
Strategy
Part of A store conversion plan built on a defined journey and predeclared stop rules
Ecommerce conversion channels: one shared record for web or seller app, marketplace and assisted routes
A shared record for comparing web or seller app, marketplace and assisted conversion routes, with published UK fees, order states, payment evidence and limits.
What to take away
- Compare web, app, marketplace and assisted routes against one seller, product, customer population and journey.
- Fill one shared record before ranking anythingseller roles, order states, price, payment and exit.
- Price each route from published UK figures, such as Shopify Basic at £25 a month and eBay final value fees of roughly 10% to 12.8%.
- Define one observed rate for every routeeligible completions divided by eligible starts, same period and timezone.
- No winner is declared, because product, staffing and the gates you pass decide the fit.
The shared record every route must fill
A comparison of these four routes works only when each fills the same fields. The record is a method, not a scoreboard.
Website or seller app
- Seller and contract
- Merchant is the seller
- Order states
- Session, basket, checkout, capture, refund
- Price evidence
- Plan, hosting and payment fees
- Payment
- Merchant account via Stripe or PayPal
- Evidence
- Own analytics and server logs
Online marketplace
- Seller and contract
- Platform sells or intermediates
- Order states
- Platform order and dispute states
- Price evidence
- Referral or final value fee
- Payment
- Platform collects, then pays out
- Evidence
- Platform reports and contract terms
Assisted route
- Seller and contract
- Merchant, agent enters order
- Order states
- Enquiry, quote, confirmation, payment
- Price evidence
- Agent seat and messaging cost
- Payment
- Payment link, invoice or terminal
- Evidence
- Transcript, CRM record, consent log
Enter the legal seller and payment provider for each route. Record order states from browse through refund where they exist, since that costing guide prices one decision from sourced inputs once the state list is complete.
Published UK fees belong in the price field. Shopify Basic costs £25 a month, and Amazon Professional costs £25 a month plus referral fees of roughly 7% to 15% by category.
Channel fees at a glance
Shopify Basic
- Monthly fee
- £25
- Transaction fee
- varies
- Extra fee
- hosting separate
Amazon Pro
- Monthly fee
- £25
- Transaction fee
- 7% to 15%
- Extra fee
- referral fees
eBay basic
- Monthly fee
- none
- Transaction fee
- 10% to 12.8%
- Extra fee
- £0.30
WhatsApp Business
- Monthly fee
- free
- Transaction fee
- free
- Extra fee
- none
What each route costs and who carries the risk
eBay charges no monthly store fee on its basic plan, with final value fees of roughly 10% to 12.8% plus £0.30. Etsy charges a 6.5% transaction fee, and TikTok Shop UK referral fees run at about 5% to 9%. WhatsApp Business is free to use.
On a direct website, WooCommerce is a free WordPress plugin with hosting typically £10 to £30 a month. Stripe charges UK cards about 1.5% plus 20p, and PayPal about 2.9% plus 30p.
Direct website payment costs
- 1.5% + 20pStripe UK card fee
- 2.9% + 30pPayPal UK fee
- £25/monthShopify Basic
- £10-£30/moWooCommerce hosting
The ICO's storage and access technology guidance applies if you plan tags across these routes, because the review is purpose-specific.
Assisted routes add a seat cost, with Zendesk Suite Team at about $55 per agent a month and Shopify Inbox free on Shopify. The government's online selling guidance lists the business, product, price and delivery information a relevant seller must give.
Example: one shared record for a £40 order
- Fix the seller, product, customer population and England rules for a £40 sale.
- Enter the seller role per routemerchant on the website, platform on the marketplace, agent on the assisted order.
- Log order states from browse to refund, and note which fields the platform will not release.
- Put the published fee in the price field and state the VAT basis.
- Compare gross order value, settled cash, refunds, variable cost and net contribution on one basis.
Treat the routes as delivery models before you commit: conversion delivery models compared sets an in-house baseline beside five alternatives, which is the same discipline this record demands.
Which route fits which seller
- [Direct websitefor](/ecommerce-conversion-strategy-and-planning-guide-for-2027) sellers needing control of pricing, bundles and first-party data who can staff maintenance.
- Online marketplacefor sellers with no audience who want demand before building one, accepting platform rules and slower payout.
- Seller appfor repeat buyers who track orders, not first-time discovery.
- Assisted routefor high-consideration purchases and buyers that self-service fails.
Money and evidence: the limits of one rate
Use common fields for gross order value, settled cash, refunds, disputes and net contribution. State the VAT basis and exclude anything that cannot be normalised.
The FCA's guidance on using payment service providers explains that protections differ by provider. Check chargeback fees and settlement timing for each route.
Apply one observed-rate definition across all four channels, because that attribution comparison shows how identity logic, duplicates and late events change the count.
Common questions
Which route should an England-based seller choose?
There is no single winner. A marketplace brings demand fastest and a website brings control, while an app suits repeat buyers and an assisted route covers buyers self-service fails.
What should be recorded before comparing routes?
The legal seller and payment provider, the platform role, and the order states each route exposes. Add price evidence, payment route, fulfilment and refund owners.
How is conversion rate defined across routes?
Eligible completions divided by eligible starts, for one period and one timezone. State how duplicates, bots, staff activity and missing records are handled.
When can a route be chosen?
Once the consumer, privacy, accessibility, payment and tax gates pass independently. A better observed rate justifies investigation, not a switch.



