Ecommerce Web Development Services

Updated September 2026 · Written and maintained by the Progression Agency strategy team

An online store is an operations problem wearing a website costume. Shipping, tax, stock, returns and payments decide more about whether it works than any platform choice does. This covers the build sequence that keeps projects on schedule, where revenue is genuinely lost, the compliance you cannot skip, and how to choose a platform without buying complexity you will pay for indefinitely.

The mistake almost everyone makes first

Ecommerce projects begin with a platform decision. Which system, which theme, which developer. Six months later the difficulty turns out to be that shipping costs vary by weight and destination in a way the setup cannot express, that stock is tracked in two places that disagree, and that nobody decided who handles returns.

None of that is a website problem. It is operations, and it decides everything the website then has to do. Settling it first makes the platform decision straightforward — and deliberately boring, which is what it should be.

What an ecommerce build really is — Teams choose a platform in week one and discover in month six that the hard parts were tax, stock and returns the whole time.

What this page assumes

That you are building or rebuilding a real store selling real products, rather than evaluating software in the abstract. The sequence below is ordered by what blocks what, not by what is most interesting to decide.

Settle operations before anything else

Are you ready to build a store? — Every ‘required’ item here blocks launch. The one nobody has decided is usually returns, and it is the one customers ask about first.

Shipping: cost and timing per product, per destination, including the awkward items
Returns: who pays, who handles them, what the window is, where it is published
Tax: where you have obligations, and how they will be calculated at checkout
Stock: one system of record, and how the site learns about changes
Fulfillment: who picks and packs, and what their process needs from the site
Payments: which methods your buyers actually expect to use
Product data: descriptions, dimensions, materials, variants, what is included
Photography: what exists, what has to be shot, and by when

Returns is the one nobody has decided

It is also the first thing a first-time customer looks for. An unpublished or vague returns policy suppresses conversion on every product page at once, and no amount of design compensates for it. Decide it, write it, and link it from everywhere.

Product content is the critical path

On almost every ecommerce project, the schedule is determined by how quickly product descriptions and photography get done — and it is always slower than planned. Starting it in week eight is why launches slip by a quarter. Start it in week three, in parallel with everything else.

Shipping — decide it first. Cost and timing per product, not 'we will work it out'.

Returns — the forgotten one. Customers ask before they buy, not after.

Tax — genuinely complex. Obligations can exist in states you have never visited.

Stock — one source of truth. Two systems disagreeing is the classic failure.

Photography — the critical path. Always slower than anyone plans for.

Descriptions — not the manufacturer's. Duplicated copy ranks for nothing.

Where the money is actually lost

Where ecommerce revenue is actually lost — Nothing on this list is a platform limitation. All of it is build and operations decisions.

Checkout is the largest single line

Friction at checkout costs more revenue than any other factor, and it is the most fixable. Forced account creation, too many fields, unclear costs, vague error messages and unexpected shipping charges each remove a share of people who had already decided to buy.

Speed is a revenue line, not a technical metric

Slow pages, particularly on mobile and particularly on category pages, cost sales continuously and invisibly. Nobody complains; they simply leave. Because the loss is spread across every session, it rarely gets attributed correctly.

None of this is a platform limitation

Every item on that chart is a build and operations decision. Teams that blame the platform for conversion problems usually migrate, rebuild the same decisions on new software and discover the numbers barely move.

Build the checkout first

The homepage is where design attention goes and the checkout is where the money is. Building in that order is backwards, and it is the norm.

Offer guest checkout — forced account creation costs real orders
Show shipping cost and delivery timing before the final step, never as a surprise
Ask only for what you need to take payment and ship the order
Support the payment methods your buyers actually use, including wallets
Write specific error messages; ‘something went wrong’ loses the sale
Keep the cart visible and editable without leaving the flow
Make the return and delivery policies reachable from within checkout
Test on a real phone, on mobile data, with a real card

Surprise costs are the classic killer

A shipping charge that appears only at the final step is among the most reliable ways to end a session. Showing it early costs some sessions that were never going to convert and saves many that would otherwise be lost at the last moment.

Every field has a price

Each additional input removes a percentage of people. Company name, second address line, phone number and ‘how did you hear about us’ are all worth asking whether you genuinely need them before the order, or whether they can wait until afterwards.

Checkout — build it first. Where the money is won and lost.

Guest checkout — always offer it. Forced accounts cost real orders.

Costs upfront — no surprises. Surprise shipping ends sessions.

Fewer fields — every one costs. Ask only for what you need to ship.

Payment options — meet expectations. Cards, wallets, and what your buyers use.

Errors — be specific. 'Something went wrong' loses the sale.

