Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Insurance Distribution Channel Cost and Acquisition Efficiency

Abstract. Insurance distribution incurs economic costs in finding, informing, assessing, converting, and retaining policyholders. These costs are distributed among insurance undertakings, agencies, brokers, banking networks, digital platforms, ancillary vendors, and customers. This monograph examines the measurement of acquisition efficiency across distribution arrangements, differentiating resource consumption from contractual remuneration, acquisition from ongoing service, and observed channel performance from causal channel contribution. It develops a comparative account of cost allocation, conversion attrition, risk and product composition, persistency, lifetime contribution, competition, and fair value. Established cost-accounting and causal-inference methods are applied to transparent hypothetical cohorts. Empirical studies of US property-liability insurance, German life insurers, Taiwanese bancassurance, and rural Indian rainfall insurance are interpreted within their original populations and evidentiary limits. International supervisory standards and European and United Kingdom rules illustrate the relationship between distribution expenditure and customer outcomes. The analytical synthesis establishes why a single acquisition-cost ranking cannot represent economic efficiency without specifying outputs, risk, service obligations, time, and the counterfactual. The study is an interpretive scientific synthesis, with no proprietary cohort dataset, independently replicated published estimates, or claim to a new theory.

Keywords: Insurance distribution; customer acquisition cost; insurance intermediaries; channel economics; activity-based costing; bancassurance; independent agents; insurance brokers; direct insurance; embedded insurance; distribution platforms; conversion; persistence; causal inference; profit efficiency; fair value; cohort analysis; customer welfare.

Introduction

Distribution efficiency is the relationship between resources committed to placing insurance and the economically and socially relevant outputs produced through that commitment. A distribution system generates contracts, information, access, service, and continuing relationships through which households and firms obtain risk transfer. Measurement starts from the particular transaction, the actors responsible for its execution, and the benefits the transaction produces during its life.

Insurance acquisition differs from an ordinary retail sale because the commodity includes a contingent contractual obligation. Payments received at inception may precede substantial claims by months or years. A successful sale can therefore have positive, negative, or still uncertain economic consequences for the insurer, distributor, and policyholder. The distributor may receive an immediate commission while the insurer assumes uncertain losses. The buyer may receive protection that is extremely valuable even when no claim is made. The economic value of the placement depends on whether the cover fits the risk, remains effective over the relevant period, and can be used when the insured event occurs.

A comparative study requires a defined output. A channel may be efficient in securing inexpensive inquiries but inefficient at converting those inquiries into valid coverage. It may perform well on first-year premium and poorly on persistency. An expensive professional placement can deliver a materially different contractual product from an inexpensive standardized online purchase. These differences cannot be eliminated by converting all expenditure into a single percentage of premium. They are part of the object of study.

The research proceeds through an interpretive review of selected insurance economics literature, regulatory primary materials, and established methods in managerial accounting and causal analysis. Institutional claims are limited to the jurisdictions and editions examined. Abstract-level access to several historical papers supports their stated findings without implying that original tables, robustness tests, or coefficients were independently reproduced. Numerical examples use invented data and currency units, not observations of firms or market benchmarks. The international comparison identifies mechanisms and differences in evidence rather than constructing an exhaustive ranking of countries. The research boundary excludes financial-reporting recognition of acquisition cash flows, actuarial premium expense loading, and a comprehensive legal treatment of remuneration disputes; those questions belong to related research.

The central inquiry is whether the channel produces valuable, persistent coverage at a resource cost justified by its incremental contribution and whether the comparison identifies an actual channel effect. The answer integrates operational accounting, insurance risk, consumer outcomes, and empirical identification.

Distribution Functions and Comparative Cost Taxonomy

Distribution as a production process

An insurance placement requires recognition of a potential exposure, communication of the product, collection of information, selection or design of coverage, a premium quotation, underwriting acceptance, contractual agreement, payment, documentation, and confirmation of effective policy status. For many contracts, a distributor also participates in endorsement, renewal, claims communication, or complaint resolution. Each activity can be performed by the insurer, an external intermediary, a customer, or a technology provider. The boundaries of the channel are determined by the allocation of these activities and obligations.

A channel is consequently a sequence of linked production decisions rather than a single marketing label. The same insurer may operate an online buying interface with human advisers in a remote contact center. A broker may obtain quotes through an insurer's automated infrastructure while supplying professional analysis outside that infrastructure. A bank may provide customer access and introduction, with a separate licensed insurance intermediary performing sales functions. Embedded sales may be initiated inside a partner application and completed through an insurer's system. The transaction record must represent these distinctions.

Activity classes include demand generation, lead validation, advice, exposure collection, underwriting support, payment processing, regulatory controls, policy issue, and ongoing servicing. Costs are incurred on successful and unsuccessful transactions. They include wages, employment benefits, facilities, purchased leads, paid media, software, vendor fees, intermediary payments, training, supervisory resources, security, quality assurance, and capacity maintained for uncertain demand. Analyses of a new channel also incorporate integration, migration, and termination costs. Expenditure classification follows economic function and the question under investigation.

Agency and brokerage

An exclusive agency can share a consistent carrier-specific workflow, training and brand with the insurer. These resources simplify some transactions while making the agency relatively dependent on the insurer's product menu and underwriting flexibility. Independent agencies and brokers may compare multiple insurers, negotiate coverage terms, and transfer business between suppliers; maintaining market knowledge, relationships, and documentation requires resources. The cost and value depend on the client's exposure complexity, the diversity of available insurers, and the functions performed under applicable contracts and law.

Regan's (1997) transaction-cost analysis examines independent agency incentives arising from control of customer relationships. Its proposed advantages relate to product complexity, uncertainty and the importance of relationship-specific investments, assessed in a historical US sample. An operational implication is that comparing two channels requires establishing which party maintains the relationship, pays for market search, and undertakes specialist analysis. Compensation reflects bargaining and ownership as well as work performed. The reported historical research provides a mechanism to investigate, rather than a contemporary causal estimate for every independent agency.

