Which parts of running an online store lose value the moment somebody outside the company takes them over?
That version of the question produces a shorter list than the usual debate suggests. Work carrying commercial judgment specific to one business, such as how products are priced, which customers the company wants more of and what the brand will not do, tends to degrade when it is compressed into a brief. Work needing deep specialist skill in bursts, such as a replatforming project or a structured testing programme, often degrades in the opposite direction when it is kept inside and handled occasionally by somebody whose real job is something else.
The decision to outsource ecommerce management is usually taken one function at a time. Dividing the work by layer instead, with judgment in one place and execution in another, produces a cleaner arrangement and considerably fewer arguments in month four.
Judgment Stays Inside, Execution Can Travel
Most versions of this decision begin by splitting the business into departments and asking which boxes can move to somebody else. That framing makes the choice harder than it needs to be, because nearly every function contains two different kinds of work. Merchandising holds the judgment of what deserves promotion and the labour of building the collection page. Email holds the decision about what the brand says to a customer who has not ordered in nine months, and the labour of building the flow, checking how it renders and chasing a deliverability problem.
The layer worth protecting is the one where advantage accumulates. Judgment about your own customers improves because this company has sold these products to these people for years, and that improvement does not travel inside a handover document. HubSpot’s guidance on delegating search work states the limit directly: an agency “can’t be in your head; they don’t know what your brand means”, and it cannot reproduce “the relationships you have with other businesses, suppliers and partners”. The same guidance notes that a retailer remains “responsible for your overall growth strategy” however much of the execution sits outside.
In practice the distinction shows up in how the engagement is written. A supplier asked to decide what matters has been given something that is difficult to take back later, because the reasoning behind each decision accumulates on their side. A supplier asked to produce evidence, options and a recommendation, with the retailer making the call, leaves the understanding inside the business even when the people delivering the work change.
There is a genuine exception. A company with no internal capability cannot make good decisions simply by reserving the right to make them, and in that position buying judgment for a period is sensible. What matters is that both sides treat it as temporary and agree how the thinking moves back in. A dependency nobody has agreed to review rarely ends on its own.
Three Questions That Settle Most Cases
Three checks cover the majority of situations in which a business is weighing whether to outsource ecommerce management work. They give better answers when applied to a specific piece of work than to a whole department, and they will often split a department in two.
Does This Work Depend on Knowledge Only the Business Holds?
Supplier terms, margin structure, the reason a slow-moving product line is still stocked, the returns pattern nobody has written down, the complaint that arrives every January. A capable external team will ask about these things and learn some of them, but they cannot be transferred in a kick-off call. Work leaning heavily on that knowledge tends to produce output that is defensible in general and wrong in this particular store, which is an expensive kind of wrong because it looks reasonable in the review meeting.
Where the answer is yes, keeping the work inside is not the only available response. Sometimes the better move is to get the knowledge out of individual heads first. WD Market’s work with A.M.Ozoli started from exactly that position: pricing depended on “the memory of a few key people”, with critical business data scattered across “15+ disconnected Excel files”. In that state neither an external supplier nor a new internal hire could have taken on quoting, because there was nothing to take on except other people’s recall.
Is the Skill Needed Continuously or in Bursts?
This question decides most development and specialist roles, and it is often answered with a cost comparison when a utilisation estimate would be more useful. A store may need senior platform development for a handful of weeks in a year, spread across a replatforming decision, a peak-season hardening exercise and two integrations. Hiring for that pattern produces an expensive person who is under-occupied, which is uncomfortable for the budget and worse for the individual, who tends to leave and take the store knowledge with them.
The reverse holds where demand is steady. Customer service in a normal week, catalogue maintenance, order exceptions and supplier chasing arrive every day at a fairly predictable volume, and work of that shape usually becomes cheaper and better inside once there is enough of it to occupy somebody.
The awkward cases sit in the middle, at roughly half a full-time role, where neither answer is clean. A retained arrangement with fixed monthly capacity, or a hire whose role is deliberately widened to absorb adjacent work, are both reasonable responses. What rarely works is buying the same skill repeatedly as a series of short one-off projects, since each supplier spends the opening part of every engagement relearning a store that has not changed much since the last one.
Would Handing It Over Move an Asset Outside Your Control?