The build sequence

The order that keeps an ecommerce build on schedule — Choosing a platform first is the industry norm and the most reliable way to discover in month six that it does not handle your tax or shipping reality.

A realistic ecommerce timeline — Product content is the critical path on almost every project. Starting it in week eight is why launches slip by a quarter.

Platform selection belongs in the middle

Once operations are settled and the catalog is structured, the requirements are concrete enough that platform choice becomes an obvious answer rather than a debate. Choosing first means choosing on marketing material and discovering the constraints later.

Ten to sixteen weeks, honestly

For a genuine store rather than a template with sample products. Projects promising four weeks are either very small catalogs, or they are deferring the operations work that will surface immediately after launch.

Choosing a platform without over-buying

Four ways to build a store — Headless is genuinely more flexible and genuinely more expensive to keep. Most stores that choose it did not need it.

Which build fits which business — The expensive corner is correct for a small number of businesses and chosen by many more than that.

What each option commits you to
Option	Suits	Ongoing reality	The catch
Hosted platform	Most stores, most of the time	Monthly fee plus transaction share	Ceiling on unusual requirements
WordPress with commerce	Content-led stores with moderate catalogs	Hosting plus real maintenance	Plugin sprawl becomes the risk
Headless	Complex catalogs and multiple channels	Continuous developer involvement	A developer for routine changes
Custom platform	Businesses whose commerce is the product	Substantial and permanent	You now maintain software
Marketplace only	Testing demand, or very small ranges	Commission on every sale	No customer relationship and no data
Marketplace plus own store	Many growing businesses	Two systems to reconcile	Stock accuracy across both

The expensive corner is right for a few and chosen by many

Headless and custom builds are genuinely more capable. They also require continuous engineering to keep running and to change anything. If nobody on your team can update a category page without a deployment, that cost recurs forever.

Marketplaces are a demand test, not a business

Selling through a marketplace is a reasonable way to find out whether people want the product. It does not give you the customer relationship, the data or the margin, which is why it works better as a first step than as a destination.

Product pages that reduce returns

Return rates are largely a content problem. Customers who received something different from what they expected were told something inadequate before they bought.

Give dimensions, weight, materials and what is included, specifically
Show multiple photographs including the product in use and at scale
Write your own descriptions rather than reusing the manufacturer’s
State stock status accurately, and never sell what you cannot ship
Put the delivery estimate on the product page, not three steps later
Link the returns policy from the product page itself
Use reviews, which do heavy lifting for unfamiliar brands
Answer the questions your support inbox already receives most often

Manufacturer copy ranks for nothing

If a hundred retailers publish identical supplier descriptions, none of them has anything distinctive to rank with. Writing your own is one of the few ecommerce content tasks with both a search benefit and a conversion benefit.

Track return rate per product

A product returning far above your average is telling you something specific about its listing — usually sizing, scale or an omitted detail. That is a cheap fix with an immediate margin effect, and almost nobody looks at the data this way.

Product detail — reduces returns. Sizes, materials, dimensions, what is included.

Real photos — beat stock. Multiple angles, in use, and honest.

Reviews — do heavy lifting. Especially for unfamiliar brands.

Stock status — must be accurate. Selling the unavailable destroys trust.

Delivery estimate — on the product page. Not a surprise three steps later.

Returns policy — link it everywhere. Reassurance, not fine print.

Speed, and why it belongs in the revenue conversation

Category pages matter most; they carry the browsing and accumulate the weight
Images are usually the largest problem — correctly sized, modern formats, lazy loaded
Audit third-party scripts; every tracking tag costs something
Test on mobile data rather than office connectivity
Measure the metrics that reflect real user experience, not a synthetic score
Make on-site search work well, because internal searchers convert far above average
Check that filters actually function, since broken filtering silently loses revenue
Re-check speed quarterly, because it decays as content and tags accumulate

On-site search is underrated

Visitors who use a site’s search box convert at a substantially higher rate than those who do not, because they arrived knowing what they wanted. Poor search results turn your highest intent visitors away, and the search query log is also the cheapest product research available.

Speed — is revenue. Especially on mobile, especially on category pages.

Images — the usual culprit. Correctly sized and modern formats.

Scripts — audit them. Every tracking tag has a cost.

Category pages — the priority. Where browsing happens and weight accumulates.

Search on site — underrated. Internal searchers convert far above average.

Filters — make them work. Broken filtering is silent lost revenue.

Compliance you cannot skip

This section is general information rather than legal or tax advice, and the tax question specifically warrants a professional.

