Which Channels Do Leading D2C Media Agencies Optimize? Paid Social, Search, TV, Affiliates & More

Key Takeaways

• Leading D2C agencies run channels as a portfolio with distinct roles — demand creation, demand capture, marketplace, proof, and retention — not as a checklist to tick.
• Paid social (Meta, TikTok) is the primary creative-led demand-creation and volume lever; Google Search and Shopping captures the demand it creates and is capped by how much demand exists.
• Retail media (Amazon) is increasingly core because so much product search now starts there; CTV scales brand demand once the efficient channels saturate.
• Affiliates and creators supply proof and CPA-based performance; email and SMS lifecycle is where D2C profit is actually made, on the repeat purchase.
• The mix is decided by incrementality and contribution margin and read with media-mix modelling and holdouts — not by last-click attribution, which over-credits capture and under-credits creation.
• The right mix evolves with scale: early brands lean on paid social and search; as they grow and efficient channels saturate, they add retail media, CTV and deeper retention.
• The mark of a leading agency is not the length of its channel list but the rigour of how it allocates across the portfolio.

Channels Are a Portfolio With Roles, Not a Checklist

The difference between an average D2C media agency and a leading one is not which channels they can run — everyone can run Meta and Google, and 'we also do TikTok and CTV now' is table stakes, not a differentiator. The difference is how they think about the mix. An average agency treats channels as a checklist to accumulate: it adds channels because they are fashionable or because a competitor is on them, spreads budget across them roughly evenly, and reports each one on its own last-click ROAS as if the channels were independent businesses competing for a prize. A leading agency treats channels as a portfolio in which each has a specific job in the funnel, allocates budget by the incremental contribution each produces rather than the credit each claims, and reads the whole mix together — because the channels are not independent, they interact, and measuring them as if they were separate systematically mis-allocates budget.

That portfolio thinking is the real answer to 'which channels do leading D2C agencies optimise?' They optimise every channel that earns its place — but they optimise them as an integrated system with defined roles, not as a pile of independent line items each chasing its own ROAS. The distinction matters enormously in practice, because the channels that create demand (paid social, CTV) and the channels that capture it (search) do fundamentally different jobs, and a last-click metric will always flatter the capture channels — which sit at the bottom of the funnel and get the click credit — while starving the creation channels that actually drove the demand the capture channels harvested. An agency that does not understand this will confidently move budget from the channels growing your business to the channels merely closing what those channels started, and call it optimisation. The rest of this guide is the role each channel plays and how the best agencies allocate across them, because understanding the roles is the prerequisite for allocating well.

Demand Creation: Paid Social Is the Engine

Paid social — Meta (Facebook and Instagram) and TikTok, with others in the mix by category — is the primary demand-creation and scaled-acquisition engine for most D2C brands, and it is where leading agencies concentrate their creative firepower, because on these platforms creative is the single biggest performance lever, larger than targeting or bidding. This is worth dwelling on, because it explains what a leading D2C agency actually spends its time doing on paid social: it functions, at its core, as a creative-testing machine. It produces user-generated content and performance video at volume, tests relentlessly across hooks, formats, angles and offers, identifies the winners with statistical discipline, and feeds those winners into scaled, catalogue-aware campaigns. The volume and velocity of creative testing is the differentiator, because on a platform where creative decides performance and creative fatigues quickly, the brand that can out-produce and out-test its competitors' creative wins, and the one that cannot out-bid its way out of a creative-supply problem loses.

Because paid social creates the demand that the rest of the portfolio captures, it typically carries the largest share of the acquisition budget in a leading agency's allocation — but, crucially, it is also the channel most under-credited by last-click attribution, because demand it creates often converts later through a search click or a direct visit that gets the last-click credit. This is the central measurement problem of D2C media, and it is why a leading agency never judges paid social on last-click ROAS alone: the platform that generated the interest gets robbed of credit by the platform that closed it, and an agency optimising on last-click will steadily de-fund the engine of its own growth. Leading agencies solve this with incrementality testing — geo holdouts and media-mix modelling that reveal paid social's true, caused contribution — and allocate accordingly, protecting the demand-creation engine even when the last-click numbers make it look less efficient than the capture channels downstream.

