Growth Hacking
7 Customer Acquisition Experiments for Your Next Growth Test
Choose an acquisition experiment from a diagnosed constraint, define the cost and downstream outcome, and decide what evidence would justify scaling it.
Editorial revision note — September 22, 2026: This practical guide replaces an earlier first-person growth story whose supporting measurement records were not available for this revision. It does not report a verified client engagement or promise a particular uplift.
Start with a constraint, not a list of tactics
Acquisition volume can rise while contribution falls. Before testing a channel, specify the customer segment, eligible audience, primary outcome, conversion lag and fully loaded cost boundary. A qualified opportunity and a paid customer are different outcomes; do not compare their acquisition costs as if the denominators were interchangeable.
Choose one constraint from observed evidence: insufficient qualified reach, weak activation, slow sales follow-up or poor repeat purchase. Record the baseline period and the decision the experiment should inform. An attractive idea without an identifiable bottleneck belongs in the backlog, not automatically in the media plan.
Seven testable options
- Referral incentives: Test whether a reward brings additional qualified customers rather than subsidising customers who would arrive anyway. Define reward eligibility, fraud controls, fulfilment costs and a minimum observation window. Where feasible, compare randomly assigned eligible customers; otherwise label the result observational.
- Specialist creators: Evaluate audience fit before follower count. Agree disclosure, attribution windows and a landing-page proposition. Compare qualified customers and contribution after creator fees; do not declare victory from impressions or a tracked-code sale alone.
- Useful interactive resources: Let visitors complete a meaningful task before asking for contact details. Track resource use, voluntarily requested follow-up and qualified opportunities separately. A calculator completion is evidence of use, not automatically purchase intent.
- Behaviour-triggered lifecycle messages: Trigger an intervention around a verified task failure or unmet need. Define consent and suppression logic, and record complaints and opt-outs. Opens alone cannot establish business value. Evaluate the downstream task or purchase outcome.
- Product sharing: Make shared output useful to the recipient and give users control over attribution or branding. Track exposure, referred activation and retention. A “powered by” label is a distribution mechanism; it does not by itself demonstrate a network effect.
- Activation support: Identify behaviours associated with later retention, then test an intervention that helps eligible users reach value. Correlation helps generate the hypothesis but does not prove that forcing a behaviour improves retention.
- Entry offers and expansion: Reduce initial commitment only when the entry product solves a real need and has sustainable economics. Track support cost, paid conversion, expansion and churn. A cheap introductory offer can attract a different customer mix rather than improve the same funnel.
Write the decision rule before launch
Record the hypothesis, assignment unit, eligible population, primary metric, guardrails, cost cap, analysis window and owner. Define a minimum useful effect and the precision required to distinguish it from noise. Use the stopping rules supported by the selected experimental method; repeatedly checking a conventional fixed-horizon significance test is not a substitute for a valid sequential design.
For a synthetic accounting example, a campaign costs €2,000 and produces 40 new paid customers within the agreed window: observed campaign CAC is €50. If the cost excludes staff, creative or sales assistance, label it campaign CAC, not fully loaded CAC. These figures do not establish incremental customers; that requires a suitable comparison design. Compare mature cohorts before estimating payback.
Methodology and limitations
This is a planning framework, not a report of seven proven wins. It does not supply a universal sample size, lift benchmark or best channel. Customer mix, seasonality, contamination, sales capacity and measurement loss can change the result. Report uncertainty and failed tests alongside successes. Use the site's incrementality decision framework to separate attributed conversions from causal impact, and the marketing-data-contracts guide to define events before launch.