Brokerage frequently incurs considerable effort before a placement exists: solicitation of insurer appetite, negotiation, risk presentation, and failed bids. Several intermediaries can participate in the chain. Evaluating a retail broker solely from successful-policy commission revenue attributes the cost of unsuccessful market exploration to an incomplete output measure. Evaluating an insurer solely from the retail commission omits its own underwriting and administration resources. A system perspective reconciles both.

Bancassurance and institutional access

Bancassurance uses a banking relationship as a route to insurance. Existing transaction systems, branch facilities, customer interactions, and identity verification can support distribution when legally permissible. The relevant economic question is the incremental consumption of those resources, together with the opportunity cost of staff time and capacity. A bank branch is not a free resource merely because its principal business finances the premises.

A negotiated distribution arrangement may include exclusivity, upfront consideration, sales incentives, service payments, or technology commitments. Revenue and expenditure can occur at different entities and different times. A comparison that assigns all bank costs to deposit and lending activity will understate the resources devoted to insurance. Conversely, attributing a large proportion of the entire branch to a small insurance activity can overstate its incremental cost. Allocation should be sensitivity-tested, and decision-relevant opportunity costs should be reported separately from full cost.

Direct, embedded, and platform channels

Direct distribution includes business concluded through an insurer's own interfaces, employees or call centers. It may carry acquisition costs for brand advertising, paid search, software development, identity checks, telephone advice, retry handling, and abandonment. Advertising that supports multiple products and renewals requires a coherent allocation rule. Digital conversion creates a potentially low incremental transaction cost only within the available technical and operational capacity.

A comparison platform organizes access to quotations. Its economic services can include search, standardization, traffic acquisition and conversion support. Revenue can derive from leads, placements, subscriptions, sponsored visibility, or arrangements with insurers. A platform's cost per lead and the insurer's cost per bound policy are different metrics. The customer may independently visit multiple sites, creating repeated expenditure and overlapping attribution. An economically informative assessment includes lead duplication, invalid inquiries, the proportion of completed quote journeys, and whether the platform actually expanded access or shifted transactions from another route.

Embedded insurance uses a noninsurance transaction or relationship as the origin of the offer. Its economy may depend on relevance, transaction timing, shared interfaces and partner incentives. A low incremental cost for presenting the offer can coexist with significant contractual remuneration, data governance, integration maintenance, customer support, returns or cancellations. Measuring attach rate as policies divided by eligible host transactions clarifies exposure to the offer; it does not eliminate the need to track valid activation and persistence. Customer understanding, suitability, the ability to decline, and effective claims access remain outputs of interest.

Cost, price and economic transfers

From the insurer's perspective a commission is a payment and therefore a contractual expense. From the distributor's perspective it is revenue that finances resources and margins. In a consolidated economic-resource account covering both parties, the intercompany transfer is eliminated and the distributor's labor, systems, purchased inputs, and relevant capital costs are counted. Adding the entire commission to the entire underlying distributor resource cost double-counts the same transfer when computing total system resource use. In contrast, adding a commission to the insurer's internal expenses is appropriate for the insurer's standalone outlay.

A partner can possess valuable scarce distribution capacity. An economic assessment of channel integration should include the opportunity cost of that capacity even where there is no separate invoice. A contractual price reflects bargaining, scarcity, risk transfer and profit, and need not equal measurable resource use. These three perspectives—payer outlay, recipient revenue and resource cost—should be maintained as separate ledgers. For a customer price and benefit assessment, the full amount charged remains relevant even if it consists partly of transfers within the chain.

Time also changes the meaning of cost. A platform integration fee may be fixed at inception, a lead fee per inquiry variable, a minimum contractual commitment fixed over a term, and a technology license step-fixed above a volume threshold. Historical average costs describe an achieved scale; prospective marginal costs depend on available capacity and contractual conditions. The cost of abandoning a channel includes obligations that survive migration. These distinctions allow acquisition performance to be interpreted within the actual production system.

Cost Denominators, Allocation and Service Boundaries

Definition of the acquisition cohort

The cohort is a defined set of eligible prospects entering a specified acquisition opportunity during a stated period. The eligibility gate precedes the outcomes under examination. A reproducible cohort definition establishes product version, jurisdiction, new versus existing customer, quote date, effective date, policy duration, exposure type, channel episode, and the window allowed for conversion. The analyst specifies whether the unit is a person, household, firm, risk, quote, contract, policy term, or coverage certificate. Group insurance especially requires this precision: one master contract may cover thousands of individuals.

For a cohort with acquisition expenditure C, valid bound policies B, eligible leads L and initial written premium P, elementary measures are C/B, C/L and C/P. An aggregate number without a denominator definition cannot be interpreted. C/B measures expense per valid placement; C/L measures expense per prospect opportunity; C/P expresses cost relative to the premium base. The last measure is dimensionless and may be expressed as a percentage. When P includes taxes, levies, or third-party items, it should be reported as gross or adjusted and made consistent across channels. A zero denominator produces an undefined ratio, not zero cost.

A risk exposure denominator provides a further view. For automobile insurance, it might be earned vehicle-years; for property, insured locations or a correctly defined amount at risk; for life insurance, an appropriately specified policy-year or insured amount. It is inappropriate to pool heterogeneous exposure units without a defensible weighting scheme. Exposure adjustment is valuable because a high-premium risk can make acquisition cost appear modest as a percentage even when its loss obligations are substantial.

Funnel reconciliation

Let leads enter a process with a quote rate q, bind rate b conditional on quote, and activation probability a conditional on binding. On a mutually exclusive reconciled cohort, expected active policies equal L × q × b × a. If acquisition expenditure is C, the expected ratio using the expected activation count is C/(Lqab); it is not the mathematical expectation of the realized random ratio. Realized active-policy cost is C divided by the actual number activated.

