A line arrives in the monthly marketing report: mobile visitors are dropping out at the delivery step. It moves to the retailer’s inbox, then to the developer who maintains the theme, who reasonably asks which change is being requested. Nobody in that chain has watched a recording of the delivery step, and nobody is paid to. Six weeks later the same line appears again, slightly reworded.
Closing that loop takes someone who can diagnose the behaviour, decide what to change, build it and then read the result. An ecommerce development agency covers a real part of that work, though the part varies far more than the label suggests, and what is left over usually decides whether the store improves at all.
Why the Store Side of the Funnel Often Has No Supplier
Agency categories describe how contracts are actually written, and the boundaries between them are where store improvements tend to get lost. Shopify’s guide to the types of ecommerce agency separates marketing and growth agencies, focused on acquisition and retention through paid media, SEO, email, SMS and conversion rate optimisation, from development and build agencies, focused on “the architecture of ecommerce websites, design theme development, custom app integrations, and platform migrations”.
The guide is blunt about the limits of the first group: they “generally don’t build or technically maintain your ecommerce store”. The reverse holds too, since development agencies “don’t manage ad budgets”. Two competent suppliers can therefore sit on either side of the buying experience without either being accountable for it, and neither is failing at what it was engaged to do.
Internal job descriptions often reproduce the same split. BigCommerce’s breakdown of ecommerce team roles describes marketing roles as responsible for “promoting the business, build brand awareness and acquire customers”, and developers as people who “build and maintain the ecommerce website and other software systems the business uses”. Both are legitimate mandates. Deciding which changes to the buying experience are worth making sits in neither, which is why it frequently sits nowhere.
The tooling mirrors the split. Advertising dashboards report impressions, click-through rate, cost per click and conversions, which describe the journey up to the click well and the journey after it barely at all. Microsoft built its Clarity integration for Microsoft Ads so advertisers can “move beyond basic ad metrics to understand what users actually do after clicking a Microsoft Ad”. That such an integration needed building is itself informative.
What an Ecommerce Development Agency Covers, and What It Leaves Out
An ecommerce development agency is the supplier most established retailers reach for once they accept the store itself needs work, and it is a good fit for anything with a defined technical shape: a theme rebuilt around a different navigation model, a checkout extension, a replatforming exercise, an ERP integration, a performance problem that traces back to how the storefront is assembled. In each case the hard part is building the thing correctly, and a specification can be written before work starts.
The limitation is structural rather than a matter of competence. A development agency builds what it is asked to build, and the asking usually originates elsewhere. When a request arrives as a symptom rather than a specification, and “mobile visitors abandon the delivery step” is a symptom, the agency can propose a change on judgement, which is guesswork dressed as delivery, or return the problem to the person who raised it precisely because they could not answer it.
This is why a comparable budget produces excellent results at one retailer and disappointing ones at another. A business that already has someone able to specify changes and read the outcome gets a great deal from development capacity alone. A business without that person has bought throughput, and throughput applied to unverified assumptions tends to produce a busy release schedule and a flat conversion rate. Incentives compound this: an agency retained for build capacity is measured on what it ships, so its natural bias is towards shipping.
Four Partner Types and What Each Actually Owns
Most established retailers are choosing between four realistic arrangements. The useful comparison is not which is strongest overall, since that depends on what the business already holds internally, but which responsibilities each genuinely accepts.
| Partner type | Typically owns | Usually does not own | Best suited to | Main risk |
| Performance or marketing agency | Traffic, acquisition cost, channel mix, campaign creative and reporting | Storefront changes, technical maintenance, platform decisions | Businesses whose store converts acceptably and whose constraint is demand | Rising acquisition cost hides a site problem that more budget cannot solve |
| Ecommerce development agency | Build quality, integrations, platform work, technical delivery and release process | Deciding which changes are worth making, or interpreting behavioural data | Retailers with a clear internal specifier and a defined technical backlog | Fast delivery of changes nobody has validated |
| Independent CRO specialist or contractor | Research, hypotheses, test design and analysis | Implementation capacity, platform engineering, ongoing maintenance | Stores with strong development support already in place | A queue of recommendations that waits months for a developer |
| External ecommerce growth team | Diagnosis, prioritisation, design, build and measurement as one loop | Media buying and campaign management in most engagements | Retailers with traffic, no internal owner, and a store that has stopped improving | Higher entry cost, and concentration of knowledge in one supplier |
The fourth row is the model WD Market operates, so it is worth being clear about where it is the weaker answer. It carries a higher cost floor than a contractor. It rarely matches a specialist paid-media agency on channel depth, which is why it usually sits alongside one rather than replacing it. It also concentrates institutional knowledge in a single supplier, a genuine continuity exposure worth pricing in. A store with a capable ecommerce manager, a responsive development partner and a working testing habit does not need it.
