Honest agencies will be specific. "We built the real-time transit tracking module and the notification system. The core routing engine was built by the client's team." Vague agencies will redirect: "We can't discuss details due to NDA."
NDAs are real, but they rarely prevent discussing the type of work done. "We built a real-time data pipeline for a major transportation company" is NDA-safe and informative.
Layer 3: Test the quality, not just the appearance
Screenshots show design. They don't show engineering quality. Here's how to look deeper:
For web applications:
- Run a Lighthouse audit (Chrome DevTools → Lighthouse). Check performance, accessibility, and best practices scores.
- Test mobile responsiveness. Does it work on a phone?
- Check page load times. Anything over 3 seconds is below modern standards.
- Look at the URL structure and SEO basics. Competent agencies don't ignore SEO.
For mobile apps:
- Check crash rates in App Store / Google Play reviews
- Look at the last update date. An app not updated in 12+ months is either abandoned or technically stagnant.
- Test basic flows. Does onboarding work smoothly? Do transitions feel native?
For API/backend systems:
- Ask about uptime and reliability metrics
- Request information about how the system handles load
- Ask what monitoring and alerting is in place
Layer 4: Outcomes, not outputs
The hardest layer to evaluate, but the most telling. Did the project achieve what it was supposed to?
Questions to ask:
- "What business metrics improved after launch?"
- "Did the project ship on time and within budget?"
- "What happened after the initial launch? Were there iterations?"
- "Is the client still working with you? If not, why?"
Strong answers sound like: "The customer portal reduced support calls by 35%. The client expanded the project to three more countries." Weak answers sound like: "The client was satisfied with our work."
Request references. Not the references the agency provides (those are curated) — ask for references from projects similar to yours in scope, technology, and industry.
Layer 5: Team continuity
The portfolio shows what the agency built with a specific team at a specific time. Your project will be built by today's team.
Questions to ask:
- "Are the key people from these portfolio projects still at your company?"
- "Who would be the tech lead on my project? What's their portfolio?"
- "What's your team turnover rate?"
- "Can I interview the developers who'd work on my project?"
An agency with 25% annual turnover will have a completely different team every four years. The people who built the impressive portfolio project may have moved on. That doesn't invalidate the work, but it should factor into your evaluation.
Red flags in portfolios
| Signal | What it might mean |
|---|
| All projects are from 5+ years ago | The agency may have talent or capability gaps today |
| No live projects available to test | The work may not have survived production use |
| Vague descriptions of the agency's role | The agency's contribution may have been limited |
| No before/after metrics | The agency may not track outcomes |
| Every project is a different technology stack | The agency may lack deep expertise in any one area |
| Reluctance to provide references | Previous clients may not have had good experiences |
| Portfolio only shows design, no technical depth | The agency may outsource development |
Green flags in portfolios
| Signal | What it means |
|---|
| Live, actively maintained products | The work was good enough to survive production |
| Specific metrics and outcomes cited | The agency measures and delivers results |
| Long-term client relationships | Clients found the work good enough to continue |
| Willing to explain what they did vs. what others built | Honest about their role and contribution |
| Team members you'd work with have portfolio history | Your project gets experienced people, not juniors |
| Technology consistency across projects | Deep expertise, not a jack-of-all-trades |
The 30-minute portfolio evaluation checklist
If you're short on time, here's a fast evaluation you can run in a single call with the agency:
- Pick 2-3 projects from their portfolio that are most similar to yours
- For each: "What exactly did your team build, and what was the team size?"
- For each: "Is this live? Can I use it or see a demo?"
- For each: "What business outcomes did the client achieve?"
- "Can I talk to the client from one of these projects?"
- "Who from these projects would work on mine?"
If the agency answers all six with specifics, they're worth a deeper evaluation. If they deflect on more than two, keep looking.
Beyond the portfolio: signs of a strategic partner
A portfolio shows what an agency has built. How they engage with your project shows whether they'll act as a partner or as a vendor that turns specs into tickets. Watch for three signals:
They challenge your requirements. A good partner tells you when a feature will cost $200K and serve 3% of your users, then proposes a simpler version that covers most of the need.
They learn your domain. A team building a healthcare system should understand clinical workflows. A team building fintech should know the securities rules. Expect the partner to spend real time in your operations before writing code.
They plan for life after launch. Launch starts the product's life. Post-launch optimization, user feedback and ongoing architecture decisions need people who know the product's history.
Choosing wrong is expensive. Switching partners mid-project loses 3-6 months of accumulated knowledge and adds 2-3 months of onboarding for the new team.
Frequently asked questions
How many portfolio projects is enough?
Quality over quantity. Three to five projects that are similar to yours in scope, technology, and industry matter more than 50 generic case studies. If none of their portfolio matches your domain, they'll be learning on your dime.
What if the agency has NDA restrictions on their best work?
Legitimate, especially for enterprise and government projects. They should still be able to describe the type of work, the technology stack, the team size, and the challenges they solved — without naming the client. If they can't discuss anything, that's suspicious.
Should I weigh design quality or engineering quality more?
Engineering quality. Good design on top of bad code leads to a product that looks great and breaks constantly. Bad design on top of good code can always be redesigned. The best agencies deliver both, but if you have to pick one to evaluate, go deeper on the engineering.
How can I test an agency's thinking before signing?
Give them one of your real technical challenges to review. A vendor proposes a solution on the spot. A strategic partner asks clarifying questions first.
How does Globalbit demonstrate its portfolio?
Our case studies name the client, state what we built and give measurable outcomes, and every project on our portfolio page has live links and named team members who still work here. We've delivered 200+ projects across fintech, healthcare, government, retail and AI, and 80% of our clients stay with us for four years or more. Our longest client relationship spans 10+ years. A few examples of long relationships:
- [Egged](/cases/egged): a 3-month rescue took the app from 2.2 to 4.7 stars and grew into a 5-year partnership as Egged's mobile team.
- [Care Laser](/cases/care-laser): a 5+ year partnership, from a 6-month rollout across 33 clinics to ongoing development.
- [Espresso Club](/cases/espresso-club): 4 years building the brand's apps and website, now used by 350,000+ monthly active customers.
See our work and talk to our team.