A hypothetical cohort illustrates the distinction. One thousand eligible inquiries cost 20,000 currency units. Six hundred obtain a complete quote, 180 bind, and 162 activate. Costs are 20 per inquiry, 33.33 per quote, 111.11 per bind, and 123.46 per activated policy. The activated-policy cost exceeds the bound-policy cost by 12.35. If cost items are deferred until binding, the expenditure used in each ratio must still encompass all costs within the defined acquisition window. The cancellation reasons are necessary to interpret the difference: failed payment, underwriting reconsideration, customer preference, duplicate records and administrative error represent different mechanisms.

The funnel must admit competing outcomes. A quote can be declined by an underwriter, abandoned by a customer, converted through another channel, or remain unresolved at the reporting cutoff. Excluding difficult cases because they did not reach quotation creates an artificial efficiency advantage. Cohorts should be closed only after the specified attribution period; immature observations require censoring-aware treatment. A customer who submits four quotes should not be counted as four independently acquired customers.

Activity-based resource allocation

Traditional expense allocation assigns common costs according to a basis such as headcount, premium, policy counts or labor hours. The basis should represent the process consuming the resource. Premium may reasonably allocate some costs connected to transaction value, but it is generally weak as a direct proxy for the time required to check customer information. High-value risks can require very different work even at identical premiums.

Kaplan and Anderson's (2004) time-driven activity-based costing estimates the resource cost per time unit and the time demanded by transactions. For an insurance sales team with resources costing 120,000 per quarter and practical productive capacity of 60,000 minutes, the capacity cost rate is 2 per minute. An ordinary quote requiring eight minutes absorbs 16; a complex quotation requiring 35 minutes absorbs 70. If only 45,000 minutes are consumed, 90,000 is allocated to observed activities and 30,000 remains the cost of unused capacity. That residual must remain visible rather than being arbitrarily assigned to completed policies. Practical capacity is an estimate and should be supported by staffing and process observations.

Activities with substantial automated costs require additional drivers. Document verification can have a per-call software fee, a fixed infrastructure component, and a human exception cost. Marketing can be charged per impression or lead while brand investment influences multiple cohorts. Claims-service infrastructure may serve policyholders originated through several channels. A combination of direct tagging, time-based allocation, shared-resource drivers and residual capacity reporting preserves an auditable account.

The total ledger should reconcile to recorded expenditure at the agreed boundary. Errors emerge when a head-office invoice is omitted, partner integration is assigned entirely to renewal activity, or duplicated platform payments are recorded as unique acquisition resources. Transfer elimination, amortization policy, inflation adjustments, and foreign exchange conversion require documentation. Alternative cost allocation conventions should be shown when their plausible range changes channel rankings.

Acquisition and ongoing service

Some activities occur before contract issue, others after it, and some are performed continuously. An intermediary may earn a first-year commission partly to compensate for future service. Renewal remuneration may partly reward the original acquisition relationship. Timing alone cannot identify the economic function. Separate activity-based attribution is preferable when reliable records exist; otherwise the allocation assumption and uncertainty must be reported.

Standalone acquisition cost is useful for monitoring the sales process. Full relationship cost is useful for assessing durable channel contribution. The relationship ledger includes endorsements, renewal outreach, service inquiries, complaint remediation, cancellations, collection costs and relevant claims assistance. Claims indemnity belongs to the risk-bearing outcome ledger rather than the acquisition ledger. Where a distributor delivers claims assistance, the labor and technology supporting that assistance remain a distribution service cost. These boundaries prevent the same function from being charged twice while preserving the full analysis of profitability.

Channel Selection, Conversion and Customer Mix

Selection mechanisms

Customer and product characteristics jointly determine channel access, willingness to engage, quotation probability and binding. Customers seeking a standardized policy may prefer self-service; commercial buyers with heterogeneous exposures may need professional negotiation. Risk appetite, language access, accessibility, geography, previous relationships and payment capacity affect which customers enter a channel. Insurers can route applications based on the expected complexity of the case. These conditions make observed channel populations different even before acquisition begins.

A portfolio of easier risks can appear inexpensively acquired because cases require less labor and meet underwriting rules more frequently. A portfolio of complex risks can require additional information, yet its higher resource use may be appropriate to the work. Measuring raw cost per policy without describing the cohort confounds resource efficiency with selection into the distribution process. A fair descriptive comparison first stratifies pre-channel characteristics and identifies the populations with plausible access to alternative routes.

Product and channel design are often jointly chosen. In embedded insurance, a narrowly defined cover may be packaged differently from a standalone policy. A bank may sell a policy with particular benefit limits, term structure or access conditions. Price differences cannot be attributed to distribution without checking coverage, risk classification, underwriting, contractual exclusions and services. Holding the product fixed isolates one useful channel comparison; evaluating the channel-plus-product bundle answers a different question.

Conversion, price and effort

Conversion reflects the suitability of the offer, the terms, the price, the customer's resources and trust, and the process required to reach agreement. Customers may reasonably decide against coverage. A conversion target that counts every refusal as process failure can incentivize pressure on inappropriate customers and distort information quality. The economically relevant objective is effective placement among people for whom the product is appropriate, subject to the customer's informed choice.

Cole and colleagues' (2013) rainfall insurance experiments investigated why rural Indian households adopted an unfamiliar risk-management product at low rates. The institutional working-paper summary and published abstract identify experimental variation in price and nonprice factors. This evidence makes the distinction between prospective demand and acquisition process concrete. It concerns a specific rainfall product and historical setting; it cannot supply a universal direct-versus-agent conversion parameter. For distribution research, it illustrates why access, trust and constraints must be measured alongside price and why observational comparisons can miss demand-side mechanisms.

Conversion analysis should use common time windows, deduplicate opportunities, and retain unsuccessful outcomes. For a campaign claiming substantial incrementality, the untreated baseline is essential. An attributed sale may involve a customer who would have bought anyway. Paid advertising may divert organic traffic. A referral channel can reach a genuinely new buyer or simply capture a transaction another route would have completed. Descriptive conversion is meaningful for capacity and workflow planning, whereas economic incrementality requires a counterfactual.

