DECAY.
Relationships don't ghost. They fade.
Nobody wakes up and decides to abandon a friend. It happens in the space between intentions. This is the research behind why relationships decay — and how Relay detects it before anyone notices.
Relationships are not binary. They don't end — they erode. The question is whether anyone's watching.
The research behind relationship decay.
Relationships follow predictable exponential decay.
Burt's landmark study tracked 345 bankers over four years and found that bridge relationships — the cross-group ties that carry the most social capital — have a catastrophically high decay rate. The longer a tie has survived, the more likely it persists, but new and weak ties vanish at astonishing rates. Decay slows over time (a "liability of newness" effect), but the default trajectory of any unattended relationship is dissolution.
Roberts and Dunbar confirmed this in a longitudinal study of students transitioning from high school to university: friendship quality declined measurably over 18 months. The decline was mitigated by effort — but effort meant different things for different people. For women, talking more frequently prevented decay. For men, doing activities together mattered more than talking.
You have 150 relationships. You actively maintain 5.
Dunbar's layered model — confirmed across mobile phone datasets of 35 million people and Facebook networks of 185,000+ users — shows human social networks organize into concentric circles: ~5 intimate ties, ~15 close friends, ~50 good friends, ~150 casual friends. Each layer requires progressively less contact frequency but still requires some maintenance. We devote roughly two-thirds of our available social time to just 15 people.
The implication for ministry: a volunteer with 5 students is already at cognitive capacity for their inner circle. The students in layer 2 (close but not closest) are the ones most vulnerable to decay — they're important enough to matter, but not close enough to get automatic attention.
Social isolation kills at the same rate as smoking 15 cigarettes a day.
A 2023 meta-analysis of 90 prospective cohort studies covering 2.2 million individuals found that social isolation was associated with a 32% increase in all-cause mortality, a 34% increase in cardiovascular mortality, and a 24% increase in cancer mortality. Loneliness — the subjective experience — independently contributed a 14% increase. These effects held across gender, geography, and follow-up length.
This isn't an elderly-only problem. The mortality signal was actually stronger in populations under 65. The students and young adults Relay serves are in the highest-risk window for social isolation to do lasting damage.
Two-thirds of international students experience loneliness. It's structural, not personal.
Sawir et al. identified three distinct forms of loneliness in international students: personal loneliness (loss of family contact), social loneliness (loss of networks), and cultural loneliness (absence of familiar cultural and linguistic environment). Cultural loneliness persists even when personal and social support is adequate — same-culture networks help but can't fully compensate.
Post-COVID, the gap widened. CCMH data from 188 U.S. counseling centers shows international students now present with significantly higher social isolation than domestic peers, and the disparity has grown since 2020. This isn't a "homesickness" problem. It's a structural deficit in the relational infrastructure universities provide.
The single best predictor of relationship survival is how often you show up.
Dunbar's 18-month longitudinal study found that emotional closeness tracks directly with contact frequency — and that the relationship between the two is not just correlational but causal. Increasing contact frequency actively prevented friendship decay during life transitions. The effect was strongest for relationships in the "sympathy group" layer (~15 people), exactly the tier where Relay's volunteers and students operate.
Critically, the type of contact matters differently by gender. But the universal finding is simple: relationships that lose regular contact decay. Relationships that maintain it survive. The technology that makes "showing up" easier — by reducing the cognitive load of remembering who to contact and when — directly addresses the mechanism of decay.
Relationship Decay Score (RDS): 0–100.
Six weighted attributes. Two penalty modifiers. Computed daily per volunteer–student pair. No surveys. No self-reporting. Just behavioral signals mapped to research.
| Attribute | Weight | Research basis |
|---|---|---|
| Recency — days since last contact, exponential decay | 25 pts | Burt (2000) decay functions; 60-day half-life |
| Frequency trend — contact rate, last 90 vs. prior 90 days | 25 pts | Roberts & Dunbar (2011) contact–closeness link |
| Reciprocity — who initiates, and how balanced | 20 pts | Dunbar (2016) call frequency reciprocity layers |
| Context breadth — distinct interaction types (text, meal, event…) | 15 pts | Dunbar (2015) activities vs. frequency gender effects |
| Emotional labor — conversation direction balance | 10 pts | Derived from Dunbar's maintenance cost framework |
| Milestone responsiveness — did key events get acknowledged | 5 pts | Holt-Lunstad (2015) social support quality indicators |
Penalty modifiers (up to −10 each): Network density shift — student losing connections even if one relationship looks active. Response latency creep — reply times stretching before frequency drops.