Where it earns its place is narrower than most agency marketing implies. A drop-off at the delivery step may be a usability defect, a shipping rule that surprises people, a slow address lookup, or a platform constraint that makes the intended design impossible. Those four diagnoses point at four different teams, and separating diagnosis from delivery makes each of them someone else’s problem.
Signs the Store Experience Has No Supplier
The gap is easier to recognise from its symptoms than from an organisation chart, because the chart usually looks complete. These patterns often appear together:
- Conversion stays flat across several quarters while traffic grows. The business is buying growth rather than earning it, which becomes expensive at the point where acquisition costs rise faster than average order value.
- Behavioural tools are installed but rarely opened. An unreviewed recording account usually indicates that nobody has been given time for the work, rather than that the store has no problems worth finding.
- Site changes are described as requests rather than decisions. Where improvements move through a ticket queue with no stated hypothesis, the store will change but may not improve, and nobody will be able to say which it did.
- The marketing report and the platform report disagree, and nobody resolves it. Unreconciled numbers often mean no single person is accountable for what the store earns, as opposed to what each channel claims.
- Nobody can name the last three changes made to improve conversion. If the answer is a redesign from two years ago, the store has had maintenance rather than improvement, and the two are easily confused in budget discussions.
None of these is conclusive alone. Two or three together usually indicate the buying experience has been treated as infrastructure to keep running rather than as an asset with an owner. A useful first test is to write down which named person or supplier is accountable for conversion rate, then notice how long the answer takes. Our article on who owns ecommerce growth examines the same gap from the internal side.
Why Asking the Existing Developer Rarely Closes the Gap
The obvious move is to send the problem to whoever maintains the store. That works when the request is already well formed and works poorly when it is not, and marketing reports tend to produce the second kind.
A developer asked to fix mobile checkout abandonment has been handed an outcome, not a task. Doing it properly means reviewing recordings, segmenting by device and browser, checking whether the pattern predates a recent release, testing payment methods on real hardware and forming a view on which of several plausible causes is operating. That analytical work competes directly with a maintenance queue that has deadlines attached and visible consequences when it slips.
Growth work rarely feels as urgent as a broken payment method or a campaign launch. Given a queue containing both, most people reasonably clear the urgent items first, and the analysis moves to next month indefinitely. The failure is not individual. It is a scheduling reality wherever diagnosis and maintenance share one person’s capacity.
Internal expert input required: add a short, non-confidential example of a request that stalled between a client’s marketing agency and their developer, including what closing the loop actually involved and how long the diagnosis took once someone owned it.
Key takeaway: a store improvement request that arrives as a symptom needs someone whose job is to turn it into a specification. Without that step the request either waits, or it becomes a change built on a guess that nobody afterwards can evaluate.
Access, Data and Continuity: the Mechanics That Decide Whether It Works
These engagements often succeed or stall on unglamorous details settled in the first fortnight. The permissions a partner holds determine what it can diagnose, and the account ownership arrangement determines what the retailer keeps when the relationship ends.
Store access is a category decision, not a password
On Shopify, external partners are expected to use collaborator accounts rather than staff logins. The documentation describes collaborators as “Shopify Partners who you’ve allowed to access your store or organization”, notes they “don’t count towards your store’s user limit”, and sets out a request flow in which the partner supplies a four-digit access code and the owner approves it and assigns a role that can be changed at any time. Sharing a staff account instead gives up all of that control.
WooCommerce reaches a similar place differently. Its roles and capabilities document defines a Shop manager role with “full access to all WooCommerce settings and configuration” and the ability to “manage products” and “manage orders”. That is an operational role rather than a development one, so template or plugin work needs broader access, granted deliberately and alongside a staging environment.
