A DTC brand can hire an excellent media buyer and still miss the next growth decision if nobody owns inventory, offer economics, creative approvals, and the definition of a new customer. It can also hire an agency with strong channel skills and get little value if every test waits on an internal merchandising decision that never arrives. The question is not whether agencies or employees are inherently better. It is where the work and decision rights should live in this particular commerce business.
This comparison is for a founder or growth lead deciding how to operate paid growth over the next planning period. It covers an in-house team, an external agency, and a deliberately designed hybrid. It does not assume a revenue threshold, universal salary figure, or guaranteed cost advantage for any model.
The short answer: keep commercial truth inside; place execution by constraint
An e-commerce brand should retain final authority over product claims, inventory priorities, discount rules, customer and margin definitions, account ownership, and investment decisions. Hire internally when the critical work is continuous, tightly coupled to merchandising and operations, and large enough to keep the right specialists busy. Use an agency for a bounded capability gap when the brand has a decision owner and can give the partner timely evidence and access. A hybrid is often viable, but only when each side knows who proposes, approves, executes, verifies, and records a decision.
This is a governance recommendation, not a claim that an agency will lower CAC or that a hire will work faster. Both models fail when the company delegates a business decision without supplying the business facts.
What are the three operating models?
In-house means employees own most of the paid-growth planning and execution. It does not mean that nobody uses contractors, a platform specialist, or a creative studio. Agency-led means an external partner owns a defined scope of execution or diagnosis under a contract; it should not mean that the partner silently becomes the owner of the store, ad accounts, or financial truth. Hybrid means a named internal owner and named external operators divide work at explicit interfaces. It is not a euphemism for unclear ownership.
For a DTC company, those interfaces are unusually concrete. A campaign can drive demand for a product that cannot ship promptly. A promotion can make platform ROAS look attractive while compressing contribution. A returning customer can be credited as a purchase conversion even when the budget was intended to acquire first-time buyers. These are not problems a channel specialist can resolve from a campaign dashboard alone. Shopify's marketing reporting exposes first-time and returning customers, but its attribution model, connected cost data, and report definitions affect what appears; a dash for missing cost data is not zero spend (Shopify Help Center: marketing reports).
Map the commerce work before comparing fees
Start with the work that must happen, not job titles. One person might perform several tasks, but every task needs a decision owner and a quality check.
Workstream Why it changes the model Default company decision right Possible external contribution --- --- --- --- Product, inventory, and merchandising A winning ad can promote the wrong SKU, price, or availability state Which products and offers may be promoted Translate the approved priorities into campaigns and creative tests Margin and acquisition economics Discount, fulfillment, returns, and repeat purchase can reverse a platform-looking win Cost boundary, new-customer definition, acceptable payback, final spend Reconcile reports and propose tests against those guardrails Paid media and feed operations Campaign setup, product data, pacing, and troubleshooting can be recurrent specialist work Budget authority and approved product claims Build, operate, and document agreed channel work Creative and landing experience A concept must reflect the actual product and available proof Brand truth, substantiation, approval standard Produce variants and run a structured test backlog Customer lifecycle The acquisition promise can create returns or weak repeat behavior Customer policy, service reality, retention strategy Test messages or journeys within approved boundaries Measurement and experiments Platforms, store reports, and finance can disagree Source of truth, identity rules, experiment decision Implement checks, explain discrepancies, and recommend next evidence
This table is an editorial responsibility design, not a universal org chart. A small business may have one founder making several company decisions; a larger business may have finance, merchandising, and growth owners. What matters is that the external operator can get an answer before the relevant test window closes.
Shopify's customer-cohort report groups customers by first purchase and can show returning behavior in later periods. That makes it useful for separating the first-order story from the repeat-order story, subject to the merchant's definitions and available fields (Shopify Help Center: customers reports). For the detailed denominator reconciliation, use the blended CAC versus platform CPA framework. The operating-model decision here is who maintains that reconciliation and acts when it changes.
Compare agency, in-house, and hybrid by the binding constraint
If product and promotion decisions change daily
An internal growth lead has an advantage when new inventory, pricing, bundles, support issues, or product launches alter what may be advertised each day. They can join the conversations that create those decisions. That does not automatically require an internal media-buying department. The lead may keep offer and prioritization decisions inside while an external specialist implements a limited paid-search or paid-social scope.
Ask whether the delay is truly in campaign execution or in approval. If the agency receives an approved offer three days late, replacing it with a hire does not repair the process unless the new hire can actually make or obtain the decision faster. Conversely, an agency's specialization does not compensate for repeated missed merchandising handoffs.
