How to Vet an Ecommerce Development Agency in 2026 (Scorecard, RFP, and Red Flags)

Vetting an ecommerce development agency is procurement, not marketing. The agency you pick decides whether your store ships on time, scales through Black Friday without falling over, and survives the next platform upgrade without a rebuild. This guide gives you a 12-point scorecard, an RFP outline, the contract clauses that cause disputes, and the seven questions that flush out weak agencies in the first call.

Key Takeaways

• Shopify merchants alone cleared more than $100 billion in GMV in a single quarter, so the agency you hire is operating in a market where downtime and bad UX have direct revenue cost (Shopify Q1 2026 results).
• The average documented online cart abandonment rate is 70.22%, which makes checkout and UX competence non-negotiable evaluation criteria (Baymard Institute, 2026).
• Standish Group’s CHAOS research shows large software projects fail or run over far more often than small ones, so the most important question on any RFP is how the agency breaks the build into shippable phases (Standish Group CHAOS).

This post is about choosing the agency. For budget sizing, see ecommerce website development costs. For platform comparison (Shopify vs Magento vs WooCommerce), see the comprehensive guide to ecommerce website development. For sole-trader and sub-£15k projects, the small business website development guide covers the lower end of the market.

Why does agency selection matter more than platform choice?

Shopify merchants cleared more than $100 billion in GMV in Q1 2026 alone, with the company posting 34% year-over-year revenue growth (Shopify Q1 2026 results). A competent build on a mid-tier platform out-earns a botched build on the most expensive one. The agency is the variable that compounds.

The decisions an agency makes in week one (information architecture, checkout flow, payment provider, schema) carry forward for years. Replatforming typically costs 1.5 to 3 times the original build, so a bad first decision is paid for twice.

What does a failed ecommerce project look like?

A failed build doesn’t usually crash. It limps. The site ships six months late, two of three integrations cut, checkout bleeding conversion, and an agency that wants a six-figure change order to finish. Phase the work, ship narrow, and judge agencies on how willingly they break a build down instead of selling a single all-in scope.

What should be on your agency evaluation scorecard?

Standish Group’s CHAOS research has tracked software project outcomes since 1994 and consistently shows large IT projects fail or overrun far more often than small ones (Standish Group). Run every shortlisted agency through the same scorecard, with the same weights, before contracting. Score 1 to 5 per row; agencies totalling under 70/100 should not make the cut.

Criterion	Weight	What “5” looks like	What “1” looks like
Platform depth (Shopify Plus, Adobe Commerce, BigCommerce, WooCommerce)	5	Certified partner with named engineers and 5+ live stores on your platform	Generalist agency that “also does Shopify”
Relevant case studies (same industry, similar GMV)	5	3+ live stores in your vertical with measurable outcomes	Mostly brochure sites and one ecommerce demo
Live store access	4	Shares URLs, lets you stress-test checkout	Screenshots only, no live links
Discovery process	4	Paid discovery with written deliverable before build quote	Quotes a full build off a 30-minute call
Phased delivery plan	4	Breaks build into 2 to 4 shippable milestones	Single fixed-price quote for everything
Code ownership and exit terms	4	You own code, can move agencies without rebuild	Proprietary framework or licence locks
Performance and Core Web Vitals track record	3	Shows real LCP/INP numbers from live sites	“It depends on the theme”
Accessibility (WCAG 2.2 AA) capability	3	Has run audits, can show conformance reports	Doesn’t raise accessibility at all
SEO transfer and migration plan	3	Documented redirect, schema, and crawl plan	Treats SEO as the client’s problem
Integration experience (ERP, PIM, payments)	3	Built named integrations like NetSuite, Klaviyo, Adyen	Vague “we integrate with anything”
Contract clarity (IP, scope, change orders)	3	Sample contract supplied with quote	Contract emerges after deposit
Reference call willingness	3	Offers 2 to 3 client references for a call	Offers written testimonials only

If you’re tempted to bend a score because you like the team, write down why. That note will save you in month three.

How do you write an RFP that actually filters?

