The analytics vocabulary,
without the sales gloss

Showing two variants to comparable groups and measuring which performs better, on the same event data the rest of your analytics uses.

Acquisition, Activation, Retention, Referral, Revenue: five stages used to decide which metric a team should be arguing about this quarter.

The moment a new user first gets the value the product promised. Everything before it is setup; everything after it is retention.

The share of new users who reach that moment. The single number most onboarding work is trying to move.

People who performed at least one qualifying action in a window. The window and the action are your definition, not an industry constant.

Rolling raw events into summaries. Fast, but only answers questions you asked in advance, which is why Countly keeps the raw events too.

A written summary of what changed in the data and why, generated from live events rather than from a monthly export.

Watching metrics for values outside their normal range and raising them before someone notices in a weekly review.

Irreversibly stripping data of anything that could identify a person. Unlike pseudonymization, there is no key that brings the identity back.

Assigning an install or a conversion to the campaign, channel or referrer that produced it.

Recording interactions automatically instead of instrumenting each one. Quick to start with, expensive to govern once the schema drifts.

Revenue in a period divided by active users in that period. Useful next to retention; misleading on its own.

Analysis of what people actually did in the product (events in sequence) rather than of who they are on paper.

A group defined by actions taken, such as everyone who used a feature twice in their first week. Recomputed as the data arrives.

The share of sessions that end without a second meaningful interaction.

Splitting a metric by a property (country, plan, app version) to see which segment is moving it.

California privacy law giving residents the right to know what is collected about them, to delete it, and to opt out of its sale.

The share of users or revenue lost in a period. Its shape over time matters more than the number in any single month.

Clicks divided by impressions, for a message, a campaign or an in-product prompt.

Grouping users by when they arrived or what they did, then comparing those groups over time instead of averaging them together.

Recording what each user agreed to and enforcing it at collection, so consent is a property of the pipeline rather than a banner.

The share of people who complete a defined step. Only comparable when the denominator is defined the same way each time.

A system that unifies customer data from several sources into one profile. Countly builds those profiles in the same store the analytics runs on.

The messages, prompts and product changes aimed at a specific behavior, and the measurement of whether they worked.

Following a person across sessions, devices and channels instead of analyzing each touchpoint on its own.

Expected total revenue from a customer over the whole relationship. Used to decide what acquiring one is worth.

Distinct users active in a day, on your own definition of active.

A saved arrangement of reports. Useful as a shared reference; dangerous when it becomes the only view anyone opens.

The rules for what is collected, how it is named, who may see it and how long it is kept, decided before the data becomes a report.

Collecting only what a stated purpose needs. The cheapest way to reduce risk is not to hold the data at all.

The path an event takes from the SDK to the store: validation, transformation, consent checks and masking, in that order.

The requirement that data physically stays in a given jurisdiction. On Flex you pick the region; on Enterprise it is wherever you deploy.

The principle that data is subject to the laws of the country it sits in, and that you, not a vendor, decide which country that is.

Querying raw events on any property, without waiting for someone to build the report first.

The share of a targeted audience that reacted to a message or a prompt.

One recorded action with a timestamp, a user and a set of properties. The unit everything else on this page is built from.

The attributes carried by an event (plan, screen, version, value) that make it possible to break a metric down later.

Deciding which actions matter, naming them consistently, and sending them from every platform the product runs on.

Running controlled changes and measuring the effect on real behavior rather than on opinion.

How many of the people who could use a feature actually do, and whether they come back to it.

A switch that turns functionality on for a chosen segment without shipping a release.

Data you collect directly from your own users in your own product. Not bought, not inferred, not borrowed from a third party.

Measuring how many people get from one step to the next, and where they stop.

EU regulation governing personal data, covering the lawful basis for collection, rights to access and erasure, and real consequences for ignoring both.

Reducing precision on purpose (an age band instead of a birth date) so a record stops being uniquely identifying.

Triggering an action when a device enters or leaves a defined area.

A cycle where the output of one user’s activity produces the next user, rather than a funnel that has to be refilled from the top.

A visual overlay showing where people clicked, tapped or scrolled on a screen.

US law governing protected health information. In practice it decides where the data may run, which is a deployment question before it is a feature question.

Messages shown inside the product to a segment defined by behavior, rather than pushed to a device.

A finding that changes a decision. Everything else is a chart.

The code that emits events. Its quality sets a ceiling on every analysis done later.

A guarantee that any record is indistinguishable from at least k−1 others, so no single person can be picked out of a dataset.

A metric a team has agreed to be judged on. Useful in small numbers, meaningless in dozens.

Messaging tied to where a user is in their relationship with the product rather than to the calendar.

Total value a user produces over the whole relationship. See Customer Lifetime Value.

Behavioral analytics for apps, including sessions, crashes, performance and the device and OS context around them.

How the product turns usage into revenue, and which behaviors precede that turning.

Distinct users active in a month. Commonly used for pricing, which is why the definition deserves scrutiny.

A survey score from a single recommendation question. Most useful next to the behavior of the people who answered it.

Revenue kept from existing customers including expansion and contraction, over a period.

The one measure a company agrees best reflects delivered value, chosen so that moving it is hard to fake.

Treating web, app, email and support as one conversation with the same person instead of four separate ones.

The path from first open to first value. Usually the shortest route to a retention improvement.

Following the actual sequences people take through a product, rather than the one drawn on a whiteboard.

Changing what the product shows or says based on live behavior, computed in the same system that stores the behavior.

Analysis of how people use a product, in order to change it. Distinct from web analytics, which counts traffic.

A model where the product itself does the acquiring, converting and expanding, and sales follows the usage signal.

Replacing identifiers with a key so records can still be joined but not read as a person, reversible only with the key.

A message delivered to a device outside the app, targeted from the same behavioral data as the analytics.

Recovering a person’s identity from data that was supposed to be anonymous, usually by combining it with another dataset.

The share of a cohort still active on each day after signup. A curve that flattens is a product; one that reaches zero is a campaign.

The share of a cohort still active after a given period.

Running the software on infrastructure you own and operate, including networks with no outbound connection.

Dividing users by properties or behavior so a number becomes a statement about someone in particular.

Analytics running on infrastructure you control, where no vendor holds custody of the behavioral data.

A period of continuous use, closed after a defined gap of inactivity. The gap is a setting, not a law.

The library that collects and sends events from an app, site or device, queuing them when the network is gone.

DAU divided by MAU: how much of the monthly audience shows up on a given day.

Asking users directly, targeted by what they just did, so the answer arrives with its behavioral context attached.

How long it takes a new user, or a new customer, to get the first real result. Measured in the product, not in the contract.

Any point of contact between a person and the product or company.

The agreed list of events and properties, with names and owners. The document that keeps analytics usable in year three.

Reading a metric’s direction over time rather than its value on one day.

Distinct people in a period, counted once regardless of how often they came back.

Everything known about one person (properties, events, consent state) merged into a single record inside your deployment.

Tags appended to a link so the traffic it produces can be attributed to a source, medium and campaign.

A number that only goes up and changes no decision. Total registered users is the usual example.

Distinct users active in a week.

Measurement of site traffic and behavior. Overlaps with product analytics, but starts from the visit rather than from the user.

Data a person gives you deliberately (preferences, intent, answers to a survey) rather than data observed about them.