Behavioral Design Under Pressure

Curious about accessible marketing? Take a look at a selection of slides from my presentation at Digital Summit here, or keep reading below for the context and background on this topic—one I’m deeply passionate about. Digital Summit is a virtual conference that convenes industry professionals, thought leaders, and experts to explore emerging trends and best practices in digital marketing and technology. Von Restorff Effect

Crisis Context and Initial Approach

By July 2020, COVID-19 shutdowns had devastated businesses across the country for months, and the business insurance company I worked for faced an unprecedented challenge. State regulations were changing weekly; federal relief programs launched in waves; and business owners were drowning in existential uncertainty about coverage, liability, and survival itself. Our marketing team was tasked with creating an online assessment tool to help overwhelmed entrepreneurs understand their insurance options during the crisis.

With help from our Web and Design teams, we created the assessment. It consisted of 20 single-choice questions, one question per page, and was powered by logic that would crunch all of the answers users gave to provide them, at the end, with personalized coverage recommendations that could help their business survive the pandemic. The tool asked fundamental business questions around annual revenue, industry type, number of employees, property ownership status, and current location, i.e., standard inquiries we had previously seen our prospects respond well to.

Moreover, our methodology was grounded in three core UX principles:

1. Loss aversion

2. Authority-driven guidance

3. Urgency-based engagement

Loss aversion is the tendency for people to feel the pain of a loss more strongly than the pleasure of an equivalent gain. Here, we tapped into it by creating an experience that emphasized what users might lose if they didn’t take the assessment: in the short term, valuable insight, in the long term, perhaps their business itself. We provided authority-driven guidance through the tool, leveraging the likelihood that users would have authority bias: an above-average willingness to consider or comply with recommendations if they come from experts (which, in the insurance space, we were). And, of course, the assessment itself demanded and created urgency-based engagement. We used ‘crisis language’ to encourage people to use the tool: businesses were under extreme pressure, so business owners needed to get our tailored recommendations as fast as possible.

Original System Architecture

The assessment combined clean visual design, state-specific personalization, and branching technical infrastructure with gap-focused messaging positioned as definitive guidance. We segmented users by demographics and geography, delivered comprehensive risk assessments, and measured success through engagement metrics like completion rates and time spent—assuming these indicated purchase intent.

The methodology assumed linear user psychology during crisis: present authoritative information about risks → user acknowledges gaps → urgency drives action → qualified lead generated. Each component was designed to reinforce the others in a systematic cascade toward conversion, exactly as these psychological principles had worked in our pre-pandemic insurance campaigns.

We also briefed our Sales and Customer Success teams about the new assessment initiative, and they were very supportive of the approach. We specifically asked them to alert us if business owners who called them referenced the assessment, whether positively or negatively, so we could monitor real-world user experience beyond the digital analytics.

Where the Framework Buckled

The first structural failure occurred because we did not accurately assume or account for the psychological state of users who were entering our system to take the assessment. In creating and promoting the tool, we treated all crisis-affected business owners as if they would exhibit rational loss aversion and simply be ‘very concerned’ versions of our normal prospects. But that was not the case. We failed to consider the cognitive overload that business owners were experiencing en masse—juggling everything from sudden supply chain disruptions and revenue free fall to the brand-new yellow tape of PPP loan applications. This miscalculation caused several components to break down inside our framework.

Component-Level Breakdowns

The following analysis examines each core framework component to understand what we expected to happen versus what actually happened, along with the specific evidence that revealed our failures.

Loss Aversion Malfunction

• Design Intent: The tool started by asking business owners information about their current policy with us, and asking them to complete a few single-choice questions about their current situation and future goals. So we, channeling the Von Restorff effect, had ensured that any meaningful gaps in coverage would be visually obvious to users. While the tool’s online experience was intentionally ‘barebones’ and grayscale—straightforward and professional; we were an insurance carrier addressing our audience during a trying time—potential coverage gaps, based on the information users themselves had provided, were called out in red.

• Actual Result: Unfortunately, these gap callouts seemed to amplify anxiety in the minds of the business owners who were taking the assessment.