Composition and aggregation

Consider hypothetical direct and adviser-assisted channels selling simple and complex policies. The direct channel places 900 simple contracts at 80 each in acquisition resources and 100 complex contracts at 300 each. Its aggregate acquisition cost is 102,000, or 102 per placement. The adviser-assisted channel places 100 simple contracts at 70 each and 900 complex contracts at 250 each, using 232,000, or 232 per placement. The adviser channel is less costly within each product stratum but appears more costly overall because it handles the more complex population.

Under common weights of one-half simple and one-half complex policies, the standardized costs become 190 for direct and 160 for adviser-assisted distribution. The reversal is purely arithmetic and illustrates the risk of unadjusted channel rankings. It does not establish that all agent systems are operationally superior. Standardization uses a declared target population; it may be infeasible where a channel never handles a category. Reporting the raw and standardized results together describes both the actual portfolio and the controlled comparison.

Selection also influences claim experience. Customers gravitating to specific channels can differ in unobserved risk or information. The channel may itself improve risk presentation or influence the policy selected. If claims are used as an efficiency outcome, losses need to be evaluated under comparable definitions, maturity and exposure. An unusually high loss ratio can reflect risk composition, pricing, random variation, or protection that is delivered as intended. These explanations require different evidence.

Cohort Unit Economics and Persistence Adjustment

Contribution and discounting

Acquisition expense occurs early, while insurance service and loss obligations unfold through time. For an initial cohort of N valid policies, let S_t be the fraction expected to be in force at the beginning of period t, M_t the expected incremental contribution per in-force policy during that period conditional on its survival, and r a consistent per-period discount rate. Specify the payment timing, and define M_t as premium revenue less expected claims, claims-handling expenses, applicable reinsurance net costs, ordinary servicing costs, and other included variable obligations. A capital cost can be included if it is economically appropriate and does not duplicate the discount-rate treatment.

With inception acquisition cost C, net present contribution on a simplified end-of-period timing basis is

V=−C+N∑t=1TStMt(1+r)t.V=-C+N\sum_{t=1}^{T}\frac{S_t M_t}{(1+r)^t}.

The valuation is conditional on the horizon, retention, claim expectations, and resource boundary. If commission is included in C, that same upfront commission must not also be deducted in M_t. Renewal commission belongs in the relevant period unless already incorporated into another expense item. Contract liabilities and insurer financial-reporting revenue are distinct from this simplified prospective economic cash-flow calculation.

Persistence is the continued existence of eligible coverage. It may be measured as policies remaining in force, premium remaining, or customers retaining at least one product. Each definition answers a different question. Renewals are opportunities for policy continuation; an unchanged customer count can mask substantial shifts in coverage. Cancellation, lapse, claim termination and replacement with another product should be distinguished. Short-term policies can terminate normally without implying failure to retain.

Retention must be modeled as a cohort outcome, with the possibility of right censoring for recent contracts. A simple discrete-time rate can be calculated among policies at risk of lapse at the beginning of each interval. Competing risks matter when a policy ends for reasons unrelated to dissatisfaction. A cross-sectional count of in-force policies divided by this year's new business is not a persistence probability.

A transparent two-channel example

The following calculation is synthetic. Each route acquires 1,000 valid policies for a comparable annual contract. Route A incurs 100,000 in acquisition resources and Route B 145,000. Both deliver first-year expected contribution of 90 per initially active policy after expected claims and included service costs. At the start of year two, the respective survival fractions are 0.65 and 0.85; the year-two contribution is 100 per surviving contract. The discount rate is 5 percent; there are no cash flows beyond year two or within-year lapses in this example.

Metric Route A Route B
Inception acquisition cost per policy 100 145
Active year-two policies 650 850
Year-one contribution 90,000 90,000
Year-two contribution 65,000 85,000
Discounted contribution before acquisition 144,671.20 162,811.79
Net present contribution 44,671.20 17,811.79

The additional retention of Route B contributes 20,000 in year two before discounting, or 18,140.59 in present value. That benefit does not cover the extra 45,000 paid at inception. Route A yields 26,859.41 more net present contribution under these assumptions. A justified premium for higher servicing quality therefore depends on the actual amount and timing of the value generated. Longer continuation, richer service benefits, lower losses, or a different customer outcome could alter the conclusion.

For Route B to tie Route A solely through its year-two contribution, the required increment is 45,000 × 1.05² = 49,612.50. Given 200 additional surviving policies, the necessary year-two contribution increment per additional survivor is 248.0625 if all other contributions are unchanged. Alternatively, with a contribution of 100 and all other conditions fixed, additional retained policies would have to number 496.125 in expectation, exceeding the feasible 350 extra survivors above Route A's 650. Under the specified horizon and equal conditional margins, persistence alone cannot offset the acquisition gap.

The comparison does not value protection separately from the insurer's margin, so it cannot establish which route delivers better consumer value. It also assumes that observed retention differences are caused by channel assignment, an assumption used only for illustrative arithmetic. Real cohorts require risk-adjusted inference and uncertainty ranges for losses, retention, expenses and discounting.

Average and marginal channel economics

For a channel with fixed avoidable infrastructure cost F and marginal cost v per effective placement in a range of capacity, average cost is F/n + v. Lower marginal costs can coexist with larger average costs at low scale. Consider a direct interface with F = 240,000 and v = 45 and a partner route with F = 40,000 and v = 95, for equal insurance output and a common cost boundary. Equating these expressions gives n = 4,000 placements. At 2,000, average costs are 165 and 115; at 8,000, they are 75 and 100. The scale crossover depends on the assumptions and is not a universal technological advantage. Capacity step-ups, quality-control requirements, minimum contractual payments, partner bargaining and demand uncertainty can change the crossover.

Avoidability matters when shifting an existing portfolio. Historical engineering expenditure may be sunk and therefore irrelevant to a narrow forward-looking marginal decision, while ongoing maintenance and closure obligations remain relevant. Full long-run profitability requires enough contribution to sustain appropriate resources. Comparing sunk-cost exclusion for one route with fully allocated capital costs for another creates an inconsistent decision basis. Both full-cost and prospective incremental views should be reconciled to their respective questions.

