Research Paper 015
Why recruited partners don't produce, and what turns an intermediary into a default.
Companies count partners because partner count is easy to collect. Signed agents. Appointed brokers. Registered advisors.
Study 013 concluded that a recruited network is not a channel. This study looks at fifteen cases with disclosed production data and finds that claim is true in one regime and false in the other. Trupanion grew its distribution channel roughly fourfold between 2012 and 2021, and about three quarters of that came from adding partners. Counting them was correct.
Then active hospitals peaked in 2022 and the same metric stopped carrying information. Nothing in the number announced the change.
This study is about which regime you are in, and what moves an intermediary from a signature to a habit once recruitment stops working.
Fact means a filing, primary document, or on-the-record executive statement. Reported means a dated attributable claim in trade press or by a company about itself. Estimate means arithmetic derived here from disclosed figures, with the derivation shown.
Pattern means the mechanism repeats across cases with dated evidence in each. Heuristic means it is visible in one or two cases and is worth testing rather than assuming. Rule means it held everywhere it was tested and failed nowhere. Nothing in this study reaches rule. The closest thing across both studies is 013's finding that removing compensation below the clearing level collapses production, which is 013's result and not this one's.
Study 013 held that recruitment measures potential supply and production measures whether the channel exists. That is right in one regime. In the other, recruitment was the primary channel growth engine for nine consecutive years.
013 said: a recruited distributor network is not a producing channel, and partner count is a measure of supply rather than of output.
This study finds: Trupanion's disclosed thirteen-year series shows roughly 77% of its veterinary channel growth from 2012 to 2021 came from adding active hospitals rather than from raising output per hospital. During that period, partner count and production moved together. From 2022 onward they came apart and the channel produced less at the end of the period than at the start.
The correction: 013's framing describes the second regime and presents it as universal. The useful version tells an operator which position they are in. Below saturation, recruitment is the growth lever and count is a fair proxy. Above it, count keeps rising while output does not, and the metric gives no signal that anything changed.
Outcome: genuine correction. New evidence is stronger, not just different, because it is the same company disclosing both sides of the crossover.Everything after section 02 is about the second regime, where 013's original framing holds. The rest of this study reconciles five further points where the two papers use different vocabulary for the same mechanism or appear to disagree outright. Each is marked where it occurs.
Trupanion sells pet medical insurance through veterinary practices. Its 2024 shareholder letter, filed with the SEC as a DEFA14A in 2025, discloses the shape of its own distribution channel every year from 2012.
The table reports the average number of active hospitals, the average new pets per active hospital per month, the number of hospitals with its payment software installed, and the size of its field organization. Almost no company publishes any of this. Fact
For nine years, adding partners was the engine. Active hospitals went from 5,034 to 14,736 while output per hospital improved much more slowly. Anyone reading partner count during that period was reading a real signal.
Then active hospitals peaked at 15,952 in 2022 and fell to 15,269 by 2024. Output per hospital fell too, from 1.376 new pets per month in 2023 to 1.280 in 2024.
Nothing about the metric announced the change. The count kept being reported. It stopped carrying information about growth, and the company that had disclosed it most rigorously for a decade was the one it stopped working for.
Do not read a saturation threshold into this. Trupanion's ratios are specific to a channel with a fixed population of physical locations visited by a field organization. The transferable claim is narrower and it is stated below.
1 / Derivation. Trupanion, Inc., 2024 Shareholder Letter, filed as DEFA14A, 2025. Monthly channel flow is estimated as active hospitals multiplied by disclosed new pets per active hospital per month. 2012: 5,034 × 0.918 = 4,621. 2021: 14,736 × 1.260 = 18,567. Total growth 4.02x, of which count contributed 2.93x and productivity 1.372x. Taking logs, ln(2.93)/ln(4.02) = 77%. For the reversal: 2022 gives 15,952 × 1.332 = 21,248 and 2024 gives 15,269 × 1.280 = 19,544, a decline of 8.0%. Trupanion publishes no channel flow figure. This is our arithmetic on its disclosed series and should not be cited as a company metric.
013 separated recruitment, activation, and sustained production. This is the same distinction at higher resolution, with each state anchored to a case that discloses something about it.
Exits on awareness. Cheap, and no case in either study links awareness spend to output.
Exits on a low-friction application. Vertafore's 2025 survey found 77% of agents want better carrier onboarding and only 33% think their primary carriers do it well.
Trupanion's 2024 disclosure implies roughly 44% of the hospitals it visits every year refer nobody. This is where most of the population sits.
eXp Realty disclosed that 77% of departing agents had done zero to two transactions in the prior month. Arriving here is not the same as staying.
