A shopper finds the right product, checks delivery, starts paying and then discovers a cost they did not expect. Another customer completes the purchase but cannot get a useful answer about a delayed order. These hypothetical situations show why customer experience extends beyond website design.
For an ecommerce business evaluating a customer experience agency in London, the useful question is specific: can this partner identify a customer problem, help fix it and establish whether the change worked?
A credible proposal should connect customer research, website usability, support and post-purchase operations. It should also explain responsibilities, measurement and costs. This guide sets out a practical way to assess that proposal and build a focused improvement programme.
What should an ecommerce customer experience agency do?
Customer experience, or CX, covers how people experience your business across discovery, purchase, delivery, support and subsequent orders. Customer support is one part of that journey. Conversion rate optimisation focuses more narrowly on helping visitors complete defined actions, such as a purchase.
A CX engagement may involve both, but its scope needs to be explicit. An attractive redesign cannot resolve inaccurate stock information without changes to the systems and teams supplying it.
Ask for tangible deliverables:
- A diagnosis supported by customer feedback and behavioural evidence.
- A journey map identifying failures, dependencies and accountable teams.
- A prioritised backlog explaining what to change and why.
- An implementation plan covering design, development, content and operations.
- A measurement plan with baseline definitions and decision criteria.
Clarify whether the agency will deliver changes or advise your internal team. That distinction affects staffing, budgets and what a completed project actually means.
When does choosing a London agency matter?
A London partner may suit a business that needs in-person workshops, store visits or direct collaboration with a local team. Check who will attend, where they work and whether those activities are included in the quote.
Location alone does not demonstrate ecommerce expertise. A remote specialist may be a better fit for your platform, product category or international operations. Evaluate working-hour overlap, access to decision-makers and implementation capability alongside proximity.
For a London retailer selling overseas, ask the agency to examine destination-specific delivery promises, currencies, returns routes and support coverage. A journey designed around UK customers may leave important questions unanswered elsewhere. Include the relevant countries and languages in the research brief rather than assuming one experience serves every market.
Find the problem before buying the solution
Combine customer accounts with operational evidence
Start with recent support conversations, return reasons, product reviews, customer interviews and checkout data. Each reveals a different part of the experience. A drop-off identifies where to investigate; it rarely establishes the cause on its own.
For example, payment-stage abandonment could involve failed transactions, delivery charges, unavailable payment methods or a shopper comparing options. Ask what additional evidence would distinguish those explanations before approving a redesign.
Compare relevant segments: mobile and desktop, first-time and returning customers, domestic and international orders. Record missing data and consent-related measurement gaps. A dashboard should make those limitations visible.
Turn the journey map into an action record
For each problem, document the customer task, observed difficulty, supporting evidence, responsible owner, proposed change and success measure. Include dependencies such as carrier data or payment-provider restrictions.
A map becomes useful when someone can act on it. Assign a decision-maker to each priority and agree how unresolved issues move between marketing, development, fulfilment and support.
Prioritise changes with a worked example
Consider an illustrative clothing retailer choosing among three projects: an animated homepage, clearer delivery information and improved sizing guidance. This is a hypothetical planning exercise, not a client case study.
If support records contain repeated delivery questions and interviews reveal uncertainty before checkout, delivery information has a clearer evidence base than an animation. Frequent size-related returns could justify a separate sizing investigation.
Compare each project using four questions:
- Customer impact: Does the issue block a purchase, delay help or create avoidable effort?
- Evidence: Is it supported by several observations or only an internal opinion?
- Feasibility: Can the team implement it with reliable underlying information?
- Evaluation: Can the team assess the intended outcome without hiding adverse effects?
The delivery project might need product-page content, postcode eligibility logic and fulfilment approval. Measure delivery-related enquiries alongside checkout completion and contribution margin. Clearer promises are only useful if operations can meet them.
Use this method to explain priorities, not manufacture precise forecasts. Record assumptions and revisit them as evidence improves.
Improve the journey across purchase, support and retention
Make product discovery and checkout easier to complete
Review whether customers can find relevant products, understand variants and see important restrictions before paying. Test search, filters, mobile forms, validation messages and payment failures using realistic tasks.
Include accessibility in acceptance criteria. W3C’s Web Content Accessibility Guidelines provide a recognised framework. Agree the version and conformance level, then combine automated checks with manual evaluation of keyboard use, focus, form labels and error recovery. An automated score alone does not establish conformance.
Give support staff enough context to resolve issues
Define the information an agent needs: order details, previous contact, delivery status and actions already attempted. Limit access according to role and avoid unnecessary exposure of personal information.
If automation is proposed, identify permitted tasks and escalation conditions. A system answering delivery questions needs dependable order information. Refund exceptions, disputed charges and uncertain answers need a clear route to authorised people.
Assess repeat contacts and reopened cases alongside response time. A quick reply that leaves the customer repeating their explanation is an incomplete service outcome.