A good RFP filters out agencies that can’t read it. Keep it under 12 pages; structure it so the response itself tests the agency’s process.

Sections every ecommerce RFP needs

1. Business context. One page on your company, current revenue, channels, and why you’re rebuilding now.
2. Current state. Existing platform, traffic, conversion rate, AOV, integrations, pain points.
3. Scope and out-of-scope. What the build must include (catalog, locales, payments, integrations) and what it explicitly excludes (paid media, content writing, photography).
4. Functional requirements. Numbered list. Every line testable.
5. Non-functional requirements. Performance budget (LCP under 2.5s on a 4G connection), accessibility (WCAG 2.2 AA), uptime, peak-traffic capacity.
6. Integrations list. Name every system: ERP, OMS, PIM, marketing automation, payment gateways, tax, search, reviews.
7. Migration scope. Source platform, data volumes, historical orders, SEO redirect strategy.
8. Timeline. Soft and hard launch dates with reasoning.
9. Budget range. Include it. Agencies that quote outside your range either misread the brief or aren’t a fit.
10. Response format and evaluation criteria. Page limit, deadline, decision date, and how you’ll score the response.

Questions to include in the response

• Describe a similar project (industry, GMV, platform) and the measurable outcome.
• Walk through how you’d phase the first 12 weeks.
• What’s your discovery process and what does it cost?
• Who’s the named engineering lead and account manager, and what’s their tenure?
• Provide three client references we can call.
• Submit a sample SOW from a comparable engagement.

What are the seven questions that flush out weak agencies?

Ask every shortlisted agency these on the first or second call. Score the answers honestly.

1. “Can we talk to a client whose project went badly?”

A confident agency names one. Every agency has a project that overran or shipped late; the question is whether they can tell you what they learned. An agency that swears every project has been perfect is either lying or new.

2. “Who owns the code if we part ways?”

The answer should be “you do.” Watch for proprietary frameworks, custom CMSs, or licence terms that lock you in. If the agency uses a proprietary layer, ask for a written exit clause covering code, credentials, and data on day one of separation.

3. “Show me the staging environment for a live client.”

A real agency runs feature branches, staging, and production with CI/CD. One that pushes directly to production from a developer’s laptop will break your store on a Saturday afternoon.

4. “What’s your average Core Web Vitals score on live stores?”

LCP, INP, and CLS are public data. A good agency will have numbers ready: “Our median LCP across 30 live Shopify stores is 1.8 seconds.” A weak one deflects to “it depends on the theme.”

5. “Walk me through your checkout testing process.”

With Baymard’s documented average cart abandonment rate at 70.22% (Baymard, 2026), checkout is the most consequential surface in your store. A competent agency tests it on real devices, across browsers, with screen readers, on slow connections, in every payment method you offer. A weak agency tests in Chrome on a MacBook and ships.

6. “What does your change-order process look like?”

You want a written process: requests submitted in writing, scoped within a fixed window, costed before work starts, signed by both parties. Agencies that handle changes verbally (“we’ll add that, no problem”) send surprise invoices in month four.

7. “What’s the smallest engagement you’d take?”

This tells you where you sit in their book. At the bottom of their range, you get junior engineers and slow responses. At the top, the agency may be out of its depth. The right answer puts you in their usual mid-band.

What contract clauses cause the most disputes?

Negotiate the clauses below before signing. Refusal to move on any of them is itself a useful signal.

Clause	What weak contracts say	What you want it to say
Scope and change orders	“Scope as discussed”	Itemised spec attached as Schedule A, defined change-order process
Acceptance criteria	“Subject to client approval”	Specific testable criteria per milestone, fixed review window
IP ownership	Silent or “agency retains”	You own all custom code and assets on payment
Termination	90+ day notice, full payment	30 day notice either side, defined handover deliverables
Warranty period	None	60 to 90 days post-launch covering bugs in delivered scope
Data and credentials	Silent	All credentials, accounts, and data returned within 14 days
Liability cap	Capped at fees paid	Multiple of fees, with carve-outs for confidentiality, IP, and gross negligence

In-house vs agency vs freelancer: which model fits?