Some work can be delegated with no lasting consequence. Some carries an asset along with it. Klaviyo’s description of owned media is a workable test: owned channels are the ones a brand “controls directly”, giving “a direct line of communication” with customers rather than depending on somebody else’s platform decisions. A subscriber list, an analytics history, a store admin and an advertising account all belong in that category.
None of this is an argument against external delivery. It is an argument about whose name sits on the account. A retailer can have an agency build and run every message it sends while owning the platform, the data inside it and the relationship with the subscriber. The failure case appears during a supplier change, when the property, the tracking history or the list turns out to belong to somebody else. At that point the cost of leaving is not the notice period, it is the rebuilding.
Where It Makes Sense to Outsource Ecommerce Management Work
The table below describes tendencies rather than rules. The final column carries most of the weight, because it names the condition that flips each row, and established retailers usually find that two or three rows behave differently in their business than the general pattern would suggest.
| Function | Usually better held inside | Usually works well outside | What flips the decision |
| Pricing, margin and promotional policy | The rules themselves and the authority to change them | Building and maintaining the system that applies them | Whether the rules exist in writing or only in somebody’s judgment |
| Merchandising and range | What is stocked, featured and discontinued | Bulk catalogue production, enrichment and seasonal rebuilds | Catalogue size and how often the range changes |
| Brand voice and customer communication | What is said, to whom and in what tone | Flow build, quality assurance, deliverability and reporting | Whether anyone internal can write for this customer |
| Store development and integrations | The requirement and the acceptance criteria | Design, build, testing and release | Volume of change and how unusual the platform setup is |
| Conversion research and testing | Which problem is worth solving next | Research, hypothesis design, implementation and analysis | Traffic volume, and whether anyone internal reads behavioural data |
| Analytics and account ownership | Ownership, access control and the metric definitions | Implementation, maintenance and reporting | Nothing; ownership is the one item that should not move |
| Customer service | Policy, escalations and what recurring complaints mean | Volume handling, out-of-hours and seasonal peaks | How varied the questions are and how much product knowledge they need |
| Fulfilment and logistics | The service promise and the exception policy | Pick, pack, despatch and returns processing | Order volume, product handling requirements and market spread |
Two rows deserve particular attention. Analytics is the only line with no real trade-off, because implementation can sit anywhere while ownership sensibly stays put. Conversion work is the row most often misjudged, since the constraint is rarely the availability of ideas. A store without enough conversions to read a result inside a few weeks cannot run a serious testing programme regardless of who staffs it, and in that situation qualitative research and a technical review tend to be more productive than an experiment queue.
Customer Service: Coverage Against Consistency
Salesforce’s overview of outsourced customer service sets out both sides without much decoration. External providers bring “specialised knowledge and skills”, the “flexibility to scale operations up or down based on the business’s needs” and the ability to “operate across different time zones, offering 24/7 services”. Set against that, the same overview names loss of control over the process, “potential data security concerns” and “the need for continuous coordination” as real challenges.
For an established retailer the deciding factor is usually how varied the questions are. A store selling a narrow range with predictable queries about delivery and returns can move that volume outside with little loss, and gains genuine coverage during promotional peaks when internal staffing cannot flex. A store where answering well requires product expertise, or where a complaint is the first visible sign of a fault in a product batch, gives up something less obvious: the conversations stop reaching the people who would recognise the pattern.
The coordination cost is the part most often underestimated. Somebody internal has to own the relationship, read a sample of what is being said each week and push product and policy changes through in time. Where no one holds that responsibility, quality drifts slowly enough that the first clear signal may be a review score rather than a report.
Development Work: The Burst Pattern in Its Clearest Form
Development is where the utilisation argument is easiest to see and hardest to act on, because the emotional case for an internal developer is strong. The hybrid that tends to hold up is an internal person who owns the backlog, understands the store and can make small changes quickly, working with an external team for projects and specialist work. That arrangement addresses the weakness of both pure models.
A fully external model tends to leave nobody inside able to judge whether an estimate is reasonable or whether a proposed approach will create maintenance cost later, which turns every quotation into an act of faith. A fully internal model concentrates platform knowledge in one person and limits the business to whatever that person has seen before, which becomes visible during a migration or an integration that falls outside their experience. Neither risk is fatal, but they are different risks, and a business should know which one it has chosen.