Goosehead's franchises past their first year produced 63% more per quarter than first-year franchises in Q4 2024.
Oasis Travel Network reports 95% preferred-supplier sales concentration. UWM obtained a written loyalty commitment from most of a 12,000-broker network in 2021.
Instruments that move an intermediary from State 0 to State 2.
Instruments that move State 2 to State 3 and State 3 to State 4.
Every item in the left column is easier to buy, launch, and report than every item in the right column. Primerica's mid-2024 convention incentives produced a record recruiting quarter with no movement in productivity per representative, which is the left column working exactly as designed on a metric that is not production. Fact 013
The second transaction is not the mechanism. It is the first cheap signal that the mechanism may be working.
Nothing here suggests transaction number two causes anything. What the evidence supports is that production rises with tenure, because each repeated use costs the intermediary less attention, less uncertainty, and less reputational exposure than the one before it. The second transaction matters because it is the earliest point at which you can tell whether tenure will accumulate at all.
Goosehead publishes a quarterly table splitting production per unit by tenure. It is the clearest public evidence in either study for how much a partner's output changes once the relationship is past its first year. Fact
Total operating franchises went from 1,226 at the end of 2023 to 1,103 at the end of 2024. New units under one year old were cut roughly in half, from 183 to 90. The company describes this as pruning underperforming franchises. Fact
Multiplying each tenure cohort by its disclosed productivity gives roughly $31.1m of franchise new business in Q4 2024 against $24.0m a year earlier. Written premium rose 27.7%. Est
Goosehead's productivity is denominated in commission dollars, and personal lines premiums were rising sharply. Some of the gain is rate rather than behavior. Public data does not separate the two, which is why this is a heuristic and not a pattern.
No company in either study's cohort publicly discloses first-to-second transaction conversion, time to first transaction, or dormant partner rates on a comparable basis. Every claim about the second transaction rests on proxies: tenure splits, activation ratios, offboarding thresholds, and the composition of partners who leave. Unresolved
The absence is itself informative. The number is not publicly disclosed. Partner count is. A metric that would tell an operator whether their channel is forming habits is absent from every public disclosure across seven industries, while partner count appears throughout them.
This study originally argued the weaker claim, that workflow placement is close to necessary and not sufficient. 013 sharpened it further and the sharper version is correct. Availability is a precondition. Ordering is what produced volume.
013's evidence is twenty years of federal rulemaking. Over 90% of travel agents were on an airline-owned reservation system by 1985. Every carrier was present in every system. What was litigated and regulated from the Civil Aeronautics Board rules of 1984 through the Department of Transportation readoption in 1992 and the survival of the display-bias prohibition in the 2004 deregulation was ordering. The residual advantage has a name in the literature, the halo effect. Fact 013
Outcome: same finding, different words. This study adopts 013's term and abandons the weaker phrasing.Necessary. Every airline was in every terminal for twenty years and it was not the thing that moved bookings.
The contested variable. Regulated for two decades because ordering inside a display the agent was already looking at moved enough volume to be an antitrust question.
Root entered a channel where every agent was already appointed with multiple national carriers. It launched inside comparative raters, the software agents open to quote a customer, and integrated with agency platforms rather than recruiting agents to a Root portal.
Where in the rater does Root appear? That is the position question and this study cannot answer it. Comparative raters return results the agent then selects from, and Root's own disclosure emphasizes pricing above every other lever. Its CEO described pricing on the Q4 2025 call as the factor that lifts every channel at once, and reported a pricing model improving estimated customer lifetime value by roughly 20%. If the raters Root sits in sort ascending by premium, then Root's pricing work is its position mechanism and the two claims are the same claim. Inference Unresolved
Two things would settle it. The default sort order of the specific raters Root has integrated with, and Root's bind rate by display position. Neither is public. Until one of them is, Root is evidence for presence and consistent with position, and this study should not claim more than that.
013 uses Trupanion as evidence for two things: a paid human layer that carries the explanation burden, and software embedded at the point of checkout. This study uses Trupanion's rising software installs against its falling channel output. A reader who knows 013 will think one paper is wrong. Neither is.
The two layers do different jobs. Territory Partners create the recommendation. The software removes friction from a recommendation that already exists. Between 2022 and 2024 Trupanion added 2,548 software installations and lost output, because the constraint had moved from transaction friction to the number of hospitals willing to raise the topic at all.
