Growth hacking agencies: what they do, how to check one and the tactics to refuse
Pay for a method you can inspect: a target metric, a scored backlog and an experiment log you own.
Published by AI Knows Us · Updated 13 October 2026
A growth hacking agency runs a steady stream of small experiments across your product, website, email and ads to find the few changes that move sign-ups, sales or retention, then repeats the winners. It suits a business that already has a product people want and tracking that works. Judge one by its experiment log, its method for choosing what to test and the metrics it reports, and pay for a process you can inspect, never for a promised number.
Where the term comes from and what it means now
Sean Ellis used the phrase "growth hacker" in 2010 for a person whose job is growth above all else. The practice grew from start-ups that had to find customers with small budgets and a product they could change quickly. Today most agencies that use the label mean a working method: set a target metric, list ideas, rank them, test them quickly, keep what works. The label is looser than the method, so ask what the agency does and not what it calls itself.
Growth hacking against the neighbouring services
| Service | Main job | Typical output |
|---|---|---|
| Growth hacking agency | Find and repeat the changes that lift a funnel metric | An experiment log, tested changes, a revised funnel |
| Growth marketing consultant | Diagnose the constraint and prioritise; often hand work back | A plan and a ranked backlog; see growth marketing consultants |
| Business growth agency | Wider commercial growth work, including sales and partnerships | See business growth agency |
| Performance or demand agency | Run paid channels at a target cost | Campaigns and reports |
| Conversion specialist | Improve the pages and flows people use | Tests on pages; see conversion rate optimisation |
What a good engagement contains
- A target metric. One number, such as activated trials per month or repeat orders. Agree its current value before work starts.
- A funnel map. The steps from first visit to paying customer, with the conversion at each step. Dave McClure's acquisition, activation, retention, referral and revenue stages are a common frame.
- An idea backlog with a scoring method. ICE, which scores each idea on impact, confidence and ease, is a popular simple choice. The agency should show the scored list and explain why the top item is on top.
- A test plan for each experiment. The hypothesis, the audience, the measure, the sample needed and the date it will be called.
- A log. Every experiment, including the losers, with the result and what was learned. This is the most valuable thing you buy, and it belongs to you.
- A rhythm. A weekly review and a monthly summary against the target metric.
When to hire one and when to wait
- Hire when customers already buy or use the product, you can see the funnel in your analytics, and you can ship changes to the site or product within days.
- Wait when you have few customers and have not yet learned why they buy. Talking to twenty customers will teach you more than twenty experiments.
- Wait when tracking is broken. An experiment without trustworthy measurement produces opinions.
- Wait when nobody on your side can implement changes. An agency's ideas stall behind your release schedule.
Fee models
| Model | How it works | Watch for |
|---|---|---|
| Monthly retainer | A fixed fee for a team and a number of experiments | Ask how many tests run each month and who runs them |
| Project | A fixed scope, such as an onboarding redesign with tests | The scope should end with a measured result |
| Hourly or day rate | Time and materials | Agree a monthly cap |
| Base fee plus bonus | A smaller fixed fee plus a share tied to a metric | Define the metric and its measurement in the contract |
Avoid any provider who offers guaranteed growth or a fixed number of customers. The honest offer is a method and a pace.
How to check an agency before you sign
- Ask for a redacted experiment log. A real team has one, with hypotheses, outcomes and failed tests.
- Ask how they decide what to test first. You are listening for a scoring method and a link to your numbers.
- Ask who runs the work. Names, roles and how many clients each person handles.
- Ask for a reference whose business is like yours. Speak to them about speed, communication and what happened when a test failed.
- Ask what they will not do. A good agency can name tactics it refuses.
The tactics to refuse
Growth hacking has a reputation for shortcuts, and some of them cost businesses their accounts, their reputation or their legal standing.
- Fake reviews and bought followers. Platforms prohibit them and customers notice.
- Mass automated following, messaging or posting. It breaks platform rules and gets accounts restricted. Our notes on Instagram growth agencies and YouTube growth agencies explain how to tell real growth from purchased numbers.
- Scraped or purchased email lists. Email laws and mailbox providers punish unsolicited bulk mail. Build a list people chose to join.
- Link schemes. Google's spam policies list buying and selling links that pass ranking credit as link spam.
- Misleading claims. Countdown timers that reset, invented scarcity and fake testimonials harm the brand and break advertising rules.
Experiments that usually pay off first
- A shorter sign-up form, or one step removed from checkout.
- A welcome email series that guides a new user to the first useful action.
- A referral offer that rewards both sides, with a link that tracks credit.
- A pricing page that makes the recommended plan obvious.
- An abandoned cart or lapsed user message.
- A new landing page for each of the top three paid search terms.
Each one is small, quick to build and measurable in a few weeks. Our guide to viral marketing explains the share loops behind the referral idea.
Questions about growth hacking agencies
Is growth hacking different from digital marketing?
Digital marketing is a set of channels. Growth hacking is a method of testing, and it often crosses channels and the product itself.
How many experiments should an agency run?
That depends on your traffic, because a test needs enough visitors to reach a clear result. A site with modest traffic may run two or three tests a month, and a high traffic product more.
Do I need developers for growth hacking?
You need someone who can change the site or product quickly. That might be your team, the agency's developers or a no-code builder.
What does a growth hacking agency cost?
Quotes vary widely by scope and team. Ask for a written scope with the number of experiments, hours and reports, and compare quotes for the same scope.
Can a start-up do this without an agency?
Yes. A founder or a marketer with a backlog, a scoring sheet and a weekly review can run the method. An agency adds capacity and experience.
A new channel to put on the experiment backlog: the answers assistants give
Every growth backlog should have a line for the places buyers now ask before they search. Assistants answer "which tool should I use for this" with a short list, and the businesses on that list receive visits and sign-ups that never appear as a search query. Make it an experiment with a measure: ask ChatGPT, Gemini and Perplexity twenty buyer questions, count how often your brand is named, change one page or one outside listing, and ask again two weeks later. AI Knows Us (aiknowsus.com) keeps that count across five assistants week after week, so the result of each change is on a chart and not in someone's memory.