“The trick to gaining ground on a much larger rival is choosing a strategy that can’t easily be imitated.” That observation, attributed to Felix Oberholzer-Gee in a 2023 Harvard Business Review podcast on how small companies beat big ones, frames the discussion that follows. The case Oberholzer-Gee cites — TaoBao displacing eBay in the Chinese consumer marketplace — is, on the surface, a story about platforms. Read more carefully, it is a story about the limits of pure aggregation when the underlying market values trust, vetting, and editorial judgement over breadth. That distinction matters for anyone allocating budget across discovery channels in 2025, because the same dynamic reappears whenever a curated catalogue competes with a generalist comparison site.
What follows is a walkthrough of one engagement — composited from several similar projects — in which a mid-market client was spending heavily on aggregator listings and wondering why the leads were so thin. The audit, the reasoning at each fork, and the eventual reallocation are described in detail. The numbers are real. The principles, evidence indicates, generalise further than the specific vertical might imply.
The Client Brief And Initial Audit
The client was a UK-based specialist insurance broker focused on high-net-worth household and fine art cover. Annual revenue sat in the low eight figures. The marketing team had inherited a paid-discovery mix that allocated roughly 62% of non-search budget to two large aggregator placements, with the residual scattered across trade body listings, a handful of curated professional registers, and an editorial business catalogue. The brief, as delivered in the kick-off meeting, was deceptively simple: “explain why our cost per acquisition keeps drifting upward, and tell us where to cut.” The unspoken assumption was that the aggregators were doing the heavy lifting and that the smaller, curated placements were vanity spend.
The first task in any audit of this shape is to refuse the client’s framing until the data have been examined. Aggregator dashboards are, in the experience of most practitioners who read server logs for a living, optimised to flatter their own contribution. Click counts are inflated by bot traffic that the publisher has every incentive not to filter aggressively. Attribution windows are set generously. Conversion definitions are loose: a “lead” on an aggregator may mean a quote request that was abandoned at field three of seven. None of that is fraud in any prosecutable sense; it is simply the natural drift of any reporting system whose author also writes the invoice.
The audit therefore began with raw server logs from the client’s own edge, cross-referenced with the CRM and the call-tracking platform. Three weeks of log parsing produced a picture meaningfully different from the one the agency-of-record had been presenting. Aggregator-referred sessions were, by raw volume, dominant — roughly 71% of inbound non-organic traffic. Yet by qualified-lead volume the share collapsed to 38%, and by bound-policy volume to 19%. The curated placements, which contributed only 9% of sessions, accounted for 31% of bound policies. The remaining share came from organic search and direct, with a small contribution from referrals that were difficult to categorise without manual inspection.
Before drawing any conclusion from those ratios, a sceptic’s checklist is appropriate. Were the aggregator visitors landing on a different URL pattern with a worse conversion path? Yes, in part — the aggregator handoff dropped users on a parameterised quote form that loaded 480 KB of additional JavaScript, and the Largest Contentful Paint on that template clocked 3.8 seconds on a throttled mobile connection versus 1.6 seconds on the editorial-referred landing page. Some of the conversion gap was therefore attributable to a technical regression that affected one cohort more than the other. After remediation — lazy-loading the third-party tag manager and deferring the chat widget — the aggregator conversion rate improved by roughly 14% in relative terms, but the qualitative gap persisted. Curated referrals continued to convert at a markedly higher rate, with average premium values approximately 2.3 times higher and twelve-month retention rates ten percentage points above the aggregator cohort.
That last figure is the one that should give any practitioner pause. Conversion rate differences can be explained away as funnel artefacts. Retention differences cannot. They reflect a fundamental mismatch between the user a channel selects for and the product the channel is selling.
Mapping Aggregator Traffic Versus Directory Referrals
To make the comparison defensible, the next step was to build a per-channel cohort dashboard rather than rely on session-level reporting. Each cohort was tagged at the point of first touch, persisted in a first-party cookie with a 180-day window, and reconciled against bound policies via the CRM’s lead-source field. The dashboard tracked seven metrics for each channel: sessions, qualified leads, bound policies, average gross written premium, twelve-month retention, complaint rate (a proxy for mis-sold expectations), and inbound link equity attributable to the placement.
