Product terms in plain language. No jargon without a definition.
A
- A/B test
- An experiment that shows two versions of something to comparable, randomly split groups of users at the same time and measures which performs better on a metric decided in advance.
- Activation
- The moment a new user first gets real value from a product. Activation rates usually pass the "would this change a decision?" test that vanity metrics fail.
- Agent network
- Local shops and people who help customers complete a service in person, such as cashing mobile money in and out. In many markets the best interface is a trusted person nearby, not an app.
- Andon cord
- A cord on Toyota's production line that any worker can pull to stop the process when they spot a problem. In product teams it stands for everyone's right to raise a problem while it is still cheap to fix.
C
- Cohort
- A group of users who joined in the same period, such as the same week. Comparing cohorts shows whether retention is getting better or worse as the product changes.
- Commitment signal
- Something a customer gives up to show real interest: a follow-up meeting, an introduction, a deposit or a pilot. Compliments cost nothing, so they prove nothing; commitment costs something, so it tells the truth.
- Concierge test
- Delivering a service by hand for your first few customers, so you learn what they need before you automate anything.
- Continuous discovery
- Talking to customers every week as a habit, not only when you're stuck or before a launch, so decisions rest on current reality rather than old assumptions.
- Cost of delay
- What every week of waiting costs you. Work tied to a deadline, a season or a customer about to leave is worth more now than later, so two ideas of equal value aren't equal if one expires next month.
- Customer discovery
- Learning about customers' real problems before and while you build. Good discovery asks about the past rather than the future, chases stories rather than opinions, and looks at what people have already tried.
D
- Dark pattern
- A design trick that nudges people into doing something they didn't mean to do, such as hidden fees or making cancellation a maze. It borrows from long-term trust to inflate short-term numbers.
E
- Evaluations (evals)
- Sets of real examples with known good answers, used to measure the quality of an AI feature instead of judging it by feel.
F
- Fake door test
- A landing page or button for something that doesn't exist yet, used to measure how many people try to use it before you build it.
- Feature flag
- A switch that lets a team turn a new feature on or off for specific users without deploying new code, so releases can be rolled out gradually and reversed in seconds.
G
- Good, better, best tiers
- Three pricing tiers, each designed for a type of customer. The entry tier lowers the barrier, the middle is where most should land, and the top serves power users and makes the middle look reasonable.
- Guardrail metric
- A metric you protect while pushing a target, so you don't win the number and lose the goal. For example: grow sign-ups while keeping first-week retention healthy.
H
- Human in the loop
- A design where AI drafts, flags or suggests and a person approves or decides, used where the stakes are high. It earns trust and keeps mistakes cheap.
J
- Jobs to be done
- A way of understanding customers by the progress they are trying to make, not by who they are. Nobody wants a drill; they want a hole in the wall, so they can hang a photo, so the house feels like home.
K
- Kaizen
- The habit of continuous improvement: catching and fixing small problems early, again and again, so they never grow into big ones.
M
- Minimum viable product (MVP)
- The smallest thing you can put in front of real people to learn whether you're right. It is an experiment, not version one with fewer features.
N
- North Star metric
- One number that captures the value customers actually get, such as successful transactions for a marketplace. It gives the whole team the same direction.
O
- Onboarding
- The first experience a new user has with a product, designed to get them to the moment they first get real value (activation) as quickly as possible.
- Outcome
- The change a product creates in a customer's life or business, such as new users reaching their first win in five minutes. Outcomes are what customers feel and pay for; they are the opposite of output.
- Output
- The activity a team produces: features shipped, hours logged, tickets closed. Output is easy to count and can look impressive while nothing changes for the customer.
P
- Pricing unit
- What you charge for: per user, per transaction, per project. The best pricing units grow as the customer's success grows, so you only win when they win.
- Prioritisation
- Deciding what to work on now, next and not yet. It works best against a shared goal, which turns "is A better than B?" into "which of these moves our goal this quarter?".
- Problem statement
- One sentence that proves a problem is worth solving: a specific person, trying to reach a specific goal, is blocked by something, and it costs them time, money or peace of mind. Every blank needs real evidence.
- Product-led growth (PLG)
- A strategy where the product itself drives acquisition, activation, retention and revenue, usually through self-serve sign-up, a free plan or trial, and sharing built into the product.
- Product-market fit
- When a product pulls customers in rather than being pushed onto them. Signs include flattening retention, many users saying they'd be very disappointed without it, growth that arrives for free, and loud complaints when it breaks.
- Psychological safety
- A team climate where anyone can raise a problem or a doubt without fear. It isn't softness; it's how good teams catch problems while they're still cheap.
R
- Retention
- The share of users who keep coming back over time. Acquisition shows how good your marketing is; retention shows how good your product is. A curve that flattens means people found real value.
- Retrospective
- A blame-free review after a launch or sprint with three questions: what did we expect, what actually happened, and what will we do differently?
- RICE score
- A simple scoring model for ideas: Reach (how many people it touches), Impact (how much it moves the goal), Confidence (how sure you are) and Effort (how much work). Reach × Impact × Confidence ÷ Effort. The number isn't magic, but it turns opinions into a conversation.
- Riskiest assumption
- The belief your whole idea rests on, which would bring everything down if it were wrong. Test it first, even if it isn't the easiest one to test.
- Roadmap
- A plan of what a team will work on and why. A good roadmap names the outcome behind every item, which you can check by finishing the sentence "We're building this so that…".
S
- Statistical significance
- A measure of how unlikely a result would be if there were really no difference between the versions you tested. It tells you whether a difference is probably not just noise; it doesn't tell you whether the difference is big enough to matter.
- Success criteria
- The result you write down before an experiment that decides whether you continue, such as "if twenty percent sign up, we continue". Setting it after you see the data turns validation into hope.
- Survivorship bias
- The mistake of drawing conclusions only from the cases that made it through a filter, such as successful companies or customers who stayed, while ignoring the ones that left.
T
- Technical debt
- The future cost of shortcuts in code or design taken to move faster now. Like financial debt it can be useful, but unpaid it slows every later change and raises the risk of failure.
V
- Value-based pricing
- Setting a price from what the product is worth to the customer, such as the time or money it saves them, rather than from what it costs you to run.
- Vanity metric
- A number that goes up and to the right while nothing important changes, such as total downloads or followers. A vanity metric changes moods; a good metric changes decisions.
- Viable
- In an MVP, viable means it solves one painful job end to end for a narrow group, well enough that they would be genuinely disappointed if you took it away.
W
- Wizard of Oz test
- A product that looks automated to the customer while a person does the work behind the scenes. It tests demand and the experience without building the machinery.