← Research · Renewals

The Upline Renewal Study 2026

Austin Boardman, Data Scientist · Published

The Upline Renewal Study points to a clear renewal policy for independent agencies: offer every customer a review before renewal, and shop the market when there is a reason to. Upline's data science team built 1,500 simulated personal lines policyholders and had each one complete the same renewal survey ten times, while the agency's offer, the customer's workload, and the messenger changed from profile to profile.

Key findings

  • Customers in the simulation wanted an annual review, with shopping when something changes. 89.2% of simulated responses chose "review my policy every year and compare carriers when there is a meaningful trigger." Only 4.6% chose shopping at every renewal.
  • A carrier comparison beat a review alone. Simulated policyholders rated their likelihood of accepting a renewal offer that included a carrier comparison at 8.4 out of 10, compared with 6.9 for a review of the current policy only.
  • Customer effort was the largest controllable factor. Asking for a longer application and document review lowered simulated acceptance from 8.7 to 6.7 out of 10. A five-minute confirmation barely moved it (8.5).
  • Price changes and claims experience flagged risk far better than demographics. In out-of-sample testing, operational signals explained 65% of the variation in simulated intent to leave. Demographics alone explained 8%.
  • Coverage built more trust than price alone. Offering to review price and coverage raised simulated trust in the agency to 4.15 out of 5, compared with 3.92 for a search for a lower price.
  • Cross-sell worked best when it started with the household's need. Coverage-gap framing drew simulated interest of 6.5 out of 10, compared with 5.8 for a savings pitch.
  • The size of the savings moved switching more than the carrier's name. Moving from 5% to 20% savings raised simulated willingness to switch by 2.1 points. The spread across 16 named carriers was 1.2 points.

Every figure in this report is an average from a simulation, not a measurement of real policyholders. Scores describe stated willingness on a scale, not the percentage of customers who would act. How to read this study explains what the method can and cannot tell an agency.

How to read this study

The Upline Renewal Study uses silicon sampling, an emerging research method where a language model answers a survey from the point of view of a specific, described person. Each simulated policyholder has a fixed profile: age, household, state, income, policies, how they buy insurance, how long they have been with their agency, and what happened at their last renewal. Academic studies have shown that this approach can reproduce some patterns in human survey responses, and also that it can distort others.

Agencies rarely get to test renewal decisions side by side before trying them on real customers. A simulation can. Each profile was randomly assigned a different version of the same renewal situation, so the study can compare how a carrier comparison performed against a review alone, or how a five-minute confirmation performed against a full application, with everything else held equal.

What the study can tell an agency is which renewal choices tended to land better and by roughly how much. What it cannot tell an agency is how many real customers will act, what an agency's retention rate will be, or what any one customer will decide. The 15,000 responses come from 1,500 profiles answering ten times each, not from 15,000 people, and the results have not yet been compared against a human survey. Read every number as a hypothesis worth testing with your own customers.

The findings

1. Should an agency shop every customer at every renewal?

In the simulation, policyholders wanted to be offered a review every year, and wanted the market shopped when something gave them a reason. Asked what an agency should normally do at renewal, 89.2% of simulated responses chose an annual review with a carrier comparison when there is a meaningful trigger: a price increase, a coverage concern, a household change, or the customer's own request. Only 4.6% chose shopping at every renewal.

That preference did not make customers cool on comparison itself. When each profile was shown one specific renewal offer, any version that included a carrier comparison clearly outperformed a review of the current policy alone.

Agency approach at renewal (randomized) Likelihood to accept (0–10) Effect on trust in the agency (1–5)
Compare carriers because a meaningful trigger is present 8.4 4.27
Automatically compare available carriers 8.4 4.20
Invite the customer to opt in before comparing 8.1 4.17
Review the current policy without outside quotes 6.9 3.61

Acceptance also rose with the size of the price change. Simulated policyholders whose premium stayed flat rated the offer 7.6 out of 10. Those facing a 25% increase rated it 8.3. The bigger the increase, the more the offer to shop was worth.

What to test: Reach out to every customer before renewal, and attach a reason when you offer to compare carriers.

2. How much work can an agency ask of a customer at renewal?

Very little, at first. Customer effort produced the largest swing of any agency choice in the study.

What the customer is asked to do Likelihood to accept (0–10) Effect on trust (1–5)
Nothing additional 8.7 4.20
A five-minute confirmation of their information 8.5 4.15
A longer application and document review 6.7 3.83

A quick confirmation cost almost nothing. A full application cost two points on a ten-point scale.

What to test: Ask only for what you need to know whether a better option exists. Collect full documents once the customer can see something worth moving to.

3. Should renewal outreach lead with price or coverage?

Lead with both. An offer to review price and coverage together raised simulated trust in the agency to 4.15 out of 5, compared with 3.92 for an offer to search for a lower price. It also scored higher on likelihood to accept, 8.1 versus 7.7.

When simulated policyholders described what would make them hesitate to switch, the most consistent concern was a difference in coverage, followed by an unfamiliar carrier and uncertainty about how claims would be handled. A lower premium alone does not answer any of those.