Thresholds & triggers.
| Score | Status | System action |
|---|---|---|
| 80–100 | Thriving | No action. Optional positive reinforcement. |
| 60–79 | Healthy | No action. |
| 40–59 | Cooling | "Check in soon" prompt to volunteer. |
| 20–39 | At risk | "This relationship needs attention" alert. |
| 0–19 | Disconnected | Flag for pastoral/staff follow-up. |
"No one falls through the cracks" is a lie we tell ourselves.
Every campus ministry says it. No one has the infrastructure to actually do it. The problem isn't care — it's cognition. Dunbar proved that humans can't maintain more than ~5 intimate relationships at full capacity. A volunteer with 8 students is already overloaded. Without a system that watches the gaps, the quietest student always loses.
Relay doesn't replace the relationship. It replaces the part of the volunteer's brain that should be tracking 8 relationships simultaneously and can't. The decay score is a mirror: it shows you what's already happening in the spaces between your best intentions.
"Absence makes the heart grow fonder" is a myth. In social networks, absence makes the heart grow forgetful. The data is unambiguous: unattended relationships decay. The only variable is how fast. — Derived from Bhattacharya, Ghosh, Monsivais, Dunbar & Kaski (2017)
Five new studies that confirm this matters now.
Negative social ties accelerate biological aging.
Using DNA methylation clocks and ego-centric network data, researchers found that "hasslers" — people in your close network who create stress — don't just feel bad. They measurably accelerate epigenetic aging and increase multimorbidity risk. The implication: relationship quality matters as much as relationship quantity. A decaying relationship that becomes one-sided or draining is actively harmful, not just neutral.
WHO declares social disconnection a global health priority.
The WHO Commission on Social Connection published its flagship report, estimating 871,000 deaths per year globally are attributable to loneliness (2014–2019). One in six people worldwide experienced loneliness between 2014 and 2023. Adolescents (20.9%) and young adults (17.4%) reported the highest rates — exactly the population Relay serves. The Lancet editorial called social health "the neglected third pillar" alongside physical and mental health.
34% of US adults 50–80 report lacking companionship. The pandemic didn't cause it — it revealed it.
Six waves of a nationally representative survey (2018–2024) showed loneliness was already at 34% before COVID, spiked to 41% during the pandemic, and returned to 33% by 2024. Social isolation followed the same pattern: 27% → 56% → 29%. The data confirms that disconnection is structural, not situational. It existed before the pandemic and persists after it.
Chronic loneliness affects 1 in 5 older adults. It compounds over time.
A meta-analysis distinguishing chronic loneliness (persistent over time) from transient loneliness found a prevalence of 20.8% among older adults. The key finding: chronic loneliness has worse health outcomes than episodic loneliness, and it's self-reinforcing — isolation breeds more isolation. Early detection and intervention during the decay window, before loneliness becomes chronic, is the highest-leverage moment.
Social isolation increases mortality by 35%. Living alone adds another 21%.
The most comprehensive meta-analysis to date on older adult mortality found social isolation carried a hazard ratio of 1.35 and living alone 1.21 for all-cause mortality. The effects were consistent across multiple countries (US, UK, China). The researchers called for social isolation to be treated as a clinical risk factor — not a lifestyle descriptor — comparable to hypertension or smoking.
Isn't this just a CRM for ministry?
CRMs track transactions. Relay tracks relationships. A CRM tells you the last time you emailed someone. Relay tells you that the relationship is cooling because contact narrowed to one channel, initiation became one-sided, and a milestone went unacknowledged — even though the email log looks "active." The decay score is a diagnostic, not a database.
Can you really measure relationship health without asking the student?
Not perfectly. But you can detect decay with high confidence using behavioral proxies — the same way a doctor reads vitals before asking "how do you feel." Recency, frequency, reciprocity, and context breadth are all observable without surveys. The research from Dunbar's lab confirms these are the primary predictors of relationship quality. We're not replacing the conversation. We're telling you when to have it.
What about privacy?
Relay doesn't read messages. It doesn't analyze content. It tracks three things: who contacted whom, what type of interaction, and who initiated. That's it. From those three data points across time, we derive the entire decay curve. The model runs on metadata, not meaning.
Why exponential decay instead of linear?
Because that's what the data shows. Burt (2000) and Dunbar (2011) both found that relationship quality drops steeply early and levels off later — exactly matching an exponential curve. A linear model would underestimate early-stage decay (the most dangerous period) and overestimate late-stage decay (where the relationship has already stabilized or died). The half-life model captures reality.
Does this work for non-student contexts?
The decay research is universal — Burt studied bankers, Dunbar studied high schoolers transitioning to university, Holt-Lunstad's mortality meta-analysis covered 2.2 million people of all ages. The score framework applies anywhere relationships matter and someone needs to be told "check in now." Churches, nonprofits, chaplaincy, mentoring programs, alumni networks.