What applies to most US online stores
Area	What it means	Where to start
Sales tax	Obligations can exist in states you have no physical presence in	Check economic nexus thresholds per state
Payment security	Card data must be handled to PCI standards	Use a compliant processor; never store card data
Accessibility	Ecommerce is heavily litigated in this area	Build to WCAG rather than retrofitting
Privacy and tracking	Disclosure and consent for analytics and advertising	A truthful privacy policy and a real consent mechanism
Terms and returns	Must be clear, findable and actually followed	Publish them and link from product pages and checkout
Product claims	Advertising claims must be substantiated	FTC advertising guidance
Age or license restrictions	Some categories carry extra obligations	Check before launch, not after

Sales tax is the one that surprises people

Following the Supreme Court’s decision in South Dakota v. Wayfair, states may require collection from sellers with no physical presence, based on sales thresholds that differ by state. A store can therefore acquire obligations in places it has never been, simply by growing. Get advice specific to your situation — this is not it.

Accessibility is a live commercial risk

Online retail sees a disproportionate share of accessibility litigation. Building to recognized standards during the project is far cheaper than retrofitting, and it also improves usability for everyone — the same work that makes a checkout usable with a keyboard makes it faster for everybody.

Sales tax — registration matters. Thresholds vary by state and change.

PCI — use a compliant processor. Do not touch card data yourself.

Accessibility — a real legal exposure. Ecommerce is heavily litigated.

Privacy — disclose tracking honestly. Consent handled properly.

Terms — and returns in writing. Clear, findable, and followed.

Get advice — for tax specifically. This page is not tax advice.

What an ecommerce build costs

Where the money goes on a store build
Component	Share of budget	Why it costs what it does	What cutting it produces
Operations and requirements	A tenth	Shipping, tax, stock and returns worked out properly	Rebuild in month six
Catalog structure and data	A fifth	Variants, attributes and clean product data	A store nobody can navigate
Product content and photography	A quarter	The slowest and most underestimated line	High returns and nothing to rank with
Design and templates	A fifth	Category, product and checkout templates	Generic, and it shows
Build and integrations	A fifth	Payments, tax, shipping, stock, fulfillment	Manual work forever after launch
Testing and accessibility	Under a tenth	Real orders, refunds, devices, standards	Support load and legal exposure
Launch and monitoring	A small slice	Redirects, monitoring, fast fixes	Lost rankings on a replatform

Content is a quarter of the build, and it is not optional

Businesses routinely budget for design and development and treat product content as something that will happen alongside. It is a quarter of the real work, it determines both your return rate and your search visibility, and it is the reason schedules slip.

Integrations are where scope quietly grows

Each connection — tax, shipping rates, stock, accounting, fulfillment — is straightforward alone and compounding together. Listing every system that has to know about an order, at the start, is what keeps this line honest.

Testing before launch

Place a real order with a real card, all the way through
Process a refund on that order and confirm both sides look right
Walk the full return flow as a customer would
Test a failed payment, because that path is used more than people expect
Confirm order confirmation emails arrive and are not filed as spam
Check that stock decremented correctly and in the right system
Test on real devices on mobile data, including an older phone
Verify tax is calculated correctly for several destinations

Test the unhappy paths

Successful orders get tested. Refunds, failed payments, out-of-stock races and address validation errors usually do not, and those are exactly the paths that generate support load and lost trust in the first week.

Real orders — test with them. Including refunds and failures.

Failed payment — test that path. It happens more than you think.

Return flow — walk it yourself. End to end, as a customer.

Email receipts — check they arrive. And that they are not spam-filed.

Mobile — test on real devices. Not a resized browser.

Then monitor — daily for two weeks. Early problems are cheap problems.

Measuring a store honestly

What to watch after launch
Metric	What it tells you	Common misreading
Conversion rate	Whether the site turns visits into orders	Reported as one number instead of split by device
Average order value	Whether people buy more per visit	Ignored in favor of chasing traffic
Cart abandonment	Where checkout hurts	Treated as a target rather than a diagnostic
Return rate by product	Whether listings set accurate expectations	Averaged, hiding the specific problem products
Repeat purchase rate	Whether the business compounds	Neglected while acquisition is optimized
Contribution after costs	Whether any of it is profitable	Revenue reported as though it were margin
On-site search terms	What people want that you may not stock	Rarely looked at, and free

Split conversion rate by device or it will mislead you

Mobile and desktop conversion rates usually differ substantially. A blended figure moves whenever the traffic mix moves, which produces confident conclusions about changes that never happened.

Contribution is the number that matters

Revenue after advertising, platform fees, payment fees, shipping and returns. It is the only figure that answers whether the store is a business, and it is routinely absent from ecommerce reporting that leads with revenue growth.

Conversion rate — the headline. But segment it by device.

Average order value — the other lever. Often easier to move than traffic.

Abandonment — diagnostic. Tells you where checkout hurts.