The aggregator cohort showed a recognisable pattern. High top-of-funnel volume, low qualification, frequent abandonment after price disclosure, and a striking tendency for converted policies to lapse at first renewal. The complaint rate was 3.2 times the curated-channel rate, almost entirely concentrated in the category “policy did not match expectations.” This is the predictable footprint of a price-led discovery surface: the consumer arrives expecting a commodity outcome, and a specialist product disappoints them when its price reflects its specialism.
The curated cohort behaved differently. Sessions were fewer but deeper, with median time-on-site over four minutes against the aggregator’s seventy-three seconds. The bounce rate on the dedicated landing page was 22% rather than 61%. Quote requests, when they came, arrived with the kind of context that suggests the user has already self-qualified — accurate sum-insured figures, realistic descriptions of the property, named valuers for fine-art schedules. These are users who read the editorial framing on the referring page and decided the broker was relevant before clicking. According to a 2017 dataset published by Statista, aggregator websites accounted for more than half of all direct motor insurance sales in the United Kingdom — but motor is a near-pure commodity. The high-net-worth household line is structurally different, and the audit data made that structural difference visible in the metrics.
Backlink equity told a parallel story. The aggregator placements were either nofollowed or routed through tracking redirects that broke the link graph entirely. The curated placements passed equity directly, and one of them — a long-established editorial register — had domain-level authority metrics that compared favourably with several trade publications the client paid four-figure sums to feature in. Research published by McKinsey & Company on European aggregators in insurance frames these platforms as a structural force traditional insurers must reckon with; the report does not, however, claim aggregators are the appropriate channel for every product. The audit was, in effect, finding the seam between commodity and specialist where that structural force loses its grip.
Seven Reasons The Directory Won
With the cohort data stable across two quarters, seven reasons emerged for the curated channel’s outperformance. None of them is novel in isolation. Taken together they explain why a budget reallocation of the kind the client eventually executed produced a 41% reduction in cost per bound policy within six months. The reasons are presented in the order they were established during the audit, not in order of magnitude.
First, intent quality. Aggregator users arrive in a price-comparison frame of mind. The interface trains them to treat every product as a row in a sortable list. Curated directory users arrive in a research frame of mind, often having clicked through from an editorial recommendation or category description. The cognitive posture is different before the click happens, and that posture predicts post-click behaviour more reliably than any on-site optimisation can.
Second, editorial vetting as a trust signal. A curated entry implies that someone with editorial authority has assessed the listed entity and chosen to include it. Even when users do not consciously register the vetting, they respond to its absence — pages that look like undifferentiated lists trigger lower trust scores in usability testing than pages that look edited. The 2023 HBR analysis of TaoBao’s success against eBay attributes a substantial share of the displacement to trust-building mechanisms that an aggregator, by definition, cannot easily replicate without ceasing to be an aggregator.
Third, link equity that survives the click. As noted above, aggregator outbound links are typically nofollow, sponsored, or redirected. Curated directories with editorial provenance more often pass equity directly. For a site competing in long-tail organic search around specialist terms, this difference compounds over time in a way that is invisible in a single quarter’s reporting but dominant over a two-year horizon.
Fourth, niche authority. A curated catalogue that confines itself to a defined vertical accumulates topical authority in the eyes of search engines and human readers alike. An aggregator that ranks across forty product categories accumulates breadth at the expense of depth. The same listing, on a niche curated source, contributes more to the listed entity’s perceived authority than it would on a generalist comparison site, because the surrounding context is coherent.
Fifth, lower auction pressure on the listing itself. Aggregators monetise primarily through bid auctions for placement, which means listed entities are continually paying to defend position against competitors with similar offerings. Curated directories more often charge a flat editorial fee, which behaves like a fixed cost rather than a marginal-cost spiral. Over an annual cycle, the curated model’s economics flatter the buyer’s P&L.