Make post-purchase communication useful
Separate necessary order information from promotional activity. Confirm what customers need to know about dispatch, delays, product use and returns, then assign ownership of each message.
For UK distance sales, GOV.UK guidance explains that customers generally have 14 days after receiving an item to notify cancellation and another 14 days to return it. Exceptions apply to certain products. Check the rules relevant to your goods and customer markets before changing return policies or automated responses.
Build retention activity around useful reasons to return. Product-care guidance or a relevant replenishment reminder may suit some categories; frequent discount messages may not. Check margin and complaints as well as repeat orders.
Set boundaries for personalisation and customer data
Personalisation should address a defined customer need, such as remembering a preference the shopper deliberately supplied. Ask what information is required, how it stays accurate and what happens when the system lacks reliable data.
UK tracking requirements need particular care in 2026. The ICO’s current guidance describes conditional exceptions to consent requirements, including certain statistical uses intended to improve a service. The statistical exception requires clear information and a simple, free way to object; it does not broadly authorise individual tracking or advertising. UK GDPR obligations may also apply.
Require a technology-by-technology assessment of purposes, data access and applicable conditions. Do not assume an analytics label makes a tool exempt. Keep advertising, profiling and session-recording proposals within that review.
Measure customer outcomes and commercial effects together
Agree definitions before implementation. The following are suggested working measures, not universal benchmarks:
- Purchase conversion: Sessions containing a completed purchase divided by eligible sessions. Specify the session rules and treatment of missing tracking.
- Repeat purchase: Customers in a defined first-order cohort who buy again within the chosen window, divided by that cohort. Allow every customer the same observation period.
- Contact rate: Order-related support cases divided by orders for a defined period. Set rules for duplicate cases and delayed enquiries.
- Customer satisfaction: Positive responses divided by valid responses to a specified survey. Record timing, response volume and possible response bias.
- Commercial guardrails: Contribution margin, returns, cancellations and support workload, with agreed cost definitions.
NPS can complement this view, but a recommendation score does not explain which journey step failed. Pair survey results with comments and observed behaviour.
A before-and-after comparison can be affected by seasonality, promotions, stock availability and changes in traffic. Use a properly designed controlled experiment when feasible. Otherwise, document competing explanations and present the result as directional evidence. Low order volume may require a longer observation period; qualitative research can still reveal usability problems while data accumulates.
Compare agency proposals on evidence and responsibility
A practical scorecard can rate each criterion from zero to three: absent, asserted, demonstrated or contractually specified. This is a suggested procurement aid, not a validated industry standard.
- Relevant work: Comparable platform and operational complexity, with the agency’s contribution explained.
- Research quality: Recruitment approach, evidence sources and handling of uncertainty.
- Delivery ownership: Named roles for implementation, testing, approval and maintenance.
- Measurement: Baselines, experiment decisions and commercial guardrails.
- Governance: Access controls, data handling, accessibility and escalation procedures.
- Commercial clarity: Deliverables, exclusions, third-party costs and change-control terms.
Set minimum requirements for critical areas rather than allowing an impressive portfolio to compensate for missing data safeguards. For reported results, ask about the baseline, period, other simultaneous changes and permission to discuss the work.
Compare discovery, implementation and ongoing evaluation costs separately. No reliable London fee benchmark is established here. Request itemised quotes against the same brief, including software, development dependencies and handover.
Frequently asked questions
What if our ecommerce platform is being replaced?
Make migration plans part of the brief. Separate urgent fixes from work that would be discarded during migration, and preserve agreed measurement definitions across the transition.
Can an agency work without access to customer-level records?
Some discovery can use aggregated reports, redacted examples and approved interviews. Establish what each task requires before granting access; missing information may limit the conclusions.
Who should own the research recordings and design files?
Specify ownership, permitted reuse, retention and export arrangements in the contract. Address participant permissions and third-party licensing separately.
How should agencies handle multilingual research?
Agree recruitment markets and language capability. Check who moderates interviews and interprets findings, and whether translated materials will be reviewed by suitable speakers.
What happens if a test produces no clear result?
Require a documented decision: gather more evidence, revise the hypothesis or stop the change. An inconclusive finding should not be reported as a demonstrated improvement.
What should happen when the engagement ends?
Request an actionable backlog, measurement definitions, implementation documentation and an access review. Assign an internal owner to maintain the changes and monitor unresolved issues.
Build the brief around a customer problem
Before appointing a partner, write a one-page brief covering the affected journey, available evidence, priority markets, internal owners and outcome to evaluate. Use it to compare how agencies investigate and deliver.
SEO Jetty’s official website lists customer data integration, personalisation and automated support among its advertised offerings. If those capabilities match your brief, ask SEO Jetty to confirm the relevant scope and delivery responsibilities.
Sustainable ecommerce growth is an objective to test through customer and business outcomes. Begin with a specific difficulty, make an accountable change and use the evidence to decide the next step.
