Ecommerce digital agency: the four capability stacks and how to judge the work

An ecommerce digital agency is a firm that runs the commercial side of an online store: acquisition, the storefront experience, merchandising and product data, and retention. Most are strong in one of those four and competent in the rest, so the first step in hiring one is deciding which of the four is your constraint right now. Then match the engagement to your stage: a project fee for a build or a replatform, a monthly retainer for growth work, and a defined scope with named people for both. Judge the firm on contribution margin per order and repeat purchase rate, not on return on ad spend alone, because ad spend return can rise while the business makes less money.

That last point is where most agency relationships go wrong. A store can look like it is winning on a dashboard while discounting, shipping costs and returns quietly eat the gain. Agree the numbers that count before the work starts and the whole relationship gets easier.

The four capability stacks

Stack	What it covers	Hire for it when
Acquisition	Paid search, paid social, shopping feeds, marketplaces, affiliates, creative testing	You have a product that converts and need more qualified traffic
Storefront and conversion	Theme development, site speed, navigation, product and category pages, checkout, testing	Traffic is healthy and conversion rate or average order value is the problem
Merchandising and product data	Taxonomy, attributes, titles and descriptions, imagery standards, search within the site, feed quality	Shoppers cannot find products, or feeds get disapproved and categories underperform
Retention	Email and SMS flows, segmentation, subscriptions, loyalty, post purchase experience	Acquisition cost is rising and repeat rate is flat

Diagnose before you shortlist. Take last quarter's numbers and work out where the funnel actually loses money: sessions, add to cart rate, checkout completion, average order value, gross margin after shipping and returns, and the share of revenue from repeat customers. The weakest of those points at the stack you should be buying.

Match the engagement to your stage

• Launch. A fixed scope project: platform setup, theme, product data, payments and shipping, analytics, and a launch plan. Insist on a written handover that includes documentation and training, because the gap between launch and competent day to day running is where new stores stall.
• Replatform. The riskiest project in ecommerce, and the one where agency quality shows. The scope must name the URL redirect map, product and customer data migration, the historical order import, SEO preservation and the rollback plan. Ask for a replatform they have done in your platform pair, and ask what broke.
• Growth. A retainer with a quarterly plan, a named team, a testing roadmap and monthly reporting against agreed numbers. The best version includes a backlog you can see and reprioritise.
• Peak season support. A short, well defined engagement for the weeks around your biggest trading period. Book it months ahead, because good firms are full by autumn.
• Audit only. A paid audit from a specialist is the cheapest way to find out whether you need an agency at all, and it gives you a scope to brief with.

Platform specialism is a real factor

Agencies build deep habits around a platform, including app stacks, theme architecture and the shortcuts that avoid known problems. For storefront and build work, a firm that lives in your platform will move faster and break less than a generalist. For acquisition and retention work, platform specialism matters far less than category experience and media skill, so do not pay a premium for it there.

Category experience cuts both ways. A firm with several stores in your niche understands the buyer, the seasonality and the creative that works. Ask directly whether they work with a direct competitor and how they handle the separation of data and ideas.

Keep ownership of everything that matters

Write these into the contract before the first invoice:

• Your ad accounts, analytics, merchant centre, domain and store admin stay in your legal name, with the agency as a user you can remove.
• Creative source files, product photography and copy are yours on payment, in editable formats.
• Code written for your store is delivered in your repository, not theirs, with a readable commit history.
• Email and SMS lists, with consent records, remain yours and exportable at any time.
• Documentation for any custom build, good enough that another firm can pick it up.

Be careful with any agency that resists these points. Ownership clauses are where a routine supplier change turns into an expensive rebuild.

How to judge the work once it starts

Set up a monthly review on five numbers, with a target for each:

• Contribution margin per order, after discounts, shipping, payment fees and returns. This is the one that decides whether growth is worth having.
• Blended customer acquisition cost, total marketing spend divided by new customers, which cannot be flattered by platform reported conversions.
• Conversion rate and average order value by device, because mobile usually carries the problem.
• Repeat purchase rate and time to second order, the clearest early signal of whether the store is building a business or buying revenue.
• Product level performance, so merchandising decisions have evidence behind them.

Ask for the test log too: what was tried, what the result was, and what happens next. An agency that cannot show you a list of tests and outcomes is running tasks rather than a programme.

What a good proposal contains

A specific diagnosis of your store, not a general pitch. The named people who will work on it and their time per month. A first ninety day plan with sequencing. The numbers they will be judged on, chosen with you. A clear price broken into fees and media. Notice period, ownership terms and handover terms. If a proposal is mostly case study logos and a monthly figure, ask for the rest before you compare it with anything else.

Warnings worth taking seriously

Be careful with promised revenue multiples and guaranteed returns on ad spend. A firm that commits to a number it does not control is pricing risk into the fee or planning to claim credit for sales you would have made anyway. Equally, treat any proposal that leads with the number of social posts per month as an activity plan rather than a growth plan. The useful promises are about process: what will be audited, built, tested and reported, and by when.

This page covers hiring, scoping and managing an ecommerce digital agency. For running the same functions with an in house team, the hiring order and tooling decisions differ and deserve their own guide.

Asked and answered: ecommerce digital agency

What does an ecommerce agency actually do day to day?

On a growth retainer: managing and testing campaigns, improving product and category pages, fixing feed and product data problems, building email and SMS flows, running experiments on the storefront, and reporting on the agreed numbers. The proportions depend on which stack you hired them for.

Retainer or project?

Project for anything with a finish line, such as a build, a migration or a redesign. Retainer for work that compounds, such as acquisition, testing and retention. Many stores run both: a project to fix the foundation, then a retainer to grow on top of it.

How do I compare two very different quotes?

Break both into strategy hours, delivery hours, media spend and third party software. Then ask each firm which single number they expect to move in ninety days. Two quotes that looked incomparable usually separate quickly on those answers.

Should the agency manage my ad spend from their account?

No. Spend runs through accounts you own, with your billing attached, and the agency works inside them as a user. That preserves your history, your audiences and your learning if you change suppliers.

How soon should results appear?

Paid media and feed fixes can show movement within weeks. Conversion work shows up as tests conclude. Organic search and retention programmes build over months. Agree the expected shape of each curve at the start so progress is judged against the same expectation on both sides.

What is the most common reason these relationships fail?

Undefined success. When nobody wrote down which number mattered, every monthly meeting becomes a debate about which chart to look at. Ten minutes spent agreeing the measure prevents most of it.

A place shoppers choose before your store

The constraint you diagnose is usually inside the store: traffic, conversion or order value. A growing part of the decision happens before arrival, when a shopper asks an assistant which brand to buy and gets two or three names with reasons.

Agree contribution margin as the judge, and add a visibility row for those questions. Ask an assistant for the best product in each of your core categories and record who is named and which pages are cited. AI Knows Us repeats the test across assistants and tells you which review sites and marketplaces are doing the persuading.