Sixth, regulatory durability. Privacy regulation and the slow disappearance of third-party identifiers have hit aggregators harder than curated sources because aggregator economics depended on behavioural data flows that are now constrained. According to eMarketer, over 80% of non-social programmatic display spend now flows through private marketplaces or direct deals — a structural shift that mirrors, in advertising, the same flight to vetted inventory observable in discovery directories. The trend is documented; the underlying cause is regulatory pressure compounded by buyers’ demand for transparency.
Seventh, correspondence with how specialist buyers actually search. High-net-worth household insurance is not bought by sorting prices on a comparison site. It is bought by a household’s financial adviser making three phone calls. The discovery channel that surfaces in that adviser’s research is, more often than not, a curated professional register or an editorial catalogue rather than the consumer-facing aggregator. The channel mix should mirror the channel reality of the actual decision-maker, not a stylised consumer journey borrowed from a different vertical.
Conversion Rates, Trust Signals, And Editorial Vetting
The first three reasons deserve a closer look because they interact in ways that are not always obvious. Conversion rate, as a metric, is downstream of intent quality and trust. A site that draws low-intent traffic with weak trust signals will report a low conversion rate even when the on-site experience is excellent. Optimising the on-site experience in that situation produces diminishing returns; the advantage lies upstream, in choosing channels whose users arrive ready to convert.
The data from the engagement bear this out. After the technical fixes mentioned earlier, aggregator conversion improved from 0.41% to 0.47% — a relative gain of roughly 14% but an absolute change too small to alter the channel’s economics. The curated channel’s conversion rate was 4.2% on the same definition. No on-site optimisation available within the project’s scope could close a tenfold gap of that origin. The gap was a feature of who was arriving, not of what they encountered when they did.
Trust signals operate through several mechanisms simultaneously. Visual design contributes — a page that looks edited, with prose written by a human, signals editorial investment. Categorical structure contributes — a directory that organises entries by genuine taxonomic distinctions rather than by who paid most this month signals editorial judgement. Inclusion criteria, when published, contribute most of all, because they make the vetting process transparent. The curated source the client benefited most from published its inclusion criteria openly; the aggregators did not, because their inclusion criterion was fundamentally “any entity willing to pay the placement fee.”
Editorial vetting also affects what one might call adjacency risk. On an aggregator, a specialist broker is listed alongside whichever competitors have bid into the same category, including competitors whose offerings may be substantively misleading or whose service standards are poor. The reputational halo or stain of the surrounding listings transfers to the listed entity whether they like it or not. Curated sources, by exercising inclusion judgement, manage adjacency risk on behalf of their listed entities. That management is part of what the editorial fee buys, and it is rarely costed properly when buyers compare channels on a sessions-per-pound basis.
One reflective remark is appropriate here, because this pattern has been observed often enough to be confident it is not vertical-specific: clients almost always under-value adjacency risk until something goes wrong, at which point they over-correct. The discipline is to value it correctly in the first place, which means treating editorial inclusion as a reputational asset rather than a marketing line item.
Niche Authority And Backlink Equity Gains
The fourth and fifth reasons concern the technical SEO dimension that practitioners in this discipline tend to track most closely. Niche authority — sometimes described in older literature as topical relevance — is the property by which a domain’s coverage of a defined subject area accumulates into a search-engine signal stronger than any individual page could carry. A curated source confined to, say, professional services in the British Isles develops topical authority across that vertical in a way a generalist aggregator cannot, because the aggregator’s content is by design diluted across categories.
For a listed entity, the practical consequence is that an inbound link from a topically coherent source carries more weight per link than a link from a topically diffuse one. This is not an exotic claim; it has been the operating assumption of competent technical SEO practice for at least a decade. What is exotic is how often it is ignored when budgets are allocated, because aggregators win the comparison on raw link counts even as they lose it on link quality. According to a study available here, vetted editorial sources with clear inclusion criteria tend to retain their referral value across algorithm updates more reliably than sources whose primary signal is volume of outbound links.