The Option That Gets Skipped: Removing the Work
Before a task is framed as hire or outsource, it is worth asking whether it needs a person at all. Omnisend’s explanation of marketing automation puts the value in ordinary terms: it “eliminates the risk of manual error and ensures consistency and accuracy in repetitive tasks” and “reduces the workload on your marketing team, freeing them up to use their resources and time for more productive tasks”. Work that is repetitive, rule-based and high in volume is often a poor candidate for either a salary or a retainer.
The A.M.Ozoli engagement is a concrete version of this. Quoting had required roughly two hours of manual calculation per customer, and once the pricing logic was built into a system with margin-protection rules, the case study records quoting time moving to “under 30 seconds”. The interesting part for this decision is not the speed. It is that the commercial judgment stayed in the business, expressed as rules an employee or a partner could apply, instead of leaving with an individual or sitting inside a supplier’s process.
The obvious caution applies. Automating a process nobody has examined tends to make a poor process faster and more consistent, and rules written from an incomplete understanding produce errors at a scale manual work never reached. The sequence that works is to document the current workflow, decide which parts encode a real commercial rule and which exist only because somebody once did it that way, and automate the first group.
Key takeaway: the useful question is not which functions to hand over. It is which layer of each function carries judgment only your business can supply, whether the remaining work arrives steadily enough to justify a salary, and whether any of it could stop being work at all.
Handing Work Over Without Losing Control
Once a business has decided to outsource ecommerce management work of some kind, the quality of the arrangement matters more than the choice of supplier. The sequence below protects the things that are difficult to recover later.
- Write down the decision being delegated, not the task list. A brief listing deliverables produces deliverables. A brief stating which business problem the supplier is accountable for, and which decisions remain with the retailer, produces an arrangement that survives a change of scope. This single document prevents most of the disputes that surface around the third invoice.
- Agree what counts as success before agreeing the scope. Scope written first tends to become the definition of success by default, which rewards activity. Deciding first which two or three figures both sides will look at, and how long they need to become readable, changes what the supplier proposes.
- Keep every account in the company’s own name. Platform tooling supports this properly. Shopify merchants can add collaborator accounts for partners and “use permissions to control what sections of your store they can access”, as described in the user permissions documentation. Adobe Commerce offers restricted admin roles, where a user “can see and change data only for websites or stores that are associated with the role, but cannot change any global settings or data”, and accounts can be given expiration dates for temporary access. In Google Analytics, user management works on the principle that “the level at which you add a user determines that user’s initial access”, so administrator rights belong with people who will still be employed after the contract ends.
- Ask who will actually do the work. BigCommerce’s guidance on choosing an ecommerce agency includes a question worth asking in the first meeting: “Ask which services are delivered by in-house teams and which are outsourced.” A partner that subcontracts part of the work is not automatically a problem, but it changes who holds the knowledge, how quickly issues get resolved and what happens if that relationship ends.
- Start with a bounded piece of work. A diagnostic engagement or a single defined project reveals more about how a team thinks than any credentials presentation, and it costs less to end. It also produces something the business keeps regardless of what happens next, which a three-month trial retainer usually does not.
- Require the reasoning back, not only the output. Ask for the evidence behind each recommendation and for changes to be recorded as hypotheses with an expected effect. This is how the understanding accumulates inside the business rather than only inside the supplier, and it makes a later handover to an internal hire realistic.
- Fix the review point and the conditions that would end it. Agree in advance when the arrangement gets assessed, on what basis, and what result would justify bringing the work back inside. Arrangements without a review date are rarely reviewed, and the ones that quietly continue are not always the ones performing best.
What Turns a Reasonable Arrangement Into a Failure
Decisions to outsource ecommerce management are usually judged on their outcome, with little examination of how the arrangement was set up in the first place. Several recurring patterns explain a good share of the disappointment.
- Briefing by task instead of by problem. A supplier given a list of changes will make those changes, competently, and the underlying issue may remain untouched. The business then concludes that outsourcing does not work, when what failed was the instruction.
- Splitting diagnosis and implementation between two suppliers. The party that identifies a problem has no exposure to whether the fix works, and the party building it inherits conclusions it cannot interrogate. Where budget forces a split, the diagnosis should at least arrive with enough evidence for the second party to verify it.
