Brox.AI
Research Readout
Doc no. AI-2026-07
6 July 2026
Flagship study · Synthetic panel Auto insurance

Switch or Stay: what actually makes an insurance customer leave

We branched 424 synthetic customers into 16 memory-wiped versions of the same renewal — then rebuilt the letter six ways. The result: a clean, causal read on why customers churn, and the cheapest lever an insurer has to stop them.

Sector
Auto insurance
US renewal decision
Synthetic panel
424 twins
9,328 responses
Method
Counterfactual + Sequencing
two-phase design
Prepared by
Brox.AI
catrin@brox.ai

01Executive summary

The premium increase everyone worries about is only the third-biggest reason customers leave. What really moves them is a competitor in view and the memory of a badly-handled claim — and the single most common retention tactic, the “let’s review your options” call, does almost nothing.

Because every twin experienced all 16 renewal scenarios as a fresh, memory-wiped event, we can attribute switching to each factor cleanly — no confounds, no ordering bias, the same person in mutually exclusive worlds. Phase 2 then rebuilt the renewal letter six ways and found a retention win that costs nothing: change the wording, not the price.

3.1 → 7.8
Avg. likelihood of switching (out of 10): a calm renewal vs. the worst case
+2.23
Points a visible competitor adds to switch likelihood (/10) — the #1 driver
+0.14
Points a retention phone call adds to switch likelihood — effectively zero
1.28 pts
Gap in switch likelihood (/10) between the best- and worst-worded letter
How to read every number in this report

We asked each twin one core question: “How likely are you to switch insurer, on a scale of 0 to 10?” — where 0 = certain to stay and 10 = certain to leave. Every score below is an average of that answer. A figure like “+2.23” means that one factor raises the average score by 2.23 points out of 10. Lower scores = better retention.

02How we ran it — two engines, one panel

A single panel of 424 richly-seeded twins carried through both phases. The design does something conventional research physically cannot: it puts the same person into contradictory situations and measures the difference.

Phase 1
Counterfactual-Self Engine
Each twin ran all 16 renewal scenarios, memory wiped between every one. Five factors flipped on and off, so each factor’s pull on switching is isolated — a clean, Shapley-style causal read.
Phase 2
Sequencing & Priming
The renewal letter was split into four blocks and rebuilt six ways — three orders crossed with two price framings — each seen fresh, to find the wording that best holds a wavering customer.
Read-out
Attribution → message
Phase 1 finds the levers that matter; Phase 2 finds how to pull the biggest one without triggering the others. Switch intent captured 0–10 with a reasoned “why.”

The scale, in one line: every score is a customer’s self-rated likelihood of switching, 0 (definitely staying) to 10 (definitely leaving), averaged across the panel.

03Phase 1 — what drives switching

Starting from a calm renewal (a small 6% increase, no claims), the average customer rates their likelihood of switching at just 3.1 out of 10. We then measured how many points each factor adds to that score. Pile all five negatives on at once and it climbs to 7.8 out of 10.

Competitor quote in view
+2.23
Slow / bad prior claim
+2.16
22% price shock
+1.84
Retention call (“review options”)
+0.14
Loyalty recognition
−0.45

Each bar = the average number of points (out of 10) that factor adds to a customer’s likelihood of switching. So a visible competitor lifts it by 2.23; a bad claim by 2.16. Green = the only factor that lowers switching (by 0.45).

The three surprises

1

The retention call is a placebo — and can backfire

The classic “we’d like to review your options” call moved switch intent by +0.14 — i.e. slightly up. It fails to rescue a tempted customer (shock + competitor: 7.1 → 7.2 with a call) and fails to heal a claim wound (shock + claim: 7.1 → 7.2). The call mostly reminds people to shop.

2

A bad claim beats a price hike

The memory of a slow, badly-handled claim (+2.16) drives more switching than a 22% premium shock (+1.84). A bad claim plus a visible competitor — even with only a 6% increase — scores 7.8, nearly the full-blown storm. Service memory outweighs price.

3

Loyalty recognition is the only thing that holds people — but it’s weak, and lands on the wrong generation

Thanking a customer for their tenure is the sole lever that lowers switching (−0.45). It works hardest on Gen Z (−0.63) and least on Boomers (−0.43) — the opposite of the intuition that older customers reward loyalty most.

From a twin · bad claim

“It’s only a few bucks more, but after that claim dragging on and me having to keep calling, it makes me not trust them as much.”

Man · Gen X · switch 7/10
From a twin · retention call

“I’d at least hear the agent out, but I’m probably shopping it and leaning toward switching if the coverage is really the same.”

Woman · Gen X · switch 8/10

Peak churn (“the storm”) = price shock + bad claim + competitor in view = 8.0 / 10. Generational effects are otherwise modest: the drivers behave consistently across age groups.

04Phase 2 — how to say it

For the wavering customers, we rebuilt the renewal letter six ways: three orders (price-first, loyalty-first, value-first) crossed with two ways of stating the same $122 — as a loss (“$22 more”) or as a daily cost (“$4 a day”). Same facts, same price. Only the wording changed. The column below is the average switch likelihood each letter produced (out of 10) — lower means the letter kept more customers.

Renewal letter designPrice framingSwitch likelihood (/10)
Price-first“$4 a day” (daily)2.78Winner
Value-first“$4 a day” (daily)2.78
Loyalty-first“$4 a day” (daily)2.89
Price-first“$22 more” (loss)3.25
Loyalty-first“$22 more” (loss)4.00
Value-first“$22 more” (loss)4.06Danger
4

Don’t bury the price — lead with it

Price-first letters retained better (3.02) than leading with loyalty or value (3.4). The conventional “soften them up, price last” playbook is the danger zone: opening with value or loyalty and then landing “$22 more” produced the worst retention of all.

5

Reframe the number, not the order

Framing dominates ordering. The same $122 described as “$4 a day” instead of “$22 more” swings switch intent by nearly a full point. Total swing from message design alone — winner vs. danger — is 1.28 points (~13% of the scale), with zero change to the actual price.

From a twin · the danger letter

“I’d be kind of annoyed at the 22% increase, and I don’t really care about the accident forgiveness or roadside stuff.”

Woman · Gen X · value-first + loss frame · switch 6/10

05What it means — the retention playbook

Stop doing

  1. The generic “review your options” call. It doesn’t move retention and can prompt shopping.
  2. Burying the increase behind loyalty and value copy. It reads as being buttered up, and the hike lands harder.
  3. Framing the increase as a loss (“$22 more”). The most expensive sentence in the letter.

Start doing

  1. Fix the claims experience. A slow claim is the deepest, most durable driver of churn — deeper than price.
  2. Lead with the price, framed as a daily cost. The cheapest retention win available — a wording change.
  3. Reserve outreach for genuine competitor exposure, and make it substantive (a real match), not a check-in call.

The one-line story

Insurers over-invest in retention calls (useless) and in burying the price (backfires), and under-invest in claims experience and framing. The cheapest retention win on the table is a sentence: state the price first, as a daily cost.

Method note: switch intent was captured on a 0–10 scale. Effects are averaged across all 424 twins; the counterfactual design isolates each factor by holding the person constant and varying one thing at a time.