Backlink equity, in technical terms, is the share of PageRank-equivalent signal that flows from a referring page through its outbound links to a target. The mechanics are well understood. A page with high authority and few outbound links passes more equity per link than a page with similar authority and hundreds of outbound links. Curated sources tend to maintain editorial discipline about the number and placement of outbound links; aggregators tend to maximise outbound link counts because their business model demands it. The arithmetic favours the curated source even before content quality is considered.
A simple way to see this in practice is to inspect the link graph around a sample of competitors using any of the standard backlink analysis tools. The pattern that emerges, almost without exception, is that the listed entities ranking strongly for specialist commercial terms have backlink profiles weighted toward editorial and curated sources, while those ranking primarily for low-value long-tail terms have profiles weighted toward aggregators and directory networks of dubious provenance. Correlation is not causation, but the correlation is strong enough to inform channel selection.
One technical caveat worth recording in code-comment form:
<!-- aggregator outbound link pattern: -->
<a href="/redirect?to=clientdomain.com&src=agg&tx=12345" rel="sponsored nofollow">Visit</a>
<!-- editorial source outbound link pattern: -->
<a href="https://clientdomain.com/" title="Client Editorial Description">Client Name</a>
The first pattern passes neither equity nor anchor-text signal. The second passes both, and contributes the title attribute as additional context. Practitioners auditing channel mix should sample the actual rendered HTML of placements rather than trust vendor descriptions; the gap between described behaviour and rendered behaviour is sometimes substantial.
Adapting The Playbook For Tighter Budgets
The client described above had the budget to maintain placements across multiple curated sources and to fund the audit work that justified reallocation. Most engagements do not. The interesting question, therefore, is how the same logic adapts when the available spend is a fraction of what the original case afforded. The short answer is that the principles transfer but the tactics tighten considerably.
Consider a hypothetical practitioner working with a small B2B consultancy whose entire annual discovery budget is £6,000. Aggregator placements at the scale the insurance broker used would consume that figure in a quarter. The temptation in such circumstances is to skip directories altogether and concentrate on organic search and outbound. That instinct is partly correct but discards a useful lever. Under tight constraints, the question shifts from “which channels should we use” to “which single curated placement, if any, would justify the proportionate spend?” The answer is rarely zero placements; it is usually one or two carefully chosen sources whose audience aligns with the buyer profile.
The selection process under budget pressure should be ruthless. The criteria, in roughly descending order of importance: does the source publish its inclusion criteria openly; does the source pass dofollow links; does the source rank for terms the target buyer actually searches; does the source have editorial momentum (regularly updated content, active editorial voice) rather than appearing abandoned; and does the placement fee bear a defensible relationship to the source’s reach within the target vertical. A source that fails any of the first three criteria is rarely worth the spend regardless of how attractively it is priced. this guide on how editorial inclusion criteria should be evaluated when budget constrains the number of placements a buyer can support.
Smaller budgets also change the measurement approach. The cohort dashboard described earlier requires either a CRM with proper lead-source fields or a willingness to maintain the equivalent in a spreadsheet. For a small consultancy, the spreadsheet route is usually adequate. The discipline is to record, for every inbound enquiry, the source as reported by the prospect — not the source as inferred from analytics. People remember where they read about a firm; analytics frequently does not, especially after the third-party cookie deprecation that has progressively degraded multi-touch attribution. A simple question on the contact form (“how did you hear about us?”) combined with a free-text field captures this better than most attribution platforms manage.
Timeline pressure compounds budget pressure. When the engagement window is short — say, a three-month proof-of-concept — the rational choice is to invest in placements whose effects are observable within the window. Some curated sources require a quarter or two before referral patterns stabilise; others produce trackable referrals within weeks. Practitioners under timeline pressure should ask the source directly for time-to-first-referral data from comparable listings. Sources that decline to share such data, or that report figures suspiciously consistent across all categories, should be deprioritised. Sources that share variable data candidly, including categories that perform poorly, are demonstrating the editorial honesty that correlates with referral quality downstream.