- Selecting on hourly rate. Rate is comparable and therefore comfortable to compare, but it says little about how many hours a task will take or how much rework follows. Two quotations with a wide gap in hourly rate can end up costing much the same once the work is finished, and the lower rate is sometimes attached to the more expensive project.
- Leaving nobody internal responsible for the relationship. External capacity still needs direction, context and decisions. Where that responsibility is spread across three people who each have another job, the supplier ends up setting its own priorities, which is not what either side agreed to.
- Handing over the only copy of the knowledge. If the supplier holds the documentation, the analytics history and the reasoning behind past decisions, the switching cost grows quietly every month and eventually starts influencing decisions that should be commercial.
- Reviewing activity rather than outcomes. Monthly reports listing completed tasks are easy to produce and easy to approve. They also make it possible for an arrangement to run for a year without anyone establishing whether the store is performing better.
Drawing the Line in Your Own Business
The decision is easier once it stops being a choice between two organisational models. Judgment that depends on knowing this business, its customers and its suppliers belongs inside, along with ownership of the accounts and data that carry the customer relationship. Execution can sit wherever the work is done best, and the deciding factor is usually whether the skill is needed continuously or in bursts.
A practical next step is to take one function, split it into the decisions it contains and the work it generates, and test each part against the three questions above. Where a repetitive, rule-based task appears, consider whether it needs a person at all before choosing between a hire and a supplier. Retailers that decide to outsource ecommerce management successfully are usually the ones who did that separation first and wrote it down.
From a Split Decision to a Working Arrangement
If the split is clear but the execution side is not covered, an external team can take the research, build and measurement while the business keeps the decisions. WD Market works with established retailers in that arrangement through ongoing conversion and growth support, and where the priorities themselves need settling first, through ecommerce strategy work. Where the question is whether the current setup can support what the business wants to do next, a bounded technical audit is usually the cheaper place to start.
To discuss which parts of your operation are worth handing over, get in touch through the contact page with your platform, approximate monthly order volume and the functions currently causing the most friction. Shorter observations from this kind of work are published on our LinkedIn page as they come up.
Frequently Asked Questions
What should an ecommerce business never hand to an external team?
Ownership of the customer relationship and of the accounts that hold it, along with the authority to set pricing, range and brand policy. An agency can build and operate almost anything on top of those, but the subscriber list, the store admin, the analytics property and the advertising accounts should be registered to the company. Losing that ownership is rarely noticed until a supplier changes, at which point recovering it can cost more than the work being replaced was worth.
Is it cheaper to outsource ecommerce management than to hire?
It depends almost entirely on how much work there is. Where a skill is needed most weeks of the year, an employee is usually cheaper per hour once salary, employment costs and tooling are compared honestly against a retainer. Where it is needed for a few weeks a year, the comparison inverts sharply, because an under-occupied specialist still costs a full salary. Estimate weeks of genuine work before comparing rates.
Can one external team cover development, conversion work and marketing?
Some can, and the honest answer varies by supplier rather than by category. What matters is confirming which parts are delivered by the team you met and which are passed to someone else, since that affects how quickly problems get resolved. A single partner covering related disciplines removes a coordination burden that otherwise falls on the retailer, though it also concentrates dependency, so account ownership and documentation become more important rather than less.
How do we keep control when an external team has admin access?
Use the permission systems the platforms already provide instead of sharing a single administrator login. Create named accounts with the narrowest access the work requires, set expiry dates where the platform supports them, and review the access list quarterly. Keep at least one administrator on every critical system who is an employee rather than a contractor, and remove access on the day an engagement ends rather than when someone remembers.
What size of business does this decision usually affect most?
Retailers with enough volume to justify specialist work but not enough to keep several specialists occupied tend to face it most acutely. Very small stores generally have little choice but to buy capability externally, and large operations can usually fill each role. In between, the same work has to be covered with partial roles, which is where the split between judgment and execution earns its keep.
How long should a first engagement run before judging it?
Long enough to produce evidence, which depends more on traffic and order volume than on the calendar. A reasonable first milestone is a documented, prioritised set of problems with the reasoning attached, which a capable team can deliver within a few weeks. Judging revenue impact takes longer, and any supplier promising a specific figure before seeing the data is estimating rather than forecasting.