Growth Engineering in English: A/B Tests, Funnels, and Activation Vocabulary

Master the English vocabulary growth engineers use daily — AARRR metrics, activation events, holdout groups, statistical significance, funnel analysis, and experiment communication.

Growth Engineering Has Its Own Language

Growth engineering sits at the intersection of product development, data science, and marketing. It has its own vocabulary — borrowed from all three disciplines and refined into a shared language that growth engineers, product managers, and analysts use to design experiments, measure outcomes, and communicate results.

If you work in growth and English is not your first language, fluency in this vocabulary helps you contribute more effectively in experiment design discussions, share results persuasively, and read analytics reports without ambiguity.


The AARRR Framework Vocabulary

AARRR (Acquisition, Activation, Retention, Referral, Revenue) — sometimes called the “Pirate Metrics” — is a framework for mapping the user lifecycle.

TermDefinition
AcquisitionThe process of getting users to discover and visit your product
ActivationThe moment when a user experiences the core value of the product for the first time
RetentionUsers returning to the product over time
ReferralUsers recommending the product to others, generating organic acquisition
RevenueUsers paying for the product or generating monetisable activity

Activation is the most discussed and debated metric in growth engineering. An activation event is a specific, measurable action that correlates with long-term retention — the moment a user has genuinely “got it.”

Identifying the right activation event is one of the highest-leverage analytical tasks in growth. Common examples: sending a first message, completing a first project, connecting a first integration.


A/B Testing Vocabulary

TermDefinition
A/B test (controlled experiment)An experiment in which users are randomly assigned to a control group (A) or a treatment group (B)
Control groupThe group that receives the existing experience
Treatment group (variant)The group that receives the new or modified experience
Holdout groupA group permanently excluded from a feature to measure its long-term effect
Randomisation unitThe entity being randomly assigned — often a user, device, or session
Sample sizeThe number of units in an experiment; must be sufficient for statistical power
Statistical powerThe probability of detecting a true effect if one exists
Type I error (false positive)Concluding that an effect exists when it does not
Type II error (false negative)Failing to detect a true effect
p-valueThe probability of observing the measured effect if the null hypothesis were true
Statistical significanceA result is statistically significant if the p-value is below a pre-defined threshold (typically 0.05)
Minimum detectable effect (MDE)The smallest effect size the experiment is designed to detect
Confidence intervalA range of values within which the true effect size is likely to fall

Statistical significance does not mean practical significance. A change that increases activation by 0.01% can be statistically significant with a large enough sample — but it may not be worth shipping. Always report effect size alongside significance.


Funnel Vocabulary

TermDefinition
FunnelA sequence of steps users take toward a conversion goal
Conversion rateThe percentage of users who complete a given step
Drop-offThe point in a funnel where users abandon the process
Funnel analysisExamining conversion rates and drop-off points across a funnel
Top-of-funnel (TOFU)Early stages of the funnel (awareness, acquisition)
Bottom-of-funnel (BOFU)Later stages (conversion, revenue)
FrictionUnnecessary difficulty or complexity in the user journey that reduces conversion
Aha momentThe moment a user understands the product’s value — closely related to the activation event

Experiment Results Vocabulary

TermDefinition
LiftThe relative improvement in a metric due to the treatment
DegradationA negative lift — the treatment made the metric worse
Novelty effectA temporary increase in engagement due to the newness of a change, not its underlying value
Network effect (in experiments)Interference between treatment and control groups in social or referral-based products
Segment analysisBreaking down results by user cohort, geography, or other dimensions
Guardrail metricA metric that must not decline in an experiment, even if the primary metric improves
North star metricThe single most important metric the organisation is optimising for

How to Present Experiment Results

When presenting A/B test results in a meeting or document, use this structure:

  1. Hypothesis: “We hypothesised that reducing the number of steps in onboarding from five to three would increase 7-day activation rate.”
  2. Experiment design: “We ran an A/B test with a 50/50 split for three weeks, targeting new sign-ups. The minimum detectable effect was set at 5% relative lift.”
  3. Results: “The treatment group showed a 7.3% relative increase in 7-day activation rate (32.1% vs 29.9%), with a p-value of 0.032 — statistically significant at the 5% level.”
  4. Effect size and confidence interval: “The 95% confidence interval for the lift is 0.8% to 13.9% — the true effect could range from modest to substantial.”
  5. Guardrail check: “No significant degradation was observed in 30-day retention or revenue metrics.”
  6. Recommendation: “We recommend shipping the three-step onboarding to all users.”

Example Sentences

  1. “We set the holdout group at 10% to measure the compounding effect of the activation sequence over six months — we’ll need that baseline when we evaluate the full feature bundle.”
  2. “The A/B test reached statistical significance after 18 days, but I’d recommend waiting for the full three weeks to mitigate the novelty effect.”
  3. “Our drop-off analysis shows that 43% of users abandon the funnel at the payment method step — this is our highest-leverage optimisation target for Q2.”
  4. “The activation event we are using is ‘first export completed’ — it has the strongest correlation with 90-day retention in our cohort analysis.”
  5. “The 95% confidence interval for this experiment is wide enough that we cannot be confident the effect is commercially meaningful — I recommend extending the test duration and revisiting next sprint.”

Common Vocabulary Mistakes in Growth Discussions

“Significant” vs “statistically significant”: In everyday English, “significant” means important. In growth, “statistically significant” has a specific technical meaning. Avoid saying “the results are significant” when you mean “the results are statistically significant at p < 0.05.” These are different claims.

“Conversion rate” specificity: Always specify which conversion rate you are discussing. “Our conversion rate improved” is ambiguous. “Our onboarding-to-activation conversion rate improved from 29% to 31%” is specific and actionable.

Confusing correlation and causation: In experiment discussions, use causal language only when you have a randomised experiment: “The treatment caused a 7% lift.” For observational data, use associative language: “Users who completed the tutorial showed 40% higher retention — we are running an experiment to test whether the tutorial itself drives retention or whether it is selected by more motivated users.”