis, field observation
High Low Medium Diary + spot checks
Low Medium Low Event‑based interviews + rapid probes
Low Low Low Structured experiments

Core Principles of Behavioral Research

1. Actions Over Intentions

Traditional: “Would you use a fitness app?” Behavioral: “Show me your phone. Which apps do you open daily? When did you last try to exercise?”

2. Context Matters

Traditional: Survey in isolation Behavioral: Observe in natural environment where behavior occurs

3. Barriers Are Key

Traditional: “Why don’t you save money?” Behavioral: “Walk me through what happened the last time you tried to save”

4. Past Predicts Future

Traditional: “Will you change your behavior?” Behavioral: “How many times have you tried before? What happened?”

Research Methods by DRIVE Phase

Define Phase: Problem Validation Research

Method 1: Behavioral Event Interviews

Purpose: Understand when and how problems manifest

Protocol:

1. "Tell me about the last time you experienced [problem]"
2. "What exactly happened?" (probe for specifics)
3. "What did you do about it?" (actual behavior)
4. "How often does this happen?" (frequency)
5. "What have you tried before?" (solution history)

What to Listen For:

  • Specific behaviors indicating problem severity
  • Workarounds revealing unmet needs
  • Emotional language showing problem importance
  • Frequency and impact on daily life

Method 2: Problem Diary Studies

Purpose: Track problem occurrence in real-time

Implementation:

# Pseudocode for clarity
# Diary Study Protocol
diary_prompts = {
    "trigger": "When you experience [problem], log it immediately",
    "capture": [
        "Time and location",
        "What triggered it",
        "How you responded",
        "Impact on your day",
        "Attempted solutions"
    ],
    "duration": "7-14 days",
    "medium": "Mobile app, SMS, or paper",
    "incentive": "Small daily + completion bonus"
}

Analysis Framework:

  • Frequency patterns
  • Contextual triggers
  • Severity variations
  • Current coping behaviors

Research Phase: Behavior Identification

Method 3: Behavioral Observation

Purpose: See what people actually do vs. what they report

Observation Protocol:

  1. Shadow Sessions (with permission)
    • Follow users through their day
    • Note all behaviors related to problem space
    • Document context and environment
    • Avoid intervening or asking questions during
  2. Time-Motion Analysis 
    Behavior: [What they did]
Duration: [How long it took]
Frequency: [How often repeated]
Context: [Where/when/with whom]
Difficulty: [Observed struggle points]
Outcome: [What resulted]
  3. Environmental Audit
    • Physical barriers/enablers
    • Social influences present
    • Available resources
    • Competing behaviors

Method 4: Behavioral Mapping

Purpose: Identify all possible behaviors that could address the problem

Process:

graph TD
    A[Problem State] --> B{Current Behaviors}
    B --> C[Effective Behaviors]
    B --> D[Ineffective Behaviors]
    B --> E[No Action]
    
    C --> F[Why These Work?]
    D --> G[Why These Fail?]
    E --> H[What Prevents Action?]
    
    F --> I[Leverageable Behaviors]
    G --> I
    H --> I

Behavior Inventory Template:

behavior:
  name: "Specific behavior"
  current_adoption: "% who do this"
  effectiveness: "Problem-solving impact 0-10"
  
  requirements:
    ability_needed: "Skills required"
    motivation_type: "Intrinsic/extrinsic"
    environmental_needs: "Context requirements"
    time_investment: "Minutes/hours"
    
  barriers:
    - "Barrier 1"
    - "Barrier 2"
    
  enablers:
    - "Enabler 1"
    - "Enabler 2"

Integrate Phase: Solution Testing Research

Method 5: Rapid Behavioral Prototyping

Purpose: Test if solutions actually enable target behaviors

Protocol:

# Pseudocode for clarity
def rapid_behavior_test(solution_prototype, target_behavior, users, duration="3 days"):
    baseline = measure_current_behavior(users)
    introduce_prototype(users, solution_prototype)
    results = track_behavior(users, duration)
    return {
        'adoption_rate': results.tried_it / len(users),
        'success_rate': results.completed / results.tried_it,
        'sustained_rate': results.repeated / results.completed,
        'barriers_found': results.failure_reasons,
        'time_to_first_use': results.avg_time_to_try
    }

What Makes a Good Prototype:

  • Tests core behavior enablement
  • Minimal features (just enough)
  • Can deploy in <1 day
  • Measurable behavior change
  • Clear success criteria

Method 6: A/B Behavior Testing

Purpose: Compare behavioral interventions

Setup:

// Pseudocode for clarity
// A/B Test Configuration
const behaviorTest = {
  control: {
    name: "Current state",
    intervention: null,
    measurement: baselineBehaviorRate
  },
  variant_a: {
    name: "Ability-focused",
    intervention: "Simplify steps from 5 to 2",
    hypothesis: "Reducing complexity increases adoption"
  },
  variant_b: {
    name: "Motivation-focused", 
    intervention: "Add immediate reward",
    hypothesis: "Instant gratification drives behavior"
  },
  
  success_metrics: {
    primary: "% who perform target behavior",
    secondary: [
      "Time to first behavior",
      "Repetition rate",
      "Self-reported difficulty"
    ]
  },
  
  sample_size: calculateSampleSize(effect_size=0.2, power=0.8)
}

Verify Phase: Outcome Measurement

Method 7: Behavioral Analytics

Purpose: Track actual behavior at scale

Key Metrics:

# Pseudocode for clarity
behavioral_metrics = {
    'first_use_rate': 'Users who try behavior / Total users',
    'time_to_adopt': 'Average days from exposure to first use',
    'adoption_curve': 'Plot of cumulative adoption over time',
    'frequency': 'Behaviors per user per time period',
    'consistency': 'Standard deviation of behavior frequency',
    'streak_length': 'Consecutive periods with behavior',
    'completion_rate': 'Completed behaviors / Attempted behaviors', 
    'error_rate': 'Failed attempts / Total attempts',
    'time_per_behavior': 'Average duration to complete',
    'retention_curve': 'Users still active at T+30, T+60, T+90',
    'churn_points': 'When users typically stop',
    'reactivation_rate': 'Dormant users who return'
}

Behavioral Funnel Analysis:

Awareness → First Attempt → Successful Completion → Repetition → Habit Formation
   ↓             ↓                    ↓                   ↓              ↓
  85%           42%                  31%                 18%            12%
         
Where are users dropping off? Why?

Enhance Phase: Iterative Research

Method 8: Behavioral Cohort Analysis

Purpose: Understand how different user segments respond

Analysis Framework:

-- Behavioral Cohort Query Example
SELECT 
  user_segment,
  behavior_version,
  AVG(adoption_rate) as avg_adoption,
  AVG(retention_30d) as avg_retention,
  COUNT(DISTINCT user_id) as cohort_size
FROM behavioral_events
GROUP BY user_segment, behavior_version
ORDER BY avg_adoption DESC

Insights to Extract:

  • Which segments struggle most?
  • Which interventions work for whom?
  • Are there unexpected champion users?
  • What predicts long-term success?

Mixed Methods Research

Combining Qualitative and Quantitative

Research Stack:

layer_1_exploratory:
  methods: ["interviews", "observation", "diary studies"]
  purpose: "Discover behaviors and patterns"
  sample_size: "10-20 users"
  output: "Behavioral hypotheses"

layer_2_validation:
  methods: ["surveys", "analytics", "experiments"]
  purpose: "Validate at scale"
  sample_size: "100-1000 users"
  output: "Quantified behaviors"

layer_3_optimization:
  methods: ["A/B tests", "cohort analysis", "ML models"]
  purpose: "Optimize interventions"
  sample_size: "1000+ users"
  output: "Refined solutions"

Research Ethics and Bias

Stopping rules: Stop explorations after 3 consecutive sessions with no novel codes. For small effect detection (d=0.3), plan ≈ 175 per group (≈ 350 total) for A/B at α=0.05, power 0.8.

Consent: “We will observe your behavior related to X for Y minutes, record only these fields, and anonymize your data. Participation is voluntary and you can stop anytime.”

Ethical Considerations

  1. Informed Consent
    • Explain behavioral observation
    • Allow opt-out without penalty
    • Protect vulnerable populations
  2. Privacy Protection
    • Anonymize behavioral data
    • Secure storage protocols
    • Clear data retention policies
  3. Avoiding Harm
    • Don’t enable harmful behaviors
    • Consider long-term impacts
    • Provide support resources

Common Research Biases

Bias Description Mitigation
Social Desirability People report idealized behaviors Observe actual behavior
Hawthorne Effect Behavior changes when observed Extended observation periods
Selection Bias Wrong user sample Behavioral segmentation
Recency Bias Overweight recent events Longitudinal tracking
Confirmation Bias See what we expect Structured protocols

Tools and Resources

Research Toolkit

  1. Observation Tools
    • Behavior coding sheets
    • Time-stamped note apps
    • Screen/video recording (with consent)
  2. Analysis Software
    • Qualitative: NVivo, ATLAS.ti
    • Quantitative: R, Python, SPSS
    • Mixed: Dedoose, MAXQDA
  3. Behavioral Tracking
    • Analytics: Mixpanel, Amplitude
    • Heatmaps: Hotjar, FullStory
    • Custom: Build your own

Sample Size Calculations 

from statsmodels.stats.power import tt_ind_solve_power
import numpy as np

def calculate_sample_size(effect_size=0.5, power=0.8, significance=0.05):
    """
    Calculate sample size for behavior change detection
    
    effect_size: Expected behavior change (Cohen's d)
    power: Statistical power (typically 0.8)
    significance: Alpha level (typically 0.05)
    """
    n = tt_ind_solve_power(
        effect_size=effect_size,
        power=power,
        alpha=significance,
        ratio=1
    )
    
    return {
        'per_group': int(np.ceil(n)),
        'total': int(np.ceil(n * 2)),
        'note': f"Detect {effect_size} effect with {power*100}% power"
    }

# Example
sample = calculate_sample_size(effect_size=0.3)  # Small effect
print(f"Need {sample['total']} users to {sample['note']}")

Research Templates

Behavioral Research Plan Template

```