Risk-adjusted analysis and uncertainty

Expected claims must be evaluated according to the risk and coverage pertinent to the acquired cohort. For general insurance, loss maturity, reinsurance arrangements and claims inflation can materially change measured contribution. For long-duration life insurance, mortality, morbidity, surrender behavior, guarantees and capital requirements can dominate initial expense differences. A distribution score should not be inferred from an immature accident year or by substituting gross premium for earned economic revenue without adjustment.

Sensitivity analysis varies conversion, persistence, claim costs, cost-allocation weights and the cost of service. Correlations should be preserved where they matter: a change in underwriting appetite might simultaneously increase conversion and expected claims. A scenario that varies only acquisition cost can conceal the mechanisms by which an apparently efficient cohort becomes costly. For observed cohorts, statistical uncertainty in the inputs should be carried into estimated contribution, with appropriate clustering and model validation. Point estimates and ranges serve different purposes.

Causal Identification and Comparative Market Evidence

Causal estimands

A channel effect is a contrast between outcomes under alternative feasible distribution interventions for the same target population. It requires defining the intervention: assignment to direct contact, offer of advice, availability of a bank route, acquisition through a partner, or a rule determining how prospects are routed. Channel assignment can affect price, product choice, acceptance and service. The intended treatment must specify which of these are included in the effect.

Let Y(a) denote an eligible prospect's outcome under channel policy a and Y(b) under policy b. The population average treatment effect is E[Y(a) − Y(b)] for a population for whom the alternative interventions are meaningful. Outcomes can include acquisition spending per eligible prospect, probability of activation, net contribution, or a prespecified measure of suitable coverage. Cost per retained customer is a ratio of population totals, for example E[C(a)]/E[R(a)], with R(a) an indicator of retention. It differs from the expectation of a person-level ratio, often undefined because many prospects do not purchase. A policy-level estimand naturally includes spending on unsuccessful acquisition.

Hernán and Robins (2020), in Causal Inference: What If, present established methods for causal contrasts, confounding and longitudinal analysis. Applied to insurance distribution, the equivalent target experiment would specify eligibility, available channel policies, assignment, follow-up, outcomes, comparisons and analysis before observing the outcome. Their methodological tradition informs this application; no newly invented identification method is proposed.

Randomized and quasi-experimental comparisons

A randomized invitation or routing experiment can compare feasible service routes for eligible prospects. Random assignment supports an intention-to-treat estimate when followed and analyzed appropriately. Customers must remain able to decline or receive the assistance and protections required by the product and applicable law. A treatment can change actual channel use without perfectly determining it. Intention-to-treat estimates preserve the assignment comparison; per-protocol analysis requires additional assumptions.

Interference arises when demand routed to one adviser changes the waiting time of customers assigned elsewhere, or a campaign alters bids in a shared advertising auction. Cluster assignment or a market-level intervention may be more suitable than individual randomization. Capacity should be reported because a small experiment under spare capacity may underestimate the resources required for large-scale deployment. Ethical and regulatory constraints govern the eligible treatment set.

When random assignment is infeasible, a phased introduction of a platform or agency service can support a difference-in-differences analysis. Identification requires credible untreated counterfactual trends, absence of important anticipation or spillovers, and attention to changes in composition. With staggered implementation and heterogeneous effects, cohort-specific comparisons require care. A simple before-and-after decline in costs does not isolate the effect from changing demand, inflation or underwriting policy.

An instrumental variable may arise from genuinely exogenous routing variation. Relevance, independence, an exclusion restriction and, for a standard local average treatment interpretation, monotonicity require substantive justification. Distance to a bank branch, for example, might affect purchase and retention through income, local economic conditions or service access beyond the designated channel; treating it as a valid instrument solely because it predicts channel use is unjustified.

Regression discontinuity may exploit a routing rule with a threshold, provided other determinants vary continuously near the cutoff and manipulation is absent. Its estimate is local to the threshold. Matching, stratification, weighting or outcome regression can balance measured pre-treatment features where overlapping alternatives exist. They cannot automatically resolve unmeasured selection. Hierarchical models can separate firm, product, region and period variation and improve estimates for small cohorts, subject to assumptions about pooling and residual structure.

Conditioning and measurement error

Analyses limited to completed contracts condition on an outcome potentially affected by the channel and by customer risk, motivation and underwriting. Such conditioning can create selection bias even when initial routing is randomized. If the target outcome is the channel's acquisition strategy, start with eligible prospects and record zero or actual outcomes for those who do not activate. If the target is post-binding service among existing policyholders, define the separate population and the service intervention at binding.

Price and coverage can be confounders if chosen before channel assignment, or mediators if assignment influences them. Controlling for intermediates removes part of the total channel effect and can introduce bias when mediator-outcome confounders are affected by treatment. The causal question determines the model, not the fact that a variable happens to be available. A prespecified causal diagram and chronology of events help distinguish these roles.

Measurement error arises from fragmented identities, absent partner events, platform cookies, customer refusals to share data and incompatible timestamps. Last-touch attribution records a convention. It does not estimate whether a sale would occur without that touch. Analysts should keep source, assistance, binding, servicing, and renewal channel fields separately, preserving provenance and uncertainty. Data linkage requires a lawful basis and proportionate access controls.

Historical efficiency studies

Berger, Cummins, and Weiss (1997) evaluated cost and profit efficiency in US property-liability insurance. Their distribution-systems study reports support for an explanation in which independent-agency insurers produce higher-quality outputs and earn compensation for them. The finding is historically valuable because it rejects a simple inference from higher relative cost to economically inferior production within the study's interpretation. It concerns insurer-level evidence and measured outputs, not an experimental comparison of otherwise identical digital and human journeys.

Regan's transaction-cost account examined how distribution organization aligns relationship rights and activities. The conceptual mechanism complements the cost-versus-profit distinction: the same input can generate different output value when information, negotiation and service differ. These studies are from earlier market structures; modern acquisition auctions, real-time underwriting and platform data access create additional resource and selection mechanisms. Their conclusions should be used at the level of economic reasoning rather than transferred as fixed quantitative parameters.

