How to Read an Agency Portfolio

Every agency portfolio looks good. That is the entire purpose of it, and it is why comparing three of them tells you almost nothing — you are looking at three sets of work chosen and presented by the people who want your business.

The useful signal is not in how the work looks. It is in what the case study is willing to tell you, and what it quietly leaves out.

Here are six checks that take about fifteen minutes and separate agencies that ship from agencies that present.

1. Is There a Live URL You Can Open Right Now?#

Start here, because it eliminates the most cases fastest.

A case study without a link to the live site is asking you to take the work on faith. Sometimes there is a legitimate reason — an internal tool, an app that has since been delisted, a client under NDA — and a good agency will say which. What should worry you is a portfolio where several pieces have no verifiable destination and no explanation.

When there is a link, open it. Then check two things: is the site still live and does it still look like the screenshots, and who is credited in the footer. Work that was rebuilt by someone else a year later is a different claim from work that is still running.

2. Are Results Measured or Estimated — and Does the Page Say Which?#

This is the check that separates careful agencies from confident ones.

"+38% booking enquiries" is a strong claim. The question is where the number came from. Was it measured in the client's analytics, over what window, against what baseline? Or is it an estimate the agency produced from their own data?

Both are legitimate. Presenting the second as the first is not.

Look for the disclosure. An agency that writes "estimated impact, measured approximately 90 days post-launch" is telling you the epistemic status of its own numbers, which is a much better sign than an unqualified percentage. An agency with a wall of precise-looking figures and no methodology anywhere is hoping you will not ask.

Then ask anyway. "How was that measured?" is a completely reasonable question and the answer is diagnostic. "We can show you the analytics" and "it's our internal estimate, here's how we arrived at it" are both fine answers. Vagueness is not.

3. Is the Stack Named, and Does It Match What You Can Inspect?#

Most case studies list technologies. Very few are checked.

You can check them, and it takes a minute. Open the live site, view source, and look at what is actually being served — the framework's fingerprints are usually obvious. For a claimed CMS, look at how content is delivered. For a claimed payment provider, look at what the checkout actually loads.

The reason this matters is not pedantry. A stack listed inaccurately means either nobody checked the case study before publishing it, or the detail was decorative. Both tell you something about how the agency handles detail on work that is not for sale.

4. Does the Case Study State the Problem, or Only the Outcome?#

Weak case studies describe what was built. Strong ones describe what was wrong.

The difference is that a problem statement is falsifiable and specific: "enquiries arrived through Instagram DMs and every one opened with the same three questions, so the team re-typed the same replies all day." That is a real business, and the solution that follows can be evaluated against it.

"We delivered a modern, responsive website with a clean user experience" describes nothing. It could be any project. If most of a portfolio reads that way, the agency is showing you deliverables rather than thinking, and you will get deliverables.

5. Is the Work Theirs, or a Template With a Logo On It?#

Harder to check, but there are signals.

Look across the portfolio for structural repetition. Different industries should produce different page structures. If a car rental site, a law firm and a restaurant all have the same six sections in the same order, you are looking at one template sold three times.

Check the details that templates cannot fake. A booking flow with real availability. A filter set that matches how that industry actually shops. A form that asks for the three things that specific business needs. Those come from understanding a business; a template cannot produce them.

Ask who did the work. Agencies subcontract, which is normal. What matters is whether you will be working with the people who did the work you are looking at, and whether they are still there.

6. Can They Show You Something in Your Language and Your Market?#

For the UAE specifically, this is the check most often skipped.

If you need Arabic, ask to see an Arabic site they built — and then have someone who reads Arabic look at it. Machine translation is instantly visible to a native reader and invisible to everyone else. Layout that breaks under right-to-left is common in work that added Arabic at the end.

Same for market specifics. If you are a rental operator, does anything in the portfolio suggest they understand Salik reconciliation? If you run holiday homes, do they know what a DET permit is? An agency that has worked in your category will show it in the details; one that has not will talk about design.

What to Ask in the First Call#

Five questions worth more than an hour of portfolio browsing:

1. "Which of these projects went badly, and what happened?" Everyone has one. An agency that cannot name it is either inexperienced or not being straight with you.
2. "Who will actually work on this, and are they the people who built the work you showed me?"
3. "How was that result measured?" — pointing at a specific number.
4. "What happens after launch?" Maintenance, hosting, who fixes things, what it costs.
5. "Who owns the code?" The answer should be you, unconditionally, and it should be in writing.

What We Publish, and Why#

Applying this to ourselves, because an article like this is worthless from an agency that does not:

Our results are labelled as estimates. Every metric on our case studies carries "estimated impact, measured approximately 90 days post-launch" because that is what they are — our own post-launch measurement, not audited client analytics. We would rather publish a qualified number than an unqualified one.

Not every project has a metric, and those get depth instead. Some of our work has no measured lift we can stand behind, so those case studies carry a deliverables list rather than an invented figure. Pampa is the clearest example — no results tiles, and a full account of what was actually shipped, which is verifiable against the live site in a way a percentage is not.

Most projects link to the live site. Where one does not, it is because there is nothing public to link to, and the case study should say so rather than leave you wondering.

That is the standard we are asking you to hold us to, and the same one worth holding anyone else to.

Frequently Asked Questions#

Why do agencies hide client names? Usually a genuine NDA, particularly in government, finance and enterprise work. What should be disclosable even under an NDA is the shape of the project — industry, scale, scope, outcome. An agency that can describe none of that is hiding more than a name.

Are "+38%" style claims trustworthy? They are trustworthy in proportion to how they are qualified. Ask how it was measured, over what window, against what baseline. An agency that discloses the measurement was its own estimate is being more careful than one that presents an unqualified figure.

Should I ask for references? Yes, and ask for a client who left as well as one who stayed. How an agency talks about a relationship that ended tells you more than a happy reference does.

What if their best work is under NDA? Reasonable, and it should not be their whole portfolio. Ask for a walkthrough under your own NDA, or for a described version with the identity removed. "We cannot show you anything" is not an acceptable answer from an established agency.

Does portfolio size matter? Less than depth. Fifteen projects with real problem statements beats fifty screenshots. What matters more is whether any of them resemble your situation.

How do I check the technology claims? Open the live site and view source — most frameworks are identifiable immediately. Check what the checkout loads if a payment provider is named. It takes a minute and tells you whether the case study was checked before it was published.

They subcontract. Is that a problem? Not inherently — most agencies do for specialisms. The question is whether the people who built the work you are evaluating will be on yours, and who is accountable when something breaks.

What is the single biggest red flag? A portfolio where every project is a success with a precise number and no methodology. Real work includes projects that underperformed, and an agency that presents none is either very new or not being straight.

The Bottom Line#

A portfolio is a sales document, and reading it as evidence rather than as marketing takes about fifteen minutes: open the live sites, check whether results are qualified, verify one technology claim, and read whether the case study can describe a real problem.

Agencies that ship will pass comfortably. Agencies that present will not, and you will have found that out before signing rather than three months in.

If you want to run these checks on us, our case studies are here — and if something does not hold up, tell us. The full range of what we do sits under our services and the industries we build for.