What to test: Present coverage, deductibles, and the claims experience alongside the price in every comparison.

4. Which customers are most likely to leave at renewal?

The customers who just had a hard renewal or a hard claim. Simulated policyholders whose last renewal premium rose by 20% or more rated their likelihood of leaving in the next 12 months at 5.6 out of 10, and those with a negative claims experience at 5.5. The average across all profiles was 4.3.

A big price increase or a bad claim flagged the highest simulated leave intentLikelihood of leaving in the next 12 months, by renewal history (0 to 10 scale)
  1. Last renewal premium rose 20% or more5.6
  2. Negative claims experience5.5
  3. Recently considered other quotes4.8
  4. Last renewal premium rose 10 to 19%4.8
  5. Mixed claims experience4.6
  6. Last renewal premium rose under 10%4.1
  7. Positive claims experience4.1
  8. Premium stayed the same3.8
  9. Premium decreased3.5

Descriptive averages for overlapping groups of simulated profiles. Not causal effects or observed retention rates.

Upline Renewal Study 2026 · retention_risk_segments.csv · 1,500 simulated profiles, repeat responses averaged per profile

What a customer has been through predicted simulated leave intent far better than who they are. In ten-fold out-of-sample testing, operational signals such as premium change, claims experience, recent shopping, tenure, and renewal contact explained 65% of the variation in leave intent. Demographics alone explained 8%. Differences between regions were negligible, a spread of 0.12 points.

These are descriptive averages for overlapping groups, not causal effects or observed retention rates.

What to test: Prioritize outreach using price change, claims history, and recent shopping activity. Use demographics to shape how you serve a customer, not whether you reach out.

5. How should an agent bring up additional coverage?

Start with the household's possible need, then ask permission. Framing an offer around protection against a possible coverage gap drew simulated interest of 6.5 out of 10, and asking permission to review whether a product fits drew 6.4. Leading with potential savings drew 5.8.

Review offered to an eligible household Interest in learning more (0–10)
Bundle review 7.1
Umbrella review 6.3
Life-needs review 6.1

Each product was offered to a different eligible group, so this is not a ranking of products against each other. One result stood out: auto bundle offers made to property-only households scored 1.75 out of 10. Those profiles did not say whether they owned a vehicle. The offer failed because nobody asked first.

What to test: Ask about vehicles, dependents, assets, and liability exposure before naming a product. Explain the need before the price.

6. What makes a customer willing to switch carriers?

The offer itself. Simulated willingness to switch to an alternative carrier rose from 5.5 out of 10 at 5% savings to 6.8 at 10% and 7.6 at 20%. Slightly broader coverage added 0.8 points over comparable limits and deductibles, and a strong recommendation from the agent added 0.4 points over a neutral presentation.

The study also named 16 carriers as the alternative. The gap between the highest and lowest carrier average was 1.2 points, smaller than the 2.1-point gap between 5% and 20% savings. Among simulated customers of independent agents, loyalty to the agency (6.2 out of 10) ran higher than loyalty to the carrier (5.7), which supports the agent's role as the person who evaluates the tradeoffs.

On how much savings is enough, responses clustered between $100 and $499 a year, but the exact threshold shifted under robustness checks. Treat it as a sign that a nominal saving is rarely enough, not as a universal number.

What to test: Explain the savings, the coverage, and your recommendation. The carrier's name matters less than the case you make for it.

7. Will customers accept AI in the renewal process?

Yes, when a licensed person is there for the decision. An agency-branded automated assistant with licensed-agent follow-up available raised simulated trust to 3.95 out of 5, close to 4.10 for a licensed agent making the same offer. Simulated policyholders generally wanted a person involved once their information was gathered or their options were ready, not necessarily at the first touch.

The exact split of channel preferences changed with how questions were worded, so the durable finding is the sequence, not a percentage.

What to test: Let automation handle outreach and information gathering. Keep the agent as the one who explains the options and makes the recommendation.

Putting it into practice

Most agencies already believe in reaching every customer before renewal. The hard part is time: tracking dates, collecting updated information, and shopping carrier portals for hundreds or thousands of households while service calls keep coming in.

Upline is a proactive account management platform for independent P&C agencies, built for the process this study describes. Upline helps you offer to shop every customer at every renewal with no additional work for your team. It monitors upcoming renewals, reaches out to customers in the agency's name, gathers what is needed for a review, and shops carriers according to the agency's own preferences. The agent reviews the comparison, explains the coverage, and makes the recommendation. Customers hear from their agent, and the agent gets the credit.

Start offering to shop every customer at every renewal at uplineinsurance.com.

Methodology and limitations

The Upline Renewal Study was designed and run by Upline's AI engineers and data scientists in November 2026. It is a hypothesis-generating simulation, not a survey of real policyholders.

Who was simulated

The study generated 1,500 personal lines policyholder profiles in the continental United States.

How the profile buys insurance Profiles
Through an independent agent 800
Through a single-carrier agent 250
Directly from a carrier 250
Online or other 200

Regional quotas were 38% South, 24% West, 21% Midwest, and 17% Northeast, with states drawn by 2020 Census population within each region. These quotas were design choices to give each group enough profiles to compare. They are not weights that represent the U.S. insurance market. Each profile also carried a fixed history: premium change at the last renewal, recent shopping, claims experience, tenure, and the kind of renewal contact they received.