TikTok deserves specific mention as a demand-creation channel with a distinct character: it is even more creative-native and trend-driven than Meta, rewards a different, less polished, more authentically-creator style, and reaches audiences and moments Meta does not. Leading agencies treat TikTok not as 'Meta with a different logo' but as its own creative discipline, and the brands that win on it are the ones whose agencies genuinely understand its native content culture rather than repurposing Meta assets onto it. The broader point across paid social is that demand creation is a creative game, and the leading agencies are, functionally, creative operations with media buying attached — which is exactly the shape a disciplined D2C performance practice takes.

Scaling Brand Demand: CTV and TV

CTV (connected TV) and, for some brands, linear TV, are the scaled brand-demand-creation layer that leading agencies add deliberately once the efficient channels start to saturate. The logic is a natural consequence of how demand works: as a D2C brand grows, it exhausts the cheap, high-intent demand that search captures and the early-adopter, easily-reached demand that paid social converts efficiently, and further growth requires creating new demand at larger scale among audiences the efficient channels are not reaching cheaply. CTV is the channel purpose-built for that — big-screen, sound-on, attention-rich demand creation at scale, now with far more measurability and targeting than linear TV ever offered. Leading agencies bring CTV in when the brand is ready, when the efficient channels are showing saturation (rising costs, falling incremental returns), and when the brand-lift it creates can be shown to pay back — not as a vanity channel to look big, and not before the fundamentals justify it.

The discipline that separates a leading agency's use of CTV from an average one's is measurement. CTV's effect is diffuse and delayed — it creates demand that converts later across other channels — so measuring it on last-click is even more misleading than measuring paid social that way, and an agency that runs CTV and then judges it on last-click ROAS will conclude, wrongly, that it does not work. Leading agencies measure CTV's incremental effect on the whole funnel — its lift on brand search, direct traffic, and the efficiency of the other channels — through geo experiments and media-mix modelling, because those are the only methods that can capture a channel whose value shows up everywhere except in its own last-click number. The recurring theme, across paid social and CTV alike, is that the demand-creation channels are systematically under-credited by last-click and require experiment-based measurement to allocate correctly, which is precisely why the quality of an agency's measurement, not the length of its channel list, is what determines whether it uses these channels well.

Demand Capture and Marketplace: Search, Shopping, Retail Media

Google Search and Shopping are the demand-capture layer of the portfolio — they catch the intent that paid social and CTV create, plus whatever organic demand exists for the category, and they do it efficiently because they meet the buyer at the moment of intent. This is high-intent, high-converting traffic, and leading agencies value it precisely for that — but they understand its defining limitation: search is fundamentally capped by how much demand exists to capture. You cannot scale search beyond the demand that other channels and the market create, so treating search as a growth engine to be scaled in isolation is a category error; it is a capture channel to be maximised in efficiency, and its true value can only be read in the context of the demand-creation channels feeding it. A leading agency runs search with ruthless efficiency and never mistakes a rising search number for growth it created, because much of that search volume is demand the creation channels generated and search merely closed.

This is also where the last-click problem bites hardest and where a leading agency's discipline is most tested, because search — sitting at the bottom of the funnel — receives the last-click credit for conversions that paid social and CTV drove, making it look like the most efficient channel in the portfolio when in reality it is often the most dependent on the others. An agency that allocates on last-click will keep shovelling budget into search and starving the creation channels, watching its blended efficiency look great even as its growth stalls — because it is harvesting an ever-larger share of a demand pool it is no longer replenishing. Leading agencies use incrementality to see through this, funding search to capture demand efficiently while protecting the creation channels that keep the demand pool full.