The supporting detail matters. Territory Partners and associates fell from 185 in 2023 to 171 in 2024, and output fell with them. The human layer shrank and production shrank. That is evidence for 013's explanation-layer finding, taken from the same table this study uses to bound the software claim. Fact
Outcome: both hold. The separating condition is which layer you are measuring.Intermediaries with a functioning default do not wake up looking for alternatives. Something has to make them willing to reconsider, and in every dated case here it was the incumbent. Pattern
The largest wholesale mortgage lender required its broker partners to stop sending loans to two named rivals, with liquidated damages, on an eleven-day deadline. The argument was not price. It was that one rival competed with brokers for the end customer.
$1.3 trillion across 3.6 million accounts at 7,000 registered investment advisors. Schwab's chief executive called the result imperfect at the firm's own conference that autumn. Advisors cited losing familiar technology, service difficulty, and competing with their own custodian for clients.
The largest travel advisor channel incumbent changed owner for $600m, was consolidated under Zurich Cover-More, and migrated its advisor experience from Agentlink to a rebranded replacement.
The minimum client asset level for referring clients out to advisory firms rose from $500,000 to $2 million.
A platform the intermediary did not choose to leave, on a date they did not pick. Present at Schwab and at Travel Guard.
Named by advisors during the Schwab conversion and by Schwab's own chief executive.
The whole of UWM's 2021 argument, and one of three reasons advisors gave for leaving Schwab. The most durable of the four.
That is the challenger's whole job before the window opens. Windows are dated events and they close as the acquirer's platform settles.
013's most useful output was that compensation behaves as a veto with a hard floor rather than a dial with a slope. This study found two cases that look like exceptions. They are not, and the reason is worth stating precisely.
013 found: below the level that clears the distributor's opportunity cost, production collapses fast and repeatably. Medicare Advantage plans that went non-commissionable lost enrollment within months on plans customers were actively choosing. Sunnova's dealers stopped work when payment stopped. Above the floor, 013 found no case of a durable step change from paying more. Aflac's contests reactivated lapsed producers during a specific disruption without moving its long-run producer count. Primerica's mid-2024 convention incentives produced a record recruiting quarter and no movement in productivity per representative. Fact 013
This study's two apparent exceptions both changed structure rather than level. Virgin Voyages pays 16% on voyage fare with no non-commissionable fares, which changes the base the percentage is computed on. Oasis Travel Network eliminates fees and returns 100% of commission to advisors above a stated preferred-supplier threshold, which changes the fee structure sitting between the supplier layer and the advisor. Neither raised a rate. Reported
The separating condition is matchability. A percentage is a level and a competitor can match it next week. A change to the base, or to the fees deducted before the intermediary is paid, requires the competitor to restructure its own business. 013 tested level. Rate structure is a different instrument and 013 did not test it.
What this study does not claim. Neither Virgin nor Oasis is a clean test, because both bundled the structure change with other instruments at the same time. Structure change is a candidate mechanism that neither study has isolated. It is not a demonstrated exception to 013's finding, and 013's finding on rate level stands unqualified. Unresolved
Outcome: both hold under different conditions. Level is settled. Structure is untested by either study.Two large surveys of independent insurance agents appear to disagree. They are answering different questions.
69% of personal lines agents named competitive pricing as the top factor in carrier selection.
In commercial lines the ordering changes: underwriting flexibility 39%, ease of doing business 35%, pricing 26%. Corporate Insight, Commercial Insurance Monitor, fielded Q4 2025, 214 independent agents, published 9 March 2026. Reported
75% of producers and account managers called claims service a must-have when placing business.
Personal relationships followed at 60%. Asked how carriers could win more business, 84% cited underwriting responsiveness. Vertafore survey of approximately 1,300 insurance professionals, reported in Carrier Management 25 Feb 2025 and Insurance Journal 7 April 2025. Reported
This is labeled a heuristic because both figures are stated preference rather than observed behavior, and neither survey publishes its instrument. It is consistent with 013's asymmetry finding. It is not independent confirmation of it.
Deloitte's 2026 middle-market insurance distribution survey adds a segmentation worth sitting with. It identifies a price-led agent segment, describes it as the most volatile of the groups studied, and reports that personal relationships with carrier staff ranked near the bottom of seventeen decision criteria for that group. The segment most responsive to a rate offer is by the same survey's description the least retainable one. Reported
2 / Derivation. On the Q3 2025 call, 6 Nov 2025, Root stated independent agents represented 50% of partnership distribution. On the Q4 2025 call, 25 Feb 2026, SVP of Business Development Jason Shapiro stated the partnerships channel was nearly half of overall new writings in the fourth quarter. 50% of approximately 50% gives roughly 25%. The two figures come from different quarters and Root has not published a combined number. Treat as directional.