Trigo-Gamarra and Growitsch (2010), in their German life insurer comparison, used data envelopment analysis for 1997–2005 and reported that specialization offered no clear performance advantage accounting for the coexistence of distribution systems. DEA compares units against an estimated frontier using selected inputs and outputs. A frontier ranking depends on the sample, the assumptions about scale and the dimensions measured. It is an efficiency assessment conditional on those choices, not a causal assignment contrast.

Chang, Peng, and Fan (2011) found in their Taiwanese channel-efficiency comparison that traditional sales achieved a higher relative efficiency score than a comparable bancassurance route using DEA. The result demonstrates why a presumed economy from shared bank access must be tested against observed inputs and outputs. It does not contradict the possibility of cost savings in specific banking arrangements. The German and Taiwanese studies concern different markets, units of analysis and benchmarks. Neither supplies a universal order of acquisition costs.

The studies cited here were examined at the abstract and bibliographic level, and their reported results are confined to that level. No numerical effect sizes, efficiency scores, statistical significance thresholds beyond the abstract's description, or unobserved sample details are inferred. In particular, comparisons of profit efficiency between insurer groups and DEA comparisons of sales units do not measure the same outcome.

Contemporary institutional evidence and transferability

EIOPA's (2026) distribution-market structure findings draw on a survey of national authorities and report that commission remuneration remained prevalent in 2024. Its discussion acknowledges differences in registration and collection practices across jurisdictions. The data characterize institutional arrangements; they do not identify distributor resource expenditure per policy. A commission share and a channel cost share cannot be substituted for each other.

The OECD's (2026) global market trend summary distinguishes life and non-life insurance penetration patterns across jurisdictions. Aggregated market conditions help formulate comparison strata, but a national premium-to-GDP measure cannot reveal the causal resource productivity of individual channel episodes. It is possible for a market with considerable bank distribution to have high insurance penetration because of product composition, savings institutions, incomes, tax policy and other structural factors.

A cross-market design should establish comparability of insurance product, customer population, distribution roles, policy duration, risk, currency, accounting perimeter, expenses borne outside insurer accounts, and legal context. Purchasing-power adjustment may illuminate labor resources; market exchange rates may be relevant to financial consolidation. Both need explicit dates and purposes. Quantitative pooling needs evidence that the estimands are aligned. The absence of harmonized matched transaction data is a central limit of this monograph; it constrains empirical ranking without weakening the logical distinction between cost, profit, and causal contribution.

Fair Value, Competition and Regulatory Implications

The consumer's full cost and service benefit

A consumer pays a contractual premium and may incur separately charged fees, financing costs and time costs. Some charges are already embedded in premium and must not be added twice. Search effort, repeated information entry, travel, inconvenient service and the burden of resolving claims can be important, especially for low-value or difficult-to-access coverage. A complete consumer perspective records monetary outlay and relevant nonmonetary burdens, together with benefit limits, exclusions, adequacy of coverage and service quality.

Expected claim payment is one component of the value of insurance. The benefit of transferring a financially disruptive risk can exceed the expected indemnity for a risk-averse buyer. A claims-to-premium ratio alone consequently cannot establish fair value. Contract usefulness, financial protection, appropriate exclusions, quality of information and practical access to the claim process matter. Measuring service quality may require multiple indicators: complaints by exposure and transaction type, correction rates, suitability review, service resolution times and claims outcomes. Different product types require different baselines.

The FCA's (2026c) general insurance value measures provide indicators covering claims experience and complaints, with explicit cautions about portfolio composition, product age, duration, premiums and timing. These data support assessment of outcome patterns, not direct identification of the cost or causal value of a particular intermediary. A low ratio can be a screening signal requiring investigation; a high ratio can also reflect changed pricing, transient sales volumes, or timing effects. A complete fair-value review examines the relevant product and distribution arrangement.

Supervision and incentives

The IAIS (2024) Insurance Core Principles provide a global supervisory reference. ICP 18 addresses professional and transparent intermediation, while the conduct discussion in ICP 19 recognizes differences by customer and product and the relevance of consumer outcomes. These principles guide supervisory frameworks; their legal implementation depends on the relevant jurisdiction. Cost studies should treat compliance resources as part of the process required to deliver the relevant legally permissible output.

The EU Insurance Distribution Directive (European Parliament and Council, 2016, consolidated edition of October 9, 2024) includes duties relating to customer interests, remuneration incentives, and demands and needs. Its distribution approach makes the relationship between sales incentives and appropriate products an economic question as well as a conduct issue. The cited consolidated text was examined through official indexed excerpts; direct page access was challenged. The analysis uses the identified framework rather than asserting a comprehensive current legal position in every Member State. National transposition and subsequent applicability require separate jurisdiction-specific assessment.

In the United Kingdom, the FCA's (2026a) manufacturer distribution-value rules address the impact of distribution arrangements on relevant non-investment insurance products. The examined provisions require consideration of remuneration across the chain and information about services performed. The corresponding FCA (2026b) distributor product-governance rules require relevant distributors to understand the product and target market, and include fair-value considerations for non-investment insurance. The scientific interpretation is that a distribution-cost review should trace services and payments through the chain rather than assume that external remuneration is inherently excessive or that internal distribution expense inherently delivers value.

The regulatory examples apply according to their scope, effective provisions and relevant law. They are not a universal commission cap or numerical acquisition-efficiency standard. A commercial policy, an insurance-based investment product and a retail non-investment product can be subject to different requirements. The institutional analysis concerns measuring and explaining efficiency, not advising an individual firm or consumer on legal compliance.

Competition and channel bargaining

Channels compete for customers, access to insurance capacity and distribution resources. A platform can reduce search frictions while concentrating access to prospective buyers. An exclusive agent can benefit from stable product-specific investment while limiting its comparison set. A bank may offer valuable contextual access but can have bargaining power over insurers seeking that access. A broker may improve the relevance of insurer quotations while encountering incentives linked to particular providers. These mechanisms are conditional on the arrangement.