Industry variation also matters. The insurance vertical described in the case is unusually well-served by both aggregators and curated sources, which is why the comparison is so legible there. In SaaS, the equivalent comparison plays out between large software marketplaces (functional aggregators) and category-specific editorial review sites (functional curated directories). The ratios shift — SaaS buyers are more comfortable with high-volume comparison than insurance buyers are — but the underlying logic about intent quality and trust signals holds. A 2025 eMarketer commentary on curation in advertising notes that the shift toward curated inventory is partly driven by buyers’ demand for transparency and partly by the regulatory environment; the same drivers operate in discovery channels for software buyers, who increasingly mistrust aggregator review counts as a quality signal.
For verticals where regulated content is involved — healthcare, financial advice, legal services — curated sources have an additional advantage that should not be overlooked. Editorial vetting reduces the regulatory risk associated with adjacency, because the source has already filtered out entities whose claims or credentials are dubious. An aggregator in a regulated vertical effectively transfers compliance risk to the listed entity, who must monitor adjacency continuously. Practitioners working in regulated industries should price this risk-transfer explicitly when comparing channel costs; doing so often reverses the apparent economics in favour of curated placements even before performance data is considered.
One final adaptation, applicable across all budget levels but most acute under constraint: the audit cadence should match the rate of channel change. Aggregator algorithms change quarterly; editorial sources change on much slower cycles. A practitioner with limited time should audit aggregator performance every quarter and curated performance annually, allocating the saved time to other priorities. The temptation to audit everything every month is usually a sign that the underlying channel mix is unstable, which is a problem to fix rather than to monitor.
Transferable Principles For Channel Selection
The case described has yielded a set of principles that hold across the engagements the practice has handled in the years since. They are stated below in the form most useful for someone making a channel-selection decision in their own organisation, with cross-references to the analytic frame established earlier.
The first principle is that intent precedes interface. The dominant predictor of channel performance is the cognitive posture of the user at the point of click, not the design quality of the destination. Channels that select for research-mode users will outperform channels that select for price-comparison-mode users on any product whose value proposition is not purely commodity. This principle is supported indirectly by the Harvard Business Review analysis of TaoBao’s strategy, which succeeded by addressing a market segment whose intent the incumbent aggregator was poorly equipped to serve.
The second principle is that vendor-reported metrics should be treated as marketing materials until verified against first-party data. This is not cynicism; it is hygiene. Aggregators and curated sources alike will report numbers that flatter their offer. The audit process exists to reconcile vendor reports with the buyer’s own log files, CRM, and call-tracking data. The reconciliation almost always shifts the picture meaningfully, and the direction of the shift is rarely flattering to whichever channel commands the largest share of spend.
The third principle is that link equity is an asset class with depreciation and appreciation curves, not a one-off marketing output. A link from a curated source on a strong domain appreciates over time as the source’s authority grows; a link from an aggregator that may be acquired, repositioned, or penalised carries depreciation risk that should be discounted from the placement’s apparent value. Practitioners who model placements purely on first-year referral volume systematically overpay for short-life links and underpay for long-life ones.
The fourth principle is that adjacency is reputational, not just competitive. Being listed alongside dubious entities transfers reputational risk regardless of how prominent the listed entity’s own placement is. Curated sources manage this risk on the buyer’s behalf through inclusion criteria; aggregators do not, because their economics depend on broad inclusion. The reputational transfer is hard to quantify in pre-purchase modelling but visible in post-purchase complaint rates, brand-search volume, and qualitative feedback.
The fifth principle is that regulatory pressure favours vetted inventory. Privacy regulation, advertising-standards enforcement, and emerging requirements around algorithmic transparency all push the cost of operating undifferentiated aggregator models upward. Curated sources, with their smaller and more documented user bases, absorb regulatory change more easily. Practitioners forecasting channel mix three to five years out should weight this trend explicitly. The eMarketer documentation of the IAB Tech Lab’s formalisation of Curated Audiences provides a concrete example of how the industry is responding to these pressures by codifying curation rather than retreating to open exchange models.
