Adaptive Variables

Change your app today.
Ship the release later

A value your app reads from Countly instead of its code: a flag, a number, a piece of text or a JSON object. Change it for everyone, for a segment, for 10% of profiles or with each person’s own details in it, test it against another, and roll out the winner, all without a release.

In this demo car app, an experiment on how charging pins cluster on the map has a winner, and one button makes it the value every driver gets.

Part of engagement

From a flag to a winner,
without a release

The usual path of one change: shipped switched off, turned on for a few, tested against what was there, and kept only if it wins.

  1. 01

    Ship it off

    Put the new code behind a variable whose default is false. It reaches every device switched off, and nothing changes until you say so.

  2. 02

    Turn it on for 10%

    A percentage rollout picks a stable slice of profiles. Widen it to 30% and everyone already in stays in.

  3. 03

    Test it

    Run an experiment with up to eight variants and three goals. Countly calls a winner only when it is clearly better.

  4. 04

    Roll out the winner

    One button turns the winning value into what your app gets, so the result outlives the experiment.

Personalized by audience

A value for every audience,
from data you already have

A separate feature-flag tool only knows the attributes you copy into it. A condition here reads everything Countly already holds about a profile, so the drivers with a home charger, the shoppers slipping away and your beta testers can each get their own value, kept current as the data changes. The examples are from the demo car app.

Profile properties

What every SDK reports about a device and a person.

PlatformCountryLanguageDevice

Your custom properties

Anything your app or your backend sets on a profile.

Home Charger InstalledDriver LevelBattery Capacity (kWh)

Cohorts

Behavior-based audiences that update themselves, the same ones your reports use.

Daily DriversCharging RegularsNew Vehicles (30d)

Profile groups

Hand-picked lists, for beta testers and staff.

Beta FleetOTA Early Ring

App version

Compared as a version number, so 5.10.0 counts as newer than 5.9.0.

Newer thanOlder thanExactly

A share of profiles

A stable random slice, widened without reshuffling.

10%30%Own seed

Personalized for one

Not just a value per segment.
A value per person.

Any text in a value can carry a profile property or one of your custom properties. Countly fills it in on the server for the person who asked, with the fallback you set for anyone it knows nothing about. The text shown in the dashboard stays exactly as you wrote it.

  1. “Welcome back, Ada!”

    One value with the profile’s name in it. Anyone without a name on file gets “Welcome back, friend!” instead.

  2. “Your Gold perks are waiting”

    A custom property your backend sets, such as a loyalty tier, filled in per person, with an optional capital letter.

  3. “Ada, your charger is ready”

    It works in any text inside a JSON value, and in the values conditions and experiments serve, not only the default.

How a value is decided

Predictable for engineers,
simple for everyone else

Every fetch returns values already resolved for the device that asked. These are the rules that decide them.

  1. 01

    Values

    True or false, a number, text, or a JSON object, with an optional date after which it stops being served.

    Boolean · Number · Text · JSON
  2. 02

    Order

    Conditions are checked in the order you list them on each variable, and the first one that matches decides. Put the narrowest first.

    First match wins
  3. 03

    Rollouts

    A random but stable percentile per device. Give a rollout its own seed to pick an independent slice.

    1 to 99%
  4. 04

    Experiments

    A running experiment decides the value for the profiles it enrolled, ahead of every condition. The variable is locked until it completes.

    Up to 8 variants
  5. 05

    Winners

    A variant wins when the low end of its 95% improvement range clears your target and it is at least 95% likely to beat the control.

    95% range
  6. 06

    Serves

    Each variable counts how often it was served, broken down by the condition that decided it.

    Hits

Works with

Adaptive Variables, with the
rest of Countly

Every feature reads the same live data, so an answer in Adaptive Variables doesn’t stop there. These are the handoffs built in.

All features

Data ownership

Values served from
your Countly server

Every fetch goes to the server that holds your data, so the profile properties and cohorts a condition reads never have to be copied to a separate feature-flag service.

Data privacy

Self-hosted, or dedicated

Run Countly in your own cloud, your data center or an air-gapped network. Or let us run a single-tenant instance in the Google Cloud region you choose.

One less vendor

Flags, rollouts and experiments use the same SDK, the same profiles and the same cohorts as your analytics.

No default vendor access

On Enterprise, Countly has no access unless you grant it. On Flex, access is limited, granted by you and logged.

ISO 27001Certified
ISO 27017Certified
ISO 27018Certified
SOC 2Type II examined
TISAX®Participant*

*Countly Ltd. is a TISAX® Participant. TISAX® is a registered trademark of and governed by ENX Association. Learn more at enx.com/tisax

Try it

Every feature.
One data layer.