Retail media — chiefly Amazon, and increasingly other marketplaces and retail networks — has moved from optional to core for D2C, for a simple structural reason: a large and growing share of product search now begins on Amazon rather than Google, so a brand absent from retail media is absent from the place many buyers start. Leading agencies optimise the marketplace both as a demand channel in its own right and as a defensive necessity — if you are not there, a competitor or a reseller captures the buyer searching for your product. Retail media is operationally distinct (its own advertising systems, its own economics, the interplay of advertising and organic marketplace ranking) and margin-sensitive, and the best agencies run it with the same contribution-first discipline as the rest of the portfolio rather than treating it as a bolt-on. For many D2C brands, the honest picture is a blended Amazon-and-own-store economics, and a leading agency measures channel contribution across that blend rather than pretending the two are separate businesses.

Proof and Retention: Affiliates, Creators, Email & SMS

Affiliates and creators supply proof, reach and performance on a cost-per-acquisition basis, and leading agencies treat them as an integrated part of the demand-and-proof engine rather than a separate silo. Influencer and creator partnerships create the social proof and trust that make every other channel convert better — a prospect who has seen a creator they follow use your product converts more readily when your paid social ad reaches them — and creator content often feeds directly back into the paid-social testing machine as high-performing ad creative, closing a virtuous loop between organic creator reach and paid amplification. Affiliate and partnership programmes add incremental, performance-priced sales, and the discipline that separates a leading agency here is judging affiliate performance on incrementality — is this a sale you would not otherwise have made, or is the affiliate taking a commission on a customer who was already going to buy? — rather than on raw attributed volume, because affiliate channels are notorious for claiming credit on demand that was already captured.

Email and SMS lifecycle is the channel that average agencies neglect and leading ones obsess over, and the reason is fundamental to D2C economics: it is where the profit is actually made. In most D2C models, the first purchase is acquired at, near, or even below breakeven after fully-loaded CAC — the acquisition channels get you the customer at roughly the cost of the customer — and the money is made on the second, third and tenth purchase, at near-zero marginal cost, through well-built email and SMS lifecycle flows. This means retention is not a nice-to-have that sits downstream of the 'real' work of acquisition; it is the economic foundation that determines whether the entire acquisition portfolio is profitable, because if customers never come back, every channel is acquiring at a loss no matter how efficient it looks. A leading D2C agency therefore treats email and SMS lifecycle as core to the media strategy — building the welcome, browse-abandon, cart-abandon, post-purchase, win-back and replenishment flows that turn a breakeven first order into a profitable customer — and reads acquisition efficiency in the context of the retention economics, because the two are inseparable. An agency that runs brilliant acquisition and neglects retention is optimising the unprofitable half of the business and ignoring the profitable one.

How the Best Agencies Allocate Across the Portfolio

The allocation method is what ultimately separates leading agencies from the rest, because everything above — the channel roles, the last-click problem, the demand-creation-versus-capture distinction — resolves into a single practical question: how do you decide where the next rupee goes? An average agency answers by splitting budget across channels and optimising each to its own last-click ROAS, which, as we have seen, systematically over-funds the demand-capture channels that claim credit and starves the demand-creation channels that drive growth. A leading agency answers by allocating on incrementality and contribution margin, and reading the whole mix with media-mix modelling and holdout testing rather than last-click attribution. It asks, for each channel, 'how much incremental contribution does the next rupee here actually produce, accounting for this channel's role and its effect on the others?' and moves budget to equalise marginal incremental return across the portfolio — the point at which the next rupee produces the same incremental contribution wherever it goes.

In practice this means combining methods, because no single measurement approach is sufficient: holdout and geo experiments to establish the causal, incremental effect of specific channels (especially the under-credited creation channels); media-mix modelling to read the whole portfolio and its interactions over time; and platform and last-click data as tactical inputs within channels rather than as the basis for cross-channel allocation. The leading agency uses experiments to calibrate what is real and MMM to allocate across the mix, and treats last-click as useful for optimising within a channel but dangerous for deciding between them. This is more sophisticated and more expensive than reporting blended ROAS, which is exactly why most agencies do not do it, and exactly why doing it is a durable advantage — the agency that allocates on incrementality while its competitors allocate on last-click will, over time, fund the channels that grow the business while its competitors fund the channels that merely close what others started.