The supplier price of distribution services can exceed its underlying resource cost because customer access and technological integration are scarce, because contract terms allocate risk, or because the channel has bargaining power. A firm's observed commission is a poor standalone indicator of whether competition is working. Studies of competitive effects should examine entry barriers, effective choice, contractual restrictions, switching costs, and the ability to access alternative insurance providers. A reduction in a distributor's compensation may benefit consumers through lower prices, benefit insurers through higher profit, reduce service investment, or change channel participation. Incidence is an empirical question.

In multichannel systems a single customer can receive advice in one route and conclude the contract in another. The channel benefiting from the final sale may not have financed the informative work. This can affect the viability of service provision and the incentives to share data. A resource ledger should recognize actual activity; commercial settlement can then be analyzed separately. A concentration of payments in one route does not demonstrate that the route created the entire economic output.

Distribution efficiency and inclusion

Insurance access is part of the output of distribution. A channel may be inexpensive for customers who can independently navigate a standardized service while failing to reach populations requiring language support, accessibility adaptations, cash-payment options or professional assistance. The cost of appropriately serving these populations can be higher even when the incremental social benefit of coverage is substantial. Segmentation is therefore needed to distinguish exclusion of expensive prospects from productivity improvement.

The appropriate comparison is a feasible delivery system capable of meeting the defined insurance and conduct requirements for the target population. A study that restricts attention to successfully acquired low-cost customers implicitly excludes unmet demand. Where protection remains inadequate, relevant outcomes include offer reach, eligible population coverage, appropriate conversion and persistence. Distribution cost is one constraint on increasing access alongside price, risk, product design, household liquidity and trust.

Integrated Research Synthesis

Reproducible comparative workflow

The substantive relationship emerging from the literature and the analytical examples is that observed acquisition cost combines input prices, quantities of resources, distribution scope, selection, product complexity, scale, and timing. A useful comparison examines these elements in sequence.

The investigator first specifies the decision and population. A cost-reduction project for an existing insurer asks about avoidable resources and attainable service. A market-structure study may ask about coexistence of institutional forms and profit efficiency. A consumer-protection study asks how payments relate to contractual benefits and services. An allocation of marketing budget asks what incremental output an extra unit of expenditure causes. Each question has its own boundary and counterfactual.

The second stage constructs the transaction and resource ledger. Each record identifies the relevant product version, customer or risk unit, eligibility event, channel touchpoints, quote and bind dates, activation, payment, underwriting status, contractual compensation, and resource driver. Shared resources and idle capacity are reconciled to the cost base. A documented allocation procedure eliminates duplicate transfers for consolidated analysis. The ledger preserves all eligible prospects, including unsuccessful placements.

The third stage describes outputs and uncertainty. The report presents counts, conversion, expenditure per policy and exposure, composition, and supported service outcomes. Where the horizon permits, it follows retention and cash-flow contribution. The results report amounts both per cohort and per relevant denominator. Confidence intervals can be constructed using methods appropriate to the ratio, clustering and sampling design. Aggregate reports describe the uncertainty of the estimated inputs and the population to which inference applies.

The fourth stage chooses an identification strategy. For channel offers or routing, an experimental or credible quasi-experimental design may estimate incremental impact. For retrospective benchmarking, standardization enables limited like-for-like comparisons, conditional on measured characteristics and overlap. A frontier method describes relative production efficiency conditional on its definitions. Its answer should not be relabeled as a causal effect. Where the necessary comparison is absent, the report presents descriptive results and states the causal limitation.

The fifth stage interprets the results through three accounting perspectives: the insurer's outlay and risk-bearing economics, the distributor's costs and revenue, and the customer's expenditure, protection and service. These perspectives are connected by transfers and obligations. Reconciliation establishes where resources are consumed and where money flows. Conflicts among the perspectives may reveal the incidence of costs and benefits, rather than measurement error.

Minimum information and falsification tests

A channel study should be reproducible from its cohort rules, resource ledger, activity drivers, exclusions, outcome window, attribution convention, and assumptions. A data dictionary differentiates leads, unique eligible prospects, complete quotations, bound contracts and activated policies. It includes whether a cancellation is counted as a reversed bind, a separate lapse, or a continuation under another policy. This choice is recorded before calculating cost.

Checks begin with the money and event reconciliation: the sum of direct costs, allocated common costs and unallocated capacity reconciles to the included expense total; partner transfers are not duplicated; an activated policy has a valid bind event; dates follow the actual process; and each eligible unit has an auditable outcome or censoring marker. Standardization checks assess common support, weights and sensitivity to cohort definitions. Model diagnostics address calibration, influential observations, residual structure, treatment assignment and missingness. A causal design should examine pre-intervention patterns and plausible falsification outcomes when the method permits.

A result is sensitive when reasonable alternative allocation assumptions reverse it. A useful sensitivity result is itself a finding: it indicates that the apparent ranking depends on how joint resources or long-run benefits are valued. Estimates at commercially relevant scales should be distinguished from small-pilot estimates. An analyst should disclose when product equivalence, risk adjustment or service measures are too weak to support a particular ranking.

Analytical contribution and boundaries

Three conclusions are supported by the integrated reasoning. First, cost comparisons are meaningful only with congruent resource and output boundaries. The same commission can be an insurer cost, an intermediary revenue and an eliminated transfer in a consolidated resource ledger. This accounts for conflicting reports that use different perimeters.

Second, acquisition efficiency at inception is related to but distinct from the efficiency of durable coverage. Conversion attrition and retention can substantially change the economic relationship between initial expenditure and future contribution. The numerical examples establish the arithmetic conditions under which rankings reverse or remain stable. Their contribution is methodological illustration and synthesis rather than an empirical discovery.

Third, cost, profit, and customer value measure different outcomes produced by the same distribution system. Historical research shows the importance of considering output quality and market fit, and supervisory frameworks make service and customer outcome evidence relevant. A channel intervention can improve one outcome while worsening another, which necessitates explicit specification of the evaluation criterion and the feasible alternative.

