Beta and experimental feature labels
We use a hierarchy of feature labels to set expectations with our users and internally: Experimental
➡️ Beta
➡️ GA.
GA (“General Availability”) has no in-product label and is the default user assumption for an unlabeled feature.
Why we use these labels:
Labeling features correctly lets us move quickly and release early in order to validate our direction. An Experimental
or Beta
label and its corresponding user-facing documentation:
- Set expectations so users are more forgiving of bugs
- Make users less likely to assume the overall quality of Sourcegraph is as brittle as an early feature might be
- Communicate that we are particularly interested in user feedback
- Communicate that a user’s feedback is more likely to have meaningful impact due to the early stage
- Let us release paid-tier features initially for free, to get usage feedback fast or to build our sales and marketing knowledge
When to use each label:
There’s no requirement that every feature begin as Experimental
and pass through Beta
before being in GA
. This depends on the feature size, the development time, and the cost of getting a GA release wrong.
While there is no strict required lifecycle, for the sake of our customers and our customer-facing teams, we do adhere to the minimum documentation, feature flag, and support requirements defined in our docs.
For example, you can’t label a feature that’s enabled by default
Experimental
, nor a feature with no documentation aBeta
, but you can have the full necessary forBeta
and still choose to useExperimental
.
Experimental features
The Experimental
label conveys that a feature is in its earliest stage.
Usually this means we have figured out a problem we’re trying to solve but aren’t sure what the general solution looks like – the pre-feature-market-fit stage. If a user came across an Experimental
feature, they might not know why or how to use it on their own.
We might also use the Experimental
label when we’re building the first version of that validated solution but haven’t yet reached the fidelity level of a minimum lovable product.
Ask yourself:
- Is it likely our feature direction might change dramatically?
- Are we still trying to determine if this feature is worth a major investment?
- Will users need high-touch communication to use this feature?
- Are we hoping to collect user feedback at the “what is this” or “I instead want…” level?
If not sooner, you must move to Beta
in order to enable the feature by default for all users.
Beta features
We use the Beta
label to convey a feature is still unfinished, but that its solution is fairly stable and ready for broad testing.
Usually this means we have figured out what problem we’re solving and have a validated, minimum-lovable solution for functionality that solves it for many users or configurations. If a user came across a Beta
feature, they should understand both why it’s valuable and be able to immediately realize some of that value a majority of the time.
Ask yourself:
- While the UX or implementation may be brittle, does the feature do “one thing well” for most users that makes them happy?
- If the feature does one useful thing well and is not brittle, you should likely be in
GA
instead.
- If the feature does one useful thing well and is not brittle, you should likely be in
- Is this feature at a quality you would be comfortable enabling it for all users by default?
- Does the feature communicate its own value without high-touch involvement?
- Are users more likely to give feedback on the “how I interact with this” or “I also want” rather than the “what is this” level?
- Are we ready to collect broad usage feedback in order to build our sales, pricing, or marketing strategy?
General Availability (GA) features
We use the term GA to refer to features that are complete and ready for broad use. GA features:
- Are stable and can be used across our entire customer base
- Are high-quality with no known bugs or issues
- Are commercially ready for customers to adopt (and purchase if it’s an add-on product or feature)
- Have established product-market fit
- Have clear pricing and packaging
For L1 product launches, the PM and PMM for the product should work together to establish GA criteria before the launch happens and should mutually decide when the product is ready for GA.
When we use an entirely different label
Rarely, sometimes we use a different label when a feature is in Beta
in order to benefit from limiting initial rollout without invoking users’ connotations with the word “Beta.”
For example: a feature may be ready for GA on a security and functional level, but we want to test the UX before entering GA. We don’t want users to incorrectly assume the security isn’t robust by attaching a Beta
label, but we do need to communicate limited availability and possible UX polishing. Traditionally, we use the term Early Access
for these cases.
Communicating about pre-release features
Whatever label above is being used, be sure to follow the communication guidelines for pre-release features.