An advisor, broker, agent, or veterinarian appears to make the decision. In some channels the layer above them has already made most of it. Heuristic
Oasis is a travel host agency. It reports travel insurance sales up roughly 40% from 2024 to 2025, with another 40% trending year to date in 2026. The figures come from the host's own post of 6 July 2026, attributed to Andria Fredrickson, VP of Insurance Operations and Advisor Acquisitions, and from Travel Weekly coverage of 13 July 2026. These are host-agency self-reports with no published method and no independent verification. Reported
Oasis attributes the increase to advisor education, support, and confidence. 013 found on fifteen years of Primerica data that training and enablement do not produce durable productivity change. Taken at face value these contradict.
They do not contradict, because Oasis changed four things at once. A preferred-supplier points program. Fee elimination and 100% commission retention above a production threshold, which is a compensation structure change. 95% preferred-supplier sales concentration. And a named person supplying escalation help when an advisor's client has a claim problem.
What can be claimed: a bundle of four instruments coincided with two consecutive years of roughly 40% growth in one product line at one host agency, self-reported. What cannot be claimed: that education did it. Education is confounded with three other instruments and cannot be decomposed from outside. 013's Primerica finding is the specific reason to be skeptical of the attribution the host itself offers. Unresolved
Outcome: no conflict. The draft's earlier education attribution was not supportable and has been withdrawn.One structural point does survive the confound. Preferred status is usually described as a reward for producing. At Oasis the incentive structure was built first, concentration followed, and production rose afterward. In this case preferred status operated upstream of production rather than downstream of it. Inference
013 calls it the paid explanation layer: a party other than the distributor, paid by the supplier, that carries the burden of explaining an unfamiliar product so the distributor will raise it at all. Trupanion's Territory Partners are the reference case.
This study kept describing something at Oasis that operates after the sale rather than before it, when the intermediary's client has a problem and the intermediary's own reputation is exposed. That is a distinct instrument and it needs a distinct name.
Post-sale escalation layer: a paid party that absorbs the intermediary's exposure after the transaction, where the paid explanation layer absorbs their burden before it. Same family, different timing, different function.
The Oasis evidence cannot tell us which one was operating, because the named executive supplied both product education and claims help. Unresolved
Outcome: 013's name adopted for the pre-sale variant. New name defined for the post-sale variant.Preferred status means very little when the list is long. Fora Travel publishes preferred partnerships with more than 5,000 suppliers in its own advisor education materials. In the insurance channel, Vertafore's 2025 survey found the largest single segment of agents works regularly with 11 to 20 carriers, and agencies over 250 employees interact with more than 50.
The institution matters where the list is short, where it controls the intermediary's economics, and where compliance with it is rewarded. Where any of the three is missing, the individual is still the unit.
Sorted by whether the metric predicted production in the cases where both numbers were visible.
The most underused metric in either study. Goosehead's step-ups are the measured size of it.
Trupanion and Root are close to the only companies publishing anything like it.
Survives the regime change. The metric companies switch to after count fails.
Strong below saturation. Misleading above it. Roughly 77% of Trupanion's channel growth to 2021 came from count. It then reversed.
Necessary, and on its own it describes the population most likely to leave. eXp disclosed 77% of departing agents had done zero to two transactions in the prior month.
No case in either study shows a portal causing production. Presence without position.
Altruist grew against Schwab with a narrower range. Root writes one line.
Real awareness instruments. No case links either to output.
013 found no durable productivity change from training and enablement on fifteen years of Primerica data. 013
Homepoint's 9,259 approved brokers transferred intact to an acquirer in a distress sale. A partner list is not a channel.
Goosehead's NPS fell from 92 to 89 and client retention from 86% to 84% in the same year written premium rose 27.7%. Fact
Strong for recruitment. Weak above the floor. 013 found no case of a durable step change from paying more. Rate structure is a separate instrument and is untested.
The satisfaction finding is the one most likely to be resisted, so it is worth restating plainly. Satisfaction scores measure how pleasant a supplier is to deal with. In at least one well-documented year that moved opposite to flow.
The sequence the evidence supports, with two changes from the version most channel programs are built around.
Cheap. Rarely the constraint.
A precondition for the first transaction, not a habit that forms after it. Presence is necessary. Ordering is what moved volume.
Opened by the incumbent in every dated case here. It closes.
Cheap to observe. Describes the population most likely to churn.
The intermediary lent their own credibility to it. Post-sale escalation lives here.
The first cheap signal that a habit may be forming.
Cannot be bought. Only accumulated.
The supplier reached for without shopping.
Not a habit that develops after adoption, and not merely availability. 013's airline record separates presence from ordering across twenty years of rulemaking.
