Growth hacking is the only service on the shortlist where the customer is usually a founder rather than a marketing department, and that changes everything about how it should be bought. A founder between funding rounds is not looking for campaigns. They are looking for evidence that one repeatable way of acquiring and keeping customers exists, in time to put it in front of an investor. The deliverable is a conclusion, not a set of assets.
Telling real growth hacking apart from the relabelled kind
A large share of firms using the growth hacking label are running paid social campaigns under a more exciting name. That is worth knowing rather than being offended by, because paid acquisition is a perfectly good service and it is priced and judged differently. The distinction is easy to test with one question: what part of the product can you change?
Growth hacking proper works across the whole funnel, which necessarily includes things outside advertising: the signup flow, the onboarding sequence, the pricing page, the first session, the notification that brings someone back on day three, the referral mechanic. If the answer to the question is that they will optimise the ads and recommend changes elsewhere, you are buying media management. If the answer involves working inside the product with your engineers, you are buying what the label implies.
The second test is about failure. Ask for an account of an experiment programme where the conclusion was that the channel did not work. A team that runs genuine tests has many of those and talks about them without embarrassment, because a fast negative result is the product. A team that only presents successes is presenting case studies rather than a method.
What the engagement should look like structurally
- A short cycle. Work organised in fixed sprints with a review at the end of each, not a monthly report. The cadence is the mechanism; it is what forces a decision instead of a continuation.
- A written hypothesis backlog. Each item stating what is believed, what would prove it wrong, what it costs to test, and how big the effect would need to be to matter. Prioritised in the open, with you in the room.
- Instrumentation first. The first sprint is almost always unglamorous: getting events firing correctly, defining what activation actually means for your product, and making sure a cohort can be followed over time. Skipping this produces a quarter of confident conclusions drawn from broken data.
- An embedded working relationship. The team sits in your channels, attends your standups, and can ship behind a feature flag. Experimentation conducted at arm's length through a weekly call runs at a quarter of the speed.
- A learning log that belongs to you. Every test, its setup, its result and its conclusion, in a document you own. This is the durable asset of the engagement, and it is the one most often left behind in an agency workspace.
Where a Dubai based startup usually finds its constraint
Two patterns show up often enough in this market to be worth anticipating.
The first is that acquisition is rarely the binding constraint even though it is the one founders ask about. A company operating from here typically has a small home market and an ambition that spans several neighbouring ones, which means the growth question is usually about expansion readiness rather than about spending more locally. Testing whether a second market responds at all is a different experiment from improving conversion in the first, and it is cheaper to answer than most teams assume.
The second is onboarding friction created by things nobody in marketing controls. Identity verification requirements, payment methods that vary by nationality and by bank, delivery address formats that do not match how people actually describe where they live, and the point at which an app asks for a document. Drop off at these steps is frequently larger than anything an advertising budget could compensate for, and it is invisible unless the funnel is instrumented properly. A growth hacking partner earns its fee in the first month by finding this, and a partner who never looks at the product will not.
Language belongs in the same conversation. Offering the interface and the support conversation in Arabic is often treated as a localisation project for later, when it is frequently an activation lever that can be tested cheaply on a segment before anyone commits to translating everything.
Commercial terms suited to short, uncertain growth hacking work
Avoid the twelve month retainer. This work is bought in a first engagement of roughly one quarter, long enough to instrument, run several cycles and reach a defensible conclusion, with an explicit decision point at the end. Price it as a fee for the team's time rather than per deliverable, since the number of experiments is not something either side can promise honestly.
Two clauses matter more than the rate. Access: the partner needs your analytics, your product data and, in most useful arrangements, a development environment, so agree what they can touch, under what review, and who approves a release. Ownership: everything produced, including code written for tests, tracking implementations, audience definitions and the experiment record, is yours, with no residual dependency on a tool the agency owns.
Be equally clear about what a successful engagement produces. It is not a growth curve. It is a short list of things that demonstrably work, a longer list of things that demonstrably do not, and an instrumented product where the next question can be answered faster than the last one. Judge the proposal on whether it is structured to deliver that.
If the work turns out to be mainly engineering, the right partner may not be a marketing firm at all, and it is worth comparing against development firms that build product or, for mobile first businesses, teams that specialise in app development. If the conclusion of the first quarter is that one paid channel works and now needs scaling, that is a different job with different economics, handled by media buying specialists. To compare proposals from several experimentation teams against the same brief, describe the situation once and collect their offers, and for the wider definition of the category see the overview of growth hacking agencies.