The synthesis has defined limits. Access to historical economics papers was frequently restricted to abstracts, and some official legal text was available through search indexing rather than a fully accessible consolidated page. Institutional sources describe selected jurisdictions and do not establish all laws applicable in November 2026. There is no proprietary firm-level cost ledger or internationally harmonized matched prospect dataset in the study. The examples are invented and demonstrate calculations under stated assumptions. Claims require further empirical testing using the same defined estimands before being applied as quantitative performance expectations. No new named theory, doctrine, empirical estimate, or priority claim is established.

Future empirical investigation

The most useful empirical extension would collect prospect-level information across genuinely comparable insurance offers under documented channel exposure, legally permissible assignment variation, and longitudinal service observations. Researchers could join time-driven operational costs with verified partner payments and policy exposure, while maintaining appropriate privacy and governance. A multisite study could estimate the heterogeneity of incremental effects by product complexity, consumer needs, risk, institution and fulfillment method.

Formal comparisons should distinguish the effect of making a channel available from the effect of actually using it, and the effect of a bundled offer from the effect of the acquisition process with product terms held constant. A long follow-up period permits assessment of persistency, contribution and relevant service indicators. Predefined outcome families and transparent handling of missing records would reduce opportunistic selection of favorable channel metrics. The scientific objective is a comparison with interpretable uncertainty and well-defined scope, not a predetermined victory of any institutional form.

Conclusion

Insurance acquisition efficiency is an economic relationship among resource use, effective placement, contractual protection, the composition of the acquired portfolio, and subsequent outcomes. Agency, brokerage, bancassurance, direct, platform and embedded channels combine different functions, costs, relationships and access conditions. Their measured performance depends on the boundary of the investigation.

The monograph establishes the importance of reconciling insurer expenditure, intermediary revenue and underlying resource use; maintaining denominators that represent the actual acquisition process; allocating shared resources and unused capacity transparently; distinguishing placements from activated and persistent coverage; and evaluating expected net contribution with coherent risk and time assumptions. Synthetic examples show how composition can reverse aggregate rankings, how higher retention may fail to compensate for greater acquisition expenditure, and how scale changes average costs.

Historical comparative research provides evidence that channel organization and efficiency coexist in different configurations across insurance markets. Regulatory and supervisory materials establish the scientific relevance of customer services, incentive structures and outcome evidence. A defensible causal assessment specifies the intervention and its alternative, avoids conditioning on downstream selection, and investigates the assumptions supporting experimental or observational comparisons.

A distribution arrangement is economically efficient relative to feasible alternatives when it uses the relevant resources effectively to produce the specified insurance outputs, including adequate service where that is part of the obligation. Acquisition cost is indispensable to that assessment. Its scientific meaning derives from the comparative accounting, risk, service and causal structure in which the cost is measured.

References

  • Berger, Allen N., J. David Cummins, and Mary A. Weiss. 1997. “The Coexistence of Multiple Distribution Systems for Financial Services: The Case of Property-Liability Insurance.” Journal of Business 70 (4): 515–546. Institutional article record.

  • Regan, Laureen. 1997. “Vertical Integration in the Property-Liability Insurance Industry: A Transaction Cost Approach.” Journal of Risk and Insurance 64 (1): 41–62. Author abstract and citation.

  • Trigo-Gamarra, Lucinda, and Christian Growitsch. 2010. “Comparing Single- and Multichannel Distribution Strategies in the German Life Insurance Market: An Analysis of Cost and Profit Efficiency.” Schmalenbach Business Review 62 (4): 401–417. Indexed journal record.

  • Chang, Pang-Ru, Jin-Lung Peng, and Chiang Ku Fan. 2011. “A Comparison of Bancassurance and Traditional Insurer Sales Channels.” Geneva Papers on Risk and Insurance—Issues and Practice 36 (1): 76–93. Indexed journal record.

  • Kaplan, Robert S., and Steven R. Anderson. 2004. “Time-Driven Activity-Based Costing.” Harvard Business Review 82 (11): 131–138, 150. Publisher article summary.

  • Hernán, Miguel A., and James M. Robins. 2020. Causal Inference: What If. Boca Raton, FL: Chapman & Hall/CRC. Author book site.

  • Cole, Shawn, Xavier Giné, Jeremy Tobacman, Petia Topalova, Robert Townsend, and James Vickery. 2013. “Barriers to Household Risk Management: Evidence from India.” American Economic Journal: Applied Economics 5 (1): 104–135. DOI: 10.1257/app.5.1.104. Federal Reserve research record.

  • European Insurance and Occupational Pensions Authority (EIOPA). 2026. Structure of the EU Insurance Distribution Market. Findings accompanying the third report on application of the Insurance Distribution Directive; 2025 NCA survey, remuneration data for 2024. Official market structure findings.

  • European Parliament and Council of the European Union. 2016. Directive (EU) 2016/97 of 20 January 2016 on Insurance Distribution (Recast). Official Journal L 26, 2 February 2016, 19–59; consolidated text 9 October 2024, arts. 17, 19–20. Consolidated directive text.

  • Financial Conduct Authority (FCA). 2026a. FCA Handbook, Product Intervention and Product Governance Sourcebook, PROD 4.2, Manufacture of Insurance Products. Online rules, particularly 4.2.14N and related remuneration-information provisions. Official PROD 4.2.

  • Financial Conduct Authority (FCA). 2026b. FCA Handbook, Product Intervention and Product Governance Sourcebook, PROD 4.3, Distribution of Insurance Products. Online rules, particularly 4.3.2 and 4.3.2A and distribution value provisions. Official PROD 4.3.

  • Financial Conduct Authority (FCA). 2026c. General Insurance Value Measures Data 2025. Reporting year January–December 2025. Official value measures report.

  • International Association of Insurance Supervisors (IAIS). 2024. Insurance Core Principles and Common Framework for the Supervision of Internationally Active Insurance Groups, revisions adopted 5 December 2024. ICP 18, Intermediaries; ICP 19, Conduct of Business. Official ICP online text.

  • OECD. 2026. Global Insurance Market Trends 2026. Paris: OECD Publishing. Executive summary, 2 November 2026. OECD executive summary.