If a specialist gap is urgent but not yet a permanent job
An agency may be sensible for a defined feed-quality audit, Google Ads restructuring, Meta creative-testing system, attribution reconciliation, or launch experiment. The engagement should state the question, decision owner, access required, deliverables, and end condition. Do not turn intermittent uncertainty into a permanent retainer by default, but do not force one generalist to cover every platform and measurement task either.
The buyer should ask what the partner will actually do, who will do it, and what the brand must supply. “Full service” is not a work specification. The partner's cross-account experience may be useful for generating hypotheses, but it is not proof that another brand's result will transfer to your assortment, margin, and audience.
If execution volume is stable and context-heavy
An internal hire becomes more plausible when there is enough recurring work in a specific function, a qualified manager, access to creative and data support, and a reason for daily product context to matter. A single media buyer cannot also be assumed to own photography, product feeds, landing-page development, cohort analysis, customer service, and finance. Write the supporting-team requirements into the hiring plan before comparing that hire with an agency fee.
The in-house case can also weaken when the growth problem changes faster than the skill profile. If the brand needs a one-off measurement repair now and a creator production sprint next quarter, one permanent role may not cover both. A designed hybrid can keep a strong internal commerce owner while buying the varying capabilities as scoped work.
If measurement itself is the bottleneck
Do not resolve a reporting disagreement by changing the agency or hiring a channel operator first. Shopify describes marketing-attributed sales and orders as reports dependent on campaign and third-party data, and its marketing reports permit different attribution views (Shopify Help Center: measuring marketing performance). The first job may be to specify the cost boundary, customer identity, window, refunds treatment, and reporting cadence. The contribution-margin guide handles the economic calculation; this page asks who owns and validates the inputs each week.
A capacity worksheet without invented break-even numbers
The typical comparison puts an agency retainer opposite one salary and calls the lower figure a winner. That omits missing specialists, management time, turnover, creative production, software, measurement, and the cost of delayed or poor decisions. It may also assume that a retainer buys a capability that is not in its scope. Use the brand's own numbers instead.
For each workstream above, record four fields over the next planning period:
Decision frequency: how often must the business make a call, and what happens if it waits?
Recurring execution load: how many qualified hours or deliverables are actually required, including review and rework?
Context sensitivity: what product, customer, supply, or financial knowledge must the operator obtain to perform safely?
Reversibility: can the work be paused, transferred, or narrowed without losing an account, dataset, asset, or crucial learning?
Then cost the feasible operating models for the same scope. For an internal option, include compensation, recruiting and management, tools, specialist support, and the portion of creative or analytics capacity it consumes. For an agency option, include fees, internal brief and review time, production excluded from scope, implementation dependencies, and transition work. For a hybrid, include the internal owner plus the scoped external work. Do not add these as a fake precision “agency score”; use them to reveal what is missing and where a capacity assumption fails.
The break-even question is not merely “Which invoice is smaller?” It is “Which feasible model can make and implement the next material decision with the required evidence and without losing commercial control?” If neither model meets that test, defer the commitment and repair the decision system first.
A fictional example of how the worksheet changes the answer
Imagine a skincare DTC store with frequent bundle changes, a working email program, weak product-feed hygiene, and a small internal marketing lead. The founder initially compares an agency retainer with the cost of one new paid-media hire. That comparison misses two facts: the current bottleneck is product-data quality, and the brand needs an internal owner to approve which bundles can be promoted at which margin.
A plausible hybrid would have the marketing lead own offer priorities, margin guardrails, product evidence, and weekly budget approval; a scoped specialist would audit the feed, repair the approved fields with the appropriate store operator, and run a bounded acquisition test. The brand would keep the store and advertising accounts, the raw creative, and the test record. After the feed work is stable, the team would recalculate the recurring paid-media workload. It might then hire, continue a narrow agency scope, or stop the external work. Nothing in this example is a Sharply Labs client result, salary benchmark, or forecast. A different brand with stable merchandising and high-volume daily creative work could make a different choice.
Keep the store, accounts, and data recoverable
Account ownership is a procurement issue, not an onboarding footnote. Shopify documents collaborator accounts for external Partners and says a merchant controls their permissions and can remove access. It also distinguishes a collaborator relationship with a merchant-owned store from a store transferred to the partner (Shopify Help Center: collaborator accounts, client-store collaborations). Give only the permissions the work needs; do not share an owner login or transfer a store just to start an agency engagement.
Google Ads manager accounts can link to existing client accounts; linking does not convert the client's account into the manager account, and a client administrator can unlink the relationship. Ownership and user-management rights are separate settings, so inspect them deliberately (Google Ads Help: manager accounts, linking manager accounts). Merchant Center also has differentiated access roles; its documentation warns that removing a person associated with website verification can affect the claim until verification is established elsewhere (Google Merchant Center Help: people and access). These controls vary by configuration; verify the actual account before changing a role.