The build-team decision sits underneath the agency-selection decision.

Model	Best for	Year-1 cost	Main risk
Freelancer (1 to 2 people)	Stores under £200k GMV, founder-led	£15k to £40k	Bus factor of one
Boutique agency (5 to 25 people)	£200k to £10m GMV, single platform	£30k to £150k	Capacity bottlenecks in peak season
Mid-market agency (25 to 100 people)	£5m+ GMV, multi-locale, integrations	£100k to £500k	Senior team sold, juniors delivered
Enterprise agency (100+ people)	£20m+ GMV, replatforming	£300k+	Slow change cycles
In-house team	£10m+ GMV with a 3-year roadmap	£250k+ in salaries	Hiring time, retention risk
Hybrid (in-house lead + agency execution)	Most mid-market stores	Variable	Coordination overhead

For most mid-market merchants the honest answer is hybrid: in-house product owners who hold the roadmap, with an agency providing engineering capacity. That keeps platform knowledge in-house and prevents single-vendor risk.

What red flags should kill a shortlist immediately?

Any one of these is enough to drop an agency.

• Refuses to share live client URLs (screenshots only signals exaggerated portfolio).
• Quotes a fixed price without any discovery phase.
• Won’t name the engineer who’ll be on your project in the contract.
• No staging environment in their workflow (production becomes the staging environment).
• Proprietary CMS layer on top of your platform (lock-in by design).
• Account manager is the only person you ever speak to pre-contract.
• Pricing dramatically lower than every other bid (usually undisclosed offshore subcontracting).
• Refuses to provide three references.
• Pressure to sign before a deadline they invented.

How long should the selection process take?

For a build between £30k and £250k, four to six weeks from RFP issue to contract signature is the right window. Working timeline: RFP in week 1; issue to 4 to 6 agencies in week 2; score responses in week 3; pitch and reference calls in week 4; contract and SOW negotiation in week 5; sign and kick off in week 6. If the timeline is tighter, cut the longlist (issue to 3 agencies, not 6) rather than skipping reference calls or legal review.

Frequently asked questions

How do I check an agency’s portfolio if they only show screenshots?

Ask for live URLs. If the agency cites confidentiality, ask for one URL per case study with that client’s permission. Agencies with strong work have at least a few clients willing to be referenced publicly. If everything is screenshot-only, treat it as unverified and downgrade the portfolio score on your scorecard.

Should I pay for a paid discovery phase before signing a full contract?

Usually yes. A paid discovery phase (typically £5k to £25k) produces a written technical spec, integration plan, and timeline before either side commits to the full build. The deliverable is yours either way. Agencies that refuse paid discovery and quote a full build from a 30-minute call are transferring all the discovery risk into change orders later.

What’s the difference between a Shopify Partner and a Shopify Plus Partner?

A Shopify Partner is any agency registered with Shopify; the bar is low. A Shopify Plus Partner has been vetted, has Plus-tier engineering experience, and sits in a smaller directory. If you’re on Shopify Plus or moving to it, restrict your shortlist to Plus Partners. Adobe Commerce and BigCommerce run similar tiered programmes.

How do I avoid being locked in to one agency post-launch?

Three contract clauses agreed before signing. First, IP ownership: all custom code and assets transfer to you on payment. Second, credentials and data: returned within 14 days of any termination. Third, no proprietary framework on top of the standard platform stack (Liquid for Shopify, PHP for Magento, etc.) that another developer can’t read. With those three in place, switching agencies post-launch is a friction cost, not a rebuild.

Is it worth working with an offshore or nearshore agency to save cost?

It depends on how the agency runs. A nearshore agency in Eastern Europe or India with a senior local PM, English-fluent engineers, and overlapping working hours can match a London or New York agency for 40 to 60% less. An offshore subcontractor white-labelled through a Western front-of-house, with no direct engineer contact, almost always disappoints. The questions to ask are the same; the answers tell you whether the cost saving is real.