• Evidence: First, we noticed a statistically significant number of hard exits from the tool on pages where this callout feature was enabled. Secondly, the Customer Success team informed us they’d suddenly begun to experience an influx of ‘negative’ calls from business owners who referenced the online tool and seemed deeply concerned about their coverage.

• Root Cause: Our method assumed users had the cognitive bandwidth to process risk information constructively, not recognizing that stressed decision-makers exhibit “tunneling” behavior

Authority-Driven Guidance Failure

• Design Intent: In promoting the tool on various marketing channels, we touted our 100 years-plus of experience as an insurance carrier. We indicated how knowledgeable of, and compliant with, the ever-changing regulations of the mid-Covid economy we were. We positioned this brand-new online assessment as something not only extremely useful and unique (it seemed no other competitor had yet produced one) but also necessary for business owners—something it would almost be nonsensical for them to not take a few minutes to complete, given what all was at stake.

• Actual Result: The marketing might have been too heavy-handed. While the tool was indeed pretty thorough and impressive for how quickly it was created to help people work through very challenging circumstances, its promotion seemed to create unrealistic expectations.

• Evidence: During check-ins with our Sales team, account executives reported an uptick in prospects asking about questions far beyond both the scope of the tool and our scope as an insurance carrier—questions about cash flow management, supply chain alternatives and employee retention strategies, for example. While we never used data to form an explicit correlation between the launch of the tool and the observed uptick in these sorts of questions, Sales inf

• Root Cause: Authority signals during crisis situations can generate dependency rather than empowerment, causing users to expect comprehensive business guidance beyond our scope as an insurance provider

Urgency-Based Engagement Backfire

• Design Intent: Time pressure and crisis language would focus attention and accelerate decision-making using scarcity principles

• Actual Result: We triggered decision paralysis in business owners who were already overwhelmed by simultaneous urgent priorities

• Evidence: Web analytics showed substantial prospect abandonment at one question in particular:

  • “How many months can your business survive at current revenue levels before permanent closure?”

• Root Cause: Producing a sense of urgency in users who are already experiencing acute stress can compound their cognitive load rather than focus their attention or provoke a desired action

Methodology Diagnosis

Our fundamental methodological error was context blindness—applying behavioral design principles calibrated for stable market conditions to users who were experiencing unprecedented economic disruption. Restaurant owners facing permanent closure, retail stores navigating capacity restrictions, and service businesses losing 80% of revenue weren’t exhibiting standard insurance buying behavior. The framework probably would’ve worked well for baseline prospects, but it became actively harmful when applied to cognitively depleted users mid-crisis.

The methodology also suffered from variable conflation—treating engagement metrics (12-minute average session times, 340% higher completion rates) as indicators for purchase intent, when they actually could have pointed to how challenging the questions were or how desperate for information and advice the business owners were.

Deconstructed Component Analysis

What Worked (Preserved Elements)

• Personalization engine: State-specific regulatory guidance and industry-specific content delivery functioned properly and provided genuine value

• Cross-team feedback systems: Sales and Customer Success teams provided crucial real-world user experience data that our digital analytics couldn't capture, enabling us to detect problems quickly

• Technical infrastructure: Branching logic, real-time data integration, and responsive design performed as intended without technical failures

• Visual hierarchy and information architecture: Clean presentation and logical flow remained effective across user psychological states

What Failed (Discarded Elements)

• Authority positioning as comprehensive problem-solver: Promising definitive solutions and expert guidance beyond the scope of our insurance expertise

• Universal urgency application: One-size-fits-all crisis language regardless of business financial stability or user psychological state

What Needed Reconstruction (Modified Elements)

• User segmentation methodology: From demographic/firmographic categorization to psychological state-based differentiation (acute crisis vs. strategic adaptation vs. recovery planning)

• Information delivery framework: From comprehensive to incremental disclosure, based on user cognitive capacity and stress indicators

• Success metrics definition: From engagement quantity (completion rates, time spent) to outcome quality (user empowerment, decision confidence, actual problem resolution)

■

Rebuilt Method Architecture

The reconstructed methodology introduced adaptive frameworks that adjust based on user psychological state rather than applying universal behavioral triggers to crisis-affected business owners experiencing vastly different stress levels and decision-making capacity.