Every dated displacement event in this study was created by an incumbent's consolidation, migration, policy change, or decision to compete with its own channel.
If you cannot produce this in an afternoon, that is the finding. Trupanion has reported an active-partner metric since 2014 and Root reports agent penetration separately from writings.
Estimate the addressable partner universe and your penetration of it. The answer decides whether recruitment is a growth lever. No channel's threshold transfers to another.
Name the tool the intermediary has open when they choose, then name where in that tool you appear. Every airline was in every terminal for twenty years.
Split production by how long each partner has been with you. Goosehead's split is 63% for franchises and 109% for corporate agents. No step-up means something is resetting the habit.
Phone calls, not dashboards. Highest information value per unit of effort on this list, and the only route to the number nobody publishes.
Fora publishes more than 5,000 preferred suppliers. Oasis reports 95% concentration. Those are different situations and only one of them has a customer above the advisor.
Track your incumbent's consolidations, migrations, commission changes, minimum changes, and any move that puts them in competition with their own channel.
Five places the evidence cuts against the argument above, including one where the two studies do not fully reconcile.
Roughly 77% of Trupanion's veterinary channel growth from 2012 to 2021 came from adding active hospitals. Employee Navigator grew its broker base from approximately 3,000 at the time of its Ease acquisition announcement in April 2023 to more than 7,000 by 2026, while the employers it served grew from 75,000 to 195,000, so production scaled slightly ahead of count. Both figures are company-stated and the definitions may not be consistent across the two dates. Reported
A company at low penetration that stops counting partners is making a mistake. This is section 01 restated as a limit on this study rather than as a correction to the last one.
Trupanion added 2,548 software installations between 2022 and 2024 while estimated channel output fell 8%. Installation appears to remove friction from a recommendation that already exists rather than causing the recommendation. Inference
Goosehead is one company, in one line of business, in one quarter, during a hardening personal lines market that inflates commission-denominated productivity independently of behavior. It is the best public evidence for the tenure effect and it is thin evidence. Labeled heuristic throughout, deliberately.
Goosehead's Net Promoter Score fell from 92 to 89 and client retention from 86% to 84% during a year when written premium rose 27.7% and franchise production rose about 29%. One year, one company. Enough to disqualify satisfaction as a proxy for flow, not enough to establish that they normally move in opposite directions. Fact
013 established that rate level above the clearing floor does not produce durable change. This study found two cases that changed rate structure instead, at Virgin Voyages and at Oasis, and neither is a clean test because both bundled the change with other instruments at the same moment.
Virgin's own community figures, from over 20,000 members to over 64,000, are company marketing materials rather than a disclosure and are not independently verified. Reported
This is the one place the two studies do not fully reconcile. Not a disagreement, an untested question sitting between them. It is recorded as unresolved in the methodology rather than resolved by argument. Unresolved
Berkshire Hathaway Travel Protection reopened direct sales to travel advisors on 1 October 2025, after seven years outside the channel. It is a useful test because the re-entry is recent enough that none of the outcomes are known yet.
Advisor channel penetration is still low, so recruitment is currently a legitimate growth lever and partner count is currently a reasonable growth metric. This follows from section 01 and reverses the position 013 would imply.
Zurich completed its acquisition of the largest advisor-channel incumbent on 3 December 2024 and migrated the advisor experience. BHTP hired a chief revenue officer out of that incumbent in 2024 and reopened the channel in October 2025. Fact
A mid-range percentage in a channel where rates run roughly 20% to 37% is matchable, and 013 found no case of a durable step change from paying more above the floor. The portal reaches parity with a competitor that shipped one in March 2023.
BHTP is already named as the errors-and-omissions partner of Fora Travel. Whether that extends to travel insurance preferred status is not established from public sources, and Fora's preferred list runs past 5,000 suppliers. Inference
Stated carefully: BHTP re-entered the channel during an incumbent consolidation and platform migration. We are not claiming the timing was deliberate and cannot establish it from outside. Unresolved What is not uncertain is that integration windows close as the acquirer's platform settles, and this one has been open for twenty months.
Advisors who have written at least one policy, divided by advisors appointed.
Output per advisor in months one to six against months seven to eighteen. A flat line means something is resetting the habit.
Nobody in either study publishes it. BHTP has it internally and it is the highest-information number it owns.
The intervention this research points to is a post-sale escalation layer for advisor-originated business, plausibly delivered through host agency relationships. It is the one instrument in the Oasis bundle that neither 013's Primerica finding nor 013's compensation finding rules out, and it operates on the transition a portal cannot touch. It is a candidate, not a proven mechanism. Inference