The company's minimum handoff inventory should include store ownership, billing, ad and Merchant Center access, product-feed sources, pixels and conversion definitions, analytics and reporting permissions, domain and landing-page rights, creative source files and usage rights, audience rules, active tests, and a record of why material budget choices were made. The brand can let an external team operate many of these systems without surrendering the ability to inspect or recover them. Review the access map before the relationship ends, not after an account is locked or an integration stops working.
Questions that expose a weak operating-model proposal
Ask an agency or prospective hire to work through the same short case: a campaign has improved attributed purchase CPA, but first-time buyers are flat, a featured SKU is nearly out of stock, and return rates have risen. Do not ask for an instant diagnosis. Ask which facts they would request, who can authorize a pause or product change, which metric would be compared on a common window, and what they would do before increasing spend.
A credible answer should distinguish a campaign signal from a business outcome. It should ask for inventory and order/cohort data, identify the missing cost and attribution definitions, and avoid claiming causation from a dashboard movement. The MER-versus-ROAS framework shows how a business-level efficiency measure and a platform optimization signal can coexist without being interchangeable. A partner who can articulate that difference is more useful than one who recites a universal target ROAS.
Then ask who will:
decide what the product promise may say;
approve discount depth and contribution guardrails;
reconcile first-time-customer, repeat-order, and platform action counts;
own the creative brief and source files;
inspect feed and landing-page readiness before spend;
document the test and stop rule; and
make the final incremental-budget call.
If the answer to most of these is “the other side,” the problem is not the fee model. It is an unowned operating system.
Write a 30/60/90-day contract before committing to a long model
The periods here are planning checkpoints, not a promise that any platform will learn, an experiment will reach significance, or revenue will change in a fixed number of days. Adjust them to the brand's actual decision cycle.
First 30 days: establish ownership and baselines. Name the internal decision owner and external or internal operators. Inventory access, active campaigns, offers, feed sources, attribution settings, customer definitions, refund treatment, and contribution assumptions. Agree on a report the parties can reproduce and a small list of hypotheses. If the data cannot be reconciled, say so before assigning performance credit or blame.
By 60 days: test a bounded workstream. Run one meaningful change for which the owner, implementation, success measure, guardrail, and stop condition were specified beforehand. It might be feed repair, an offer/landing alignment test, or a new creative process. Record what was learned and what remained ambiguous. The point is to observe whether the operating model can make decisions, not to manufacture a “winner.”
By 90 days: choose the next model based on observed workload. Review the actual decision delays, specialist load, quality of creative and reporting, account recoverability, and the business outcomes that are measurable within that period. Decide whether the scope should stay external, move inside, be split differently, or stop. Do not compare two models on different responsibilities and call the result evidence.
The record should survive personnel and vendor changes. A one-page decision log with the question, owner, data source, action, result, uncertainty, and next call is more useful than a large deck no one can operate from.
When not to hire an agency—or a new employee—yet
Delay a broad engagement when the company cannot identify a decision owner, the product claims or offer rules are unsettled, the store and ad-account access boundary is unclear, or the acquisition target is a platform action without a compatible business definition. A narrow diagnostic may still help, but do not purchase “scale” to solve a governance problem.
Delay a permanent specialist hire when the work is intermittent, the job description combines unrelated disciplines, there is no manager who can prioritize the work, or the required creative and analytics partners do not exist. A good hire cannot make a missing operating system appear by themselves.
Do not treat an agency as a substitute for commercial accountability or an in-house team as a guarantee of speed. Choose the smallest accountable arrangement that can answer the current decision, and write down what evidence would cause you to change it.
A scoped e-commerce growth conversation
For DTC founders and growth leads deciding whether to hire, outsource, or redesign a hybrid, Sharply Labs' e-commerce growth practice can review one live acquisition workstream: its product and offer cadence, paid-channel and creative workload, contribution and new-customer definitions, account access, and decision handoffs. We would return a proposed responsibility map and a bounded first-test agenda, including unresolved data questions. Our growth services are relevant only where a defined execution or measurement gap fits. This conversation does not promise that an agency is the right answer, a lower CAC, a target ROAS, or a particular revenue result.
Sources and scope
Shopify Help Center — Marketing reports: reporting fields, attribution settings, and incomplete cost data; not proof of campaign incrementality.
Shopify Help Center — Customers reports: first-purchase cohorts and repeat behavior; availability depends on report configuration.
Shopify Help Center — Measuring marketing performance: marketing-attributed sales and order definitions; not a finance ledger.
Shopify Help Center — Collaborator accounts and client-store collaborations: merchant-controlled external access, not evidence that a specific agency setup is safe.
Google Ads Help — About manager accounts and linking an existing account: linking and ownership mechanics; inspect the live account before making access changes.
Google Merchant Center Help — Manage people and access: role and website-verification caveats; account configuration varies.