New Architectural Principles

Our rebuilt assessment expanded from 20 to 25 questions with a strategic structure: we kept the first five foundational business questions that any business owner would need to answer when seeking insurance guidance, then added five new psychological state assessment questions about recent layoffs, ability to pay essential expenses, time spent on crisis management, business survival expectations, and preferred decision-making approach. These psychological assessment questions would determine routing to persona-specific versions of the final 15 questions, which we preserved from the original assessment and covered day-to-day operations, goals, and specific use cases.

We knew that adding five questions would increase overall completion time, so we wouldn’t be able to evaluate “time spent” metrics in the same way. Therefore, we shifted our focus to quality indicators: decreasing hard exits (users clicking X rather than using the soft exit of save-and-continue), reducing negative Customer Success and Sales interactions that seemed influenced by the assessment, and increasing conversion rates among users who started the assessment. Even if fewer people finished overall, we wanted more people who began the process to complete it successfully.

1. Crisis-State Personas

For the rebuild we created three versions of the same final 15 questions, fundamentally asking about the same business information but with different levels of detail and directness appropriate to each user’s cognitive capacity. Similarly, while the underlying recommendation logic remained unchanged, we developed three versions of each result set—with the simplest guidance reserved for Survival Mode Sam and the most specific, detailed next steps designed for Recovery Mode Rachel.

• “Survival Mode Sam” (Acute Crisis): Restaurants facing closure, retail with 80% revenue loss—cognitive capacity severely limited by daily crisis management, requires simple binary guidance and immediate resource identification

• “Pivot Mode Paul” (Strategic Adaptation): Businesses successfully adapting operations—moderate cognitive capacity available for strategic thinking, can handle comprehensive analysis and planning

• “Recovery Mode Rachel” (Recovery Planning): Companies stabilizing after initial crisis—high cognitive capacity returning as operations normalize, ready for future-focused recommendations and growth protection emphasis

The vast majority of users during the crisis were Survival Mode Sams—business owners who might normally handle complex decisions very well but were now operating with limited cognitive capacity due to unprecedented, unplanned stressors that could prove fatal to their business. Most of our architectural changes focused on making the experience work (and feel) much better for this super-majority of users.

2. Cognitive Load Management

• Early Psychological Assessment: The rebuild used questions 6-10 to assess users’ psychological state and cognitive capacity, assigning them to one of the three new personas

• Persona-Specific Question Paths: Different personas would encounter different versions of the final 15 questions tailored to respect their cognitive capacity—more simply phrased questions for Survival Mode Sam vs. strategically phrased questions for Recovery Mode Rachel

• Exit Ramp: To prevent decision paralysis, the rebuild also newly provided users with an always-visible “save and continue later” feature, which enabled them to leave the assessment but keep their progress, so they could pick it back up when they were ready. A follow-up email with a deep link was sent three days later to everyone; once again eight days later to everyone minus Survival Mode Sams, whom we didn’t want to overwhelm; and once again (and finally) 14 days later only to Recovery Mode Rachels, who were least likely to feel burned out by the reminders

3. Support-First Positioning

• Resource Identification: On the assessment landing page that introduced the tool, we moved away from wording about “the unprecedented Covid-19 global crisis that is impacting businesses” to instead center all of the help and resources that were available, e.g., tax relief measures, grants for specific industries, the Paycheck Protection Program (PPP loans), and more

• Supportive Assessment Messaging: Throughout the assessment we integrated encouraging messages that reminded users there are no right or wrong answers, we’re here to help, they can stop and return anytime using the save feature, and they can complete it over the phone with Customer Success if that’s easier

• Realistic Timeline Acknowledgment: In our final recommendations (based on their answers) and follow-up emails (if they saved before completing), we softened our language from urgency-focused (“Take action today to ensure your business is protected”) to more reassuring and passive (“When you’re ready to expand your protection…”)—moving away from the implication that immediate action was needed and toward an understanding that temporary inaction is OK