What each profile answered

Every profile completed the full survey ten times, for 15,000 completed responses. Runs one through eight used the standard prompt. Run nine reversed the order of selected answer options, and run ten reworded the neutral instructions, to check whether answers depended on presentation. Results were averaged within each profile before comparison, so the unit of analysis is the profile, not the response.

The survey included randomized scenarios. Across profiles, the study varied the premium change, the agency's approach, the effort asked of the customer, how the offer was framed, who delivered it, how a cross-sell was framed, and the savings, coverage, recommendation, and named carrier in a competing quote. Because those conditions were assigned at random, the comparisons in findings 1, 2, 3, 5, 6, and 7 are like-for-like within the simulation. Finding 4 describes groups defined by profile history, so it shows associations rather than effects.

How the results were checked

  • A matched subset of 150 profiles was rerun on a second language model. The direction of the core findings held: comparison over review alone, low effort over paperwork, coverage and permission framing over savings, and larger savings raising willingness to switch. Exact channel preferences and the savings threshold were less stable.
  • Every batch was reviewed for parsing and quality issues. One truncated response was replaced with an audited rerun, and the original was retained.
  • A fresh rebuild of the analysis from the saved responses reproduced every published table exactly.

What this study cannot tell you

  • How real customers will behave. The simulation has not been calibrated against a human survey.
  • How common a view is. Percentages describe simulated responses, not the share of U.S. policyholders.
  • Retention or sales rates. A 7.6 out of 10 willingness to switch is not a 76% switch rate.
  • How real customers see specific carriers. Simulated familiarity with carrier brands mostly reflects what language models learned in training. This report does not publish carrier-by-carrier results for that reason.
  • Answers to items where responses did not vary. Nearly every simulated policyholder chose the same answer on contact timing and on reviewing details before switching. Those items are excluded.
  • Customer quotes. The survey's short written answers were generated by a model and are not quoted as customer testimony.

Frequently asked questions

What is the Upline Renewal Study? The Upline Renewal Study is a 2026 simulated policyholder study of how personal lines customers respond to the decisions an independent agency makes at renewal. It covers renewal outreach, comparison shopping, customer effort, retention risk, cross-sell, and carrier switching.

What is silicon sampling? Silicon sampling is a research method where a language model answers a survey from the perspective of a described person, such as a 60-year-old Texas homeowner with bundled home and auto coverage. Researchers use it to explore patterns quickly before running studies with people.

Is this a survey of real policyholders? No. The study simulated 1,500 policyholder profiles, each answering the same survey ten times. No real customers were surveyed, and the results have not yet been compared against a human survey.

Should independent agents shop every policy at every renewal? In the Upline Renewal Study, 89.2% of simulated responses preferred an annual review with carrier shopping when there is a reason, such as a price increase or a household change. Only 4.6% preferred shopping at every renewal. The study supports offering every customer a review and shopping when a trigger is present.

What makes customers turn down a renewal review? In the simulation, the amount of work asked of the customer mattered most. A longer application and document review lowered simulated acceptance from 8.7 to 6.7 out of 10, while a five-minute confirmation scored 8.5.

Which customers are most at risk of leaving at renewal? In the Upline Renewal Study, simulated policyholders with a 20% or larger premium increase at their last renewal, or a negative claims experience, reported the highest likelihood of leaving. Operational signals explained 65% of the variation in simulated leave intent, compared with 8% for demographics.

How should agents cross-sell at renewal? In the simulation, offers framed around a possible coverage gap drew more interest (6.5 out of 10) than offers framed around savings (5.8). Asking about the household before naming a product mattered most.

Do customers trust AI in the renewal process? In the Upline Renewal Study, an agency-branded automated assistant with a licensed agent available for follow-up scored 3.95 out of 5 on trust, close to 4.10 for a licensed agent. Simulated policyholders wanted a person involved when their options were ready.

Can I cite these findings? Yes. Please describe them as results from the Upline Renewal Study, a simulation of 1,500 policyholder profiles, rather than as a survey of policyholders.

Appendix: how to read the scales

All figures are averages across simulated profiles. Randomized comparisons are adjusted averages that account for the other scenario conditions each profile received.

Measure Survey question (paraphrased) Scale
Likelihood to accept How likely would you be to accept this renewal offer? 0 not at all likely to 10 extremely likely
Effect on trust Would this approach increase or decrease your trust in the agency? 1 decrease a lot, 3 no difference, 5 increase a lot
Leave intent How likely are you to leave your current agency or insurance arrangement in the next 12 months? 0 not at all likely to 10 extremely likely
Cross-sell interest How interested would you be in learning more? 0 not at all interested to 10 extremely interested
Willingness to switch How likely would you be to switch to the named alternative carrier? 0 not at all likely to 10 extremely likely
Loyalty How loyal are you to your current carrier, and to your agency? 0 to 10

Sources

Built with independent agents for independent agents.

Schedule your demo