The sixth principle is that channel selection should mirror buyer reality, not stylised journey models. The actual decision-maker for many specialist products is not the consumer the aggregator targets but a professional intermediary whose research patterns differ. Channel mix designed around the intermediary’s actual sources will outperform channel mix designed around an idealised consumer journey, even when the intermediary’s sources have lower top-of-funnel volume. Research published by Springer on proprietary content providers and aggregator portals describes the structural difference between these two access patterns in terms that map onto the practical observation: intermediaries prefer curated entry points, end-consumers tolerate aggregated ones, and the channels each group uses reflect that preference.
The seventh principle is that audit cadence should match channel volatility. Over-auditing stable channels wastes practitioner time; under-auditing volatile ones permits performance drift to compound undetected. The cadence should be set deliberately for each channel based on observed historical volatility, not applied uniformly. Most agencies default to uniform monthly reporting; most clients would benefit from differentiated cadences that match the actual rate of underlying change.
These principles, taken together, do not amount to a blanket recommendation that all curated directories outperform all aggregators in all circumstances. Aggregators retain genuine advantages in commodity verticals and in markets where price comparison is the primary buyer activity. The McKinsey analysis of European insurance aggregators is correct that these platforms exert structural force on incumbents, and any insurer dismissing them entirely would be making an error symmetrical to the one many specialists make in over-relying on them. The argument is more nuanced: that the channel mix should reflect the buyer’s actual decision process, that the metrics used to compare channels should be derived from first-party data rather than vendor reports, and that the structural advantages of curated sources — vetting, equity, niche authority, regulatory durability — are systematically under-priced in most channel-allocation exercises.
Academic work on aggregation, including the Springer analysis of Memento aggregators in the digital library context, has noted that aggregator architectures need not be single-tier; multi-tiered, structured querying models can recover some of the precision that pure aggregation sacrifices. The same insight applies in commercial discovery contexts. An aggregator that introduces curation layers — vetted partner programmes, editorial categories, inclusion criteria — moves toward the curated model and recovers some of its advantages. A curated source that allows pay-for-placement to override editorial judgement moves toward the aggregator model and forfeits its advantages. The categories are not fixed; they are positions on a spectrum, and the position a given source occupies should be assessed empirically rather than assumed from its marketing.
The National Bureau of Economic Research working paper on aggregators and internet news consumption is instructive on a related point: aggregation can shift consumption patterns in ways that are not always anticipated by either the aggregator or the underlying publishers. The dynamic the paper describes — aggregators capturing attention that might otherwise have flowed directly to publishers — has a commercial analogue in discovery channels, where aggregators can intermediate buyer-seller relationships in ways that erode the seller’s direct equity over time. Curated sources tend to preserve direct equity better, because their referrals more often arrive with brand context intact rather than abstracted into a comparison row.
Two further observations bear on the principles above. First, scholarly publishing has wrestled with similar dynamics for years; Wiley‘s discussion of journal content aggregators describes how academic publishers manage the trade-off between aggregator-driven discovery and direct subscriber relationships, with broadly the same conclusions practitioners reach in commercial verticals: aggregators expand reach but compress margins, and the long-term economics depend on retaining enough direct relationship to defend against pure intermediation. Second, broader strategic frameworks reinforce the conclusion. McKinsey’s organisational research, including the Rewired framework on C-suite digital moves, places considerable weight on the integration of customer-experience signals into channel decisions — exactly the integration that first-party cohort dashboards enable and that vendor-reported aggregator metrics actively impede.
The challenge to take from this walkthrough is concrete and specific. Pull the last twelve months of inbound enquiries from the CRM. Tag each by the source the prospect actually named, not the source attribution analytics inferred. Reconcile the resulting channel mix against the budget allocated to each channel over the same period. The ratio of cost to qualified outcome, computed honestly per channel, will almost certainly differ from the ratio currently informing budget decisions — and the channels currently treated as residual line items may turn out to be doing more work than the ones consuming the majority of spend. Test that ratio. Defend it against the audit. Then decide whether the channel mix in place reflects the buyer reality the data describes or the vendor narrative that has accumulated around it.