Return rate — by product. High returns mean bad product information.

Repeat rate — the durable one. Cheaper than any acquisition.

Contribution — after everything. Revenue minus ads, fees, shipping, returns.

Do you still need a website in 2026? — Google Search Central. On owning the store rather than renting an audience.

SEO for small businesses — Google Search Central. How product and category structure affects search.

Analyzing performance on Google Search — Google Search Central. Reading real performance data after launch.

Questions about ecommerce development

Building or rebuilding a store?

We start with shipping, tax, stock and returns, structure the catalog before anything is designed, build the checkout first — and report contribution, not revenue.

Talk to us about ecommerce

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Frequently asked questions

What is the most common mistake in an ecommerce project?

Choosing a platform first. Shipping, tax, stock and returns decide what the website has to do, and settling those first makes the platform decision straightforward. Choosing first means discovering the constraints in month six.

What blocks launch most often?

Product content and photography. It is the critical path on nearly every project and always slower than planned. Starting it in week eight is why launches slip by a quarter.

Where is ecommerce revenue actually lost?

Checkout friction is the largest single line, followed by slow pages on mobile and unclear shipping costs. None of it is a platform limitation — it is all build and operations decisions.

Should I offer guest checkout?

Yes. Forced account creation costs real orders from people who had already decided to buy. You can invite them to create an account after the purchase instead.

Why does surprise shipping cost matter so much?

A charge that appears only at the final step is among the most reliable ways to end a session. Showing it early loses a few sessions that would never have converted and saves many that would otherwise be abandoned at the last moment.

How many checkout fields should I ask for?

Only what you need to take payment and ship the order. Every additional input removes a percentage of buyers. Company name, second address lines and ‘how did you hear about us’ can usually wait until after the order.

Which platform should I choose?

For most stores, a hosted platform. Headless and custom builds are genuinely more capable and require continuous engineering to change anything. If nobody on your team can update a category page without a deployment, that cost recurs forever.

Is selling on a marketplace enough?

It is a reasonable way to test whether people want the product. It does not give you the customer relationship, the data or the margin, which is why it works better as a first step than as a destination.

How long does an ecommerce build take?

Ten to sixteen weeks for a genuine store. Projects promising four weeks either have very small catalogs or are deferring operations work that will surface immediately after launch.

Why are my return rates high?

Usually a content problem. Customers who received something different from what they expected were told something inadequate beforehand — most often about size, scale or what is included. Track return rate per product and the problem listings identify themselves.

Should I write my own product descriptions?

Yes. If a hundred retailers publish identical supplier copy, none of them has anything distinctive to rank with. It is one of the few ecommerce content tasks with both a search benefit and a conversion benefit.

Does site speed really affect sales?

Substantially, and invisibly. Nobody complains about a slow store, they simply leave. Because the loss is spread across every session it is rarely attributed correctly, which is why it keeps being treated as a technical rather than a revenue issue.

Which pages should I optimize for speed first?

Category pages. They carry most of the browsing and accumulate the most weight, usually through images and third-party scripts.

Is on-site search worth investing in?

Yes. Visitors who use the search box convert at a substantially higher rate because they arrived knowing what they wanted. The search query log is also the cheapest product research available.

What are my sales tax obligations?

Potentially in states where you have no physical presence. Following South Dakota v. Wayfair, states may require collection based on sales thresholds that differ by state, so a growing store can acquire obligations in places it has never been. Get advice specific to your situation — this page is not tax advice.

Do I need to worry about PCI compliance?

Use a compliant payment processor and never store card data yourself. That handles the majority of the obligation for most small and mid-sized stores.

Is accessibility a real risk for online stores?

Yes — online retail sees a disproportionate share of accessibility litigation. Building to recognized standards during the project is far cheaper than retrofitting, and the same work makes checkout faster for everyone.

What should I test before launch?

A real order with a real card, a refund on it, the full return flow, a failed payment, confirmation emails arriving, stock decrementing in the right system, tax calculating for several destinations, and everything on a real phone.

Which test do people skip?

The unhappy paths. Successful orders always get tested; refunds, failed payments, out-of-stock races and address errors usually do not — and those generate the support load in week one.

What metrics should I watch after launch?

Conversion rate split by device, average order value, cart abandonment as a diagnostic, return rate per product, repeat purchase rate, and contribution after all costs.

Why split conversion rate by device?

Because mobile and desktop rates usually differ substantially, so a blended figure moves whenever the traffic mix moves. That produces confident conclusions about changes that never happened.

What is the single most important number?

Contribution — revenue after advertising, platform fees, payment fees, shipping and returns. It is the only figure that answers whether the store is a business, and it is routinely missing from reporting that leads with revenue growth.

Sources and further reading

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.