So 'which channels do leading D2C agencies optimise?' turns out to be the wrong question in isolation, and 'how do they decide how much to put in each?' is the right one. The channels a leading agency optimises are the ones that earn their place in the portfolio — and which ones those are, and in what proportion, is decided by the brand's margins, its stage, and its measured incrementality, not by a fixed template or the channel of the moment. The mix also evolves with scale in a predictable arc: early-stage brands lean heavily on paid social for demand creation and search for capture, because those are the efficient channels and the demand pool is not yet exhausted; as the brand grows and the efficient channels saturate, leading agencies add retail media to meet buyers where product search now starts, bring in CTV to create demand at larger scale, and invest more deeply in retention as the customer base grows large enough that repeat economics dominate. The mark of a leading agency is not the length of its channel list — anyone can list channels — but the rigour of how it decides where the next rupee goes, and the honesty with which it measures whether that decision was right.

Measuring the Portfolio in Practice: MMM, Incrementality and Attribution

Because the whole argument of this guide is that leading agencies allocate on incrementality rather than last-click, it is worth being concrete about how the three measurement approaches actually fit together in practice, since 'use incrementality and media-mix modelling' is easy to say and harder to operationalise. The three are not competitors but complements, each answering a different question at a different altitude. Multi-touch attribution and platform data answer the tactical, within-channel question — which ad, audience, creative or keyword is working inside a given channel — and they are genuinely useful for that, because inside a channel the confounding from other channels is smaller and the granularity is high. The mistake is not using attribution; it is using it to decide between channels, where its last-click bias systematically misallocates by crediting the closer over the creator.

Incrementality experiments — geo holdouts, audience holdouts, and on/off tests — answer the causal question for a specific channel or tactic: what happens to outcomes when this is present versus absent? They are the gold standard for establishing whether a channel is genuinely additive, and they are especially valuable for the demand-creation channels (paid social, CTV) that last-click chronically under-credits, because an experiment can reveal the caused lift that attribution hides. Their limitation is that they are episodic and channel-specific — you run a test, you get a read for that channel at that time, and you cannot easily run holdouts on everything simultaneously and continuously. So experiments calibrate: they tell you the true incremental value of specific channels, which you then use to correct the biases in your ongoing measurement.

Media-mix modelling answers the portfolio question — given all the spend across all channels over time, and controlling for seasonality, price, promotions and external factors, how much incremental contribution did each channel produce? MMM can read the whole portfolio and its interactions continuously in a way experiments cannot, but it is a statistical model with assumptions, so it is less precise than a clean experiment and can be led astray by poor data or bad specification. The sophisticated practice, which leading agencies follow, is to triangulate: use MMM to allocate across the portfolio, calibrate the MMM with incrementality experiments so its channel estimates are anchored to causal reality rather than to correlation, and use attribution and platform data to optimise within channels. No single method is sufficient; the rigour is in combining them, and an agency's answer to 'how do you measure across channels?' tells you immediately whether it has this rigour or is quietly running everything on last-click while using the word 'incrementality' as decoration.

Methodology & Fairness

A note on how to read this. This is an opinionated guide published by Fluxsy, not an independent ranking or audit. Where we describe how leading agencies operate we generalise from widely-observed practitioner experience, not from a specific named agency's confidential playbook, and where we name agencies we do so by public positioning, not as endorsements. We have avoided inventing statistics or 'typical' figures. The durable value is the framework — how to think about channel roles, allocation, and execution quality — which holds regardless of which specific channels, tools or agencies exist next year. Verify specifics against primary sources and your own measured data.