Analytics permissions quietly define what a partner can prove
The Google Analytics roles and data restrictions documentation describes Viewer as a role that “can see settings and data”, Analyst as adding the ability to “share created explorations to other users of the property”, and Editor as “full control of settings at the property level” without user management. Two data restrictions can also be applied, No Cost Metrics and No Revenue Metrics, which suppress those figures across reports, explorations, audiences, insights and alerts.
A partner engaged to improve revenue, but given a role carrying the No Revenue Metrics restriction, can describe behaviour and cannot connect it to money. Roles also inherit downwards from account to property by default, so an arrangement made once for convenience may persist across properties added later. Neither issue announces itself, and both tend to surface months in as a reporting argument.
Behavioural history is an asset, and it may not be in your name
Microsoft Clarity is instructive because its team management documentation is explicit about consequences. Projects support two roles, Admin and Team member, where a member has read and write access but cannot change roles, manage settings or delete the project. The documentation warns that “if you are the only project member and remove yourself, the project will be deleted”. Where an agency created the project on its own account, a year of recordings is a supplier asset, and that is one of the few things in this work that genuinely compounds.
How to Add Store-Side Ownership Without Replacing Your Marketing Agency
Retailers frequently assume the choice is binary. In most cases it is not, because the two mandates barely overlap. A workable sequence looks like this:
- Write down who is accountable for conversion rate today. If the honest answer is that it is shared between an agency and a developer, the gap is confirmed and the rest of the process has a purpose.
- Establish what the current data can and cannot answer. Check whether purchases are tracked reliably, whether behavioural tools are recording, and who holds administrative rights on each account. Scoping against broken measurement produces a plan nobody can evaluate.
- Separate the diagnostic engagement from the delivery engagement. A bounded diagnostic phase, whether an ecommerce technical audit or a structured review, gives both sides evidence before either commits to something longer, and produces a prioritised list the retailer keeps regardless of who does the work.
- Agree the interface with the marketing agency explicitly. Landing page findings, campaign calendars and planned site changes each affect the other party, and a short recurring call usually prevents more problems than a contract clause.
- Define success in figures both suppliers can see. Conversion rate by device, revenue per session and completion rate through the checkout steps keep the review honest, because channel metrics and site metrics answer different questions.
- Set a review point with a decision attached. Name the date, and name the three outcomes it can produce: carry on, change the scope, or stop. Arrangements without that date tend to survive on inertia, because ending one is socially awkward in a way that renewing one is not.
What a Combined Engagement Looks Like in Practice
WD Market’s work with RIPO International, a manufacturer selling insect nets across several European markets, shows the kind of problem that sits between the usual supplier categories. The published case study records that each market ran its own website with separate maintenance and inconsistent branding, that “the purchase flow had friction points that were causing drop-off right at the moment of commitment”, and that the site “was built purely for B2C”, leaving wholesale buyers without a journey designed for them.
The services listed on that page include “Full UX/UI Redesign”, “Multi-Country Platform Consolidation Into One Site”, “Checkout Flow Optimisation” and “B2C And B2B Journey Separation”. The reported outcome is revenue growth from “€200K → €500K annual turnover”, alongside a consolidated multi-market platform and reduced final-step drop-off.
That figure covers the engagement as a whole, and no responsible reading attributes it to any single item on the list. The transferable observation is about the shape of the problem. Consolidating country sites is platform engineering. Separating the B2B journey is commercial strategy. Removing checkout friction is behavioural work. Presented separately to three suppliers, each would have been scoped as a project, and the question of which to do first would have had no obvious owner.
Mistakes That Leave the Store Unowned for Another Year
The most common error is treating a conversion problem as a demand problem, because increasing budget is a decision a business already knows how to make. Spend sent to a store with a structural friction point tends to increase the number of people who meet the friction, so the effect often surfaces as a rising acquisition cost rather than as a visible site failure. Our analysis of stores where conversion has plateaued covers that pattern in more depth.
A second is commissioning a redesign in place of a diagnosis. Redesigns are satisfying to approve and reliably change something, but a rebuild that was not informed by evidence may move a problem rather than remove it, and it resets whatever behavioural baseline existed.
A third is selecting an ecommerce development agency primarily on rate. Rate comparisons are only meaningful when two suppliers are accepting the same responsibilities, and in this category they rarely are. Shopify’s guidance advises agreeing “what success looks like in concrete, measurable terms” before work begins and writing those goals into the contract, which compares better than an hourly figure because it forces both parties to state what is included.
The fourth is leaving the marketing agency out of the arrangement entirely. Page changes made without reference to live campaigns can invalidate creative written against the previous page, and campaign changes made without reference to a running test can contaminate its results. Both are avoidable, and neither is avoided by contract alone.
Questions Worth Asking Before You Sign
The answers matter less than whether a prospective partner can give them directly. Vagueness here tends to become vagueness in reporting later.
- Who writes the specification, and on what evidence? If the retailer provides requirements, the engagement is delivery capacity, which may be exactly right as long as everyone knows that is what has been bought.
- Which accounts will be created in our name? Analytics properties, behavioural tools, staging environments and repositories are assets, and establishing ownership at the start is far easier than at the end.
- How will we tell whether a change worked? Look for a stated method and a stated period, rather than a promise to report on results, which every supplier offers and which describes no particular practice.
- What would you need from our marketing agency? A partner with no answer has probably not worked alongside one, and that interface is where a meaningful share of the value is created or lost.
Matching the Supplier to the Work the Store Needs
Who should improve the store depends on what the business already holds. A marketing agency is engaged to acquire visitors and generally does not build or maintain the store. An ecommerce development agency builds well and needs someone to decide what to build. A specialist produces hypotheses and needs implementation capacity behind them. A combined external team closes that loop and asks the retailer to accept a higher cost floor and knowledge concentrated in one supplier.
Before comparing proposals, settle two things internally: who is accountable for conversion today, and whether the measurement in place could show that a change worked. Those answers narrow the choice more than any supplier comparison will.
From a Traffic Problem to a Store With an Owner
Where traffic is holding and the store has stopped improving, a bounded diagnosis is a safer first purchase than a retainer, because it produces something the business keeps whoever does the work that follows. A WD Market conversion rate optimisation engagement opens by establishing what the existing measurement is capable of proving, then locates where revenue is leaving the current journey and hands back a ranked set of problems with the evidence behind each one. Retailers whose question turns out to be commercial rather than technical are usually better served by ecommerce strategy consulting.
To talk through your setup and what a first engagement would cover, contact the WD Market team and include your platform, rough monthly traffic and anything already attempted. Shorter notes from this kind of work go out on our LinkedIn page as we write them.
Frequently Asked Questions
Is an ecommerce development agency the same thing as a marketing agency?
No, and the distinction is usually written into both contracts. A marketing agency brings people to the store and is measured on traffic, acquisition cost and channel performance. An ecommerce development agency builds and maintains the store itself, covering theme work, integrations, platform changes and technical delivery. Some full-service firms combine both, but that should be confirmed rather than assumed, because the labels are used loosely across the market.
Can our existing marketing agency take on store improvements instead?
Sometimes, and it is worth asking directly. Agencies offering conversion work alongside media buying do exist. Settle whether they can implement changes or only recommend them, whether they have people who work on storefronts rather than campaigns, and how a conflict would be handled if a site change were likely to reduce short-term campaign performance while improving overall revenue.
Do we need to end the marketing relationship to bring in a store-side partner?
Rarely. The two mandates overlap very little and most established retailers run both. What needs deliberate attention is the interface: a shared view of planned site changes and campaign activity, agreement on which figures both parties treat as the score, and a route for one to raise findings with the other. Left undefined, the suppliers can unintentionally interfere with each other.
What should we hand over in the first two weeks?
Administrative access to analytics and behavioural tooling, a store account with an appropriate role, access to a staging environment, and a short written history of what has already been changed and tried. The last item is most often skipped and the most valuable, because it stops a partner spending its first month proposing things the business already tested.
How long before store-side work shows up in revenue?
It depends on traffic volume more than on effort. Stores with substantial traffic can read the effect of a meaningful change within weeks, while lower-volume stores may need considerably longer before a result is distinguishable from normal variation. Any supplier offering a fixed timeline before seeing your data is guessing. A reasonable first milestone is an evidenced list of prioritised problems.
How do we judge whether a store-side partner is actually working?
Look for three things at the first quarterly review. There should be a record of what changed and why, expressed as hypotheses rather than a task list. At least one change should have had its effect measured and reported honestly, including any that did not work. And the prioritised problem list should have moved, with items closed and new ones added from evidence rather than opinion.