Hey guys! Is there any way to specify >specific_app_version: string in a targeting condition? We have some features that require users to have the latest native code and we keep appending the latest app version string to 5 or 6 features

Hi, are there any official instructions about deploying Growhtbook to kubernetes?

👋 Hola Growthbook team! We are testing the Visual Editor and we had some problems setting it up. They were all about CSP rules, I’m posting the rules we had to include here just so that you are aware other people might trip on them: • We needed to allow cdn.growthbook.io to the

script-src

policy (this one is duh 😅). • We needed to allow unpkg.com to the

script-src

policy (there is only one script served from there:

ab-designer@0.6.0

). This one is tricky, and I’d expect people not wanting to add a global CDN to their allow list as that’d render the policy useless. This is a blocker on our end. No idea what the best solution for this would be but I’d personally recommend hosting this script in your own domain if possible. • We needed to include a

frame-ancestors

policy to allow app.growthbook.io to embed an iframe pointing to our site. This one is pretty obvious but I was surprised not to see any reference to it in the docs. I hope this helps improving this new feature and I’m happy to share any other details you might need to trace these problems down

Hey, I'm trying to work with relative dates, so like apply rule for users that signed up before a month ago being dynamic, so in the next day, so when users reach 1 month, have the rule enabled. So I created the following query:

{
  "$expr": {
    "$lte": [
      "$createdAt",
      {
        "$subtract": [
          "$$NOW",
          2592000000
        ]
      }
    ]
  }
}

But I saw the operations are done manually in the code, I thought there was a mongo memory server running under the hood to allow do any mongo query. And there isn't support for the

$expr

operator (it does work outside aggregations). Is there any plan to support relative date to do something like this? E.g

$$NOW - number

Is there no way to configure the suspicious result flag on GB cloud? If not, it'd be good if there was a "view anyway" type button We're a small and early company, so it's very common that changes we make cause >50% uplift. Often we know it'll be a huge increase, but we're using GB so we can properly analyse how much

Hello! Have a problem with traffic split configuration in experiments. My configuration in the Growthbook is 100% traffic, 50/20/5/25 split. I created script to make 100 requests with 100ms delay between each and i got result like this:

{0: 29, 1: 35, 2: 18, 3: 18}

How to achieve my configured split for traffic?

Is there a reason that the conversion window of a Metric is defined on a per Metric basis rather than a per Experiment basis?

@future-teacher-7046 When defining a dimension column name in an Assignment Query GrowthBook should appropriately ‘quote’ the value.

Hey guys, had a minor mess up with a JSON feature flag and it turned out my JSON was invalid. A GB client just got a null value instead of the default. Perhaps adding quick validation there?

@future-teacher-7046 I wanted to chat about how should I go about a comment you left in the design doc

Another “nice to have” feature is support for projects. You can create multiple projects in GrowthBook and scope both the webhooks and API endpoint to only include features in a specific project. For example, you might have front-end and back-end features in separate projects and want to have a separate features JSON for each. This proxy server should ideally be able to cache and serve both.

I'm not sure if we already have an API to actually separate based on project. But if we wanted to have it as a config where endpoint depends on what api key you have, it'd work. Current approach I'm taking is that this service takes a config that's API Key and keeps the data in it and serves when asked for. And in next version, I was thinking about adding websocket support 🙂

Hey! Would be awesome if the guardrail metrics in the experiments also had links straight to the definition/sql like the regular metrics have 😊

Another idea for something that could be useful to several. Would it be possible to add tags and projects as dimensions of the “Experiments by month”-graph? We have team names as tags, and would be nice to easily see who are “the best” at running experiments 🙂 But a nice to have!

btw it'd be good if there was a way to express "time since" in the targeting conditions We have an ISO8601 string indicating when the user account was created, and we can already use string comparisons to compare with a fixed date But it'd be extremely helpful if we could express e.g. "`mojo_user_created_at` is at least 7 days ago" - without having to provide both absolute + relative measures for every date attribute

When using an activation metric, bigquery takes a very long time to compute the results. The actual 21 days I'm currently analyzing I stopped after 45 minutes. Usually a sign that a join is exploding. When leaving out the activation metric, All 7 queries (7 metrics in the test) take less than a minute. Ran some manual timings, limiting to 1 day, 1 metric. The query takes 4+ minutes. Diagnostics, subquery outputs (count(*) from ...) step by step •

__rawExperiment

is ~4 mio rows (1 day) •

_experiment

comes in at 2 mio rows •

__metric

at 1.8 •

__activationMetric0

at 16K •

__activatedUsers

1.2 mio •

__distinctUsers

- takes about 46 seconds, results in 7K rows •

__userMetric

- takes 4 minutes something, results in 7K rows There is the bottleneck. Query details report 11.2 mio rows read. at some stage Not sure how/why this is such a resource hog, maybe you can spot something.

__userMetric as (
  -- Add in the aggregate metric value for each user
  SELECT
    d.variation,
    d.dimension,
    LEAST(10, SUM(m.value)) as value
  FROM
    __distinctUsers d
    JOIN __metric m ON (m.user_pseudo_id = d.user_pseudo_id)
  WHERE
    m.conversion_start >= d.conversion_start
    AND m.conversion_start <= d.conversion_end
  GROUP BY
    variation,
    dimension,
    d.user_pseudo_id
)

Hi, my team for now still used flutter SDK 2.8.1 (need to upgrade Flutter 3 soon). But, I see growthbook package starting with Flutter 3.0.0 support. If you are interested to backward compatibility in 2.8.1 please see my PR in forked github repository. Thanks.

Maybe this will be other feedback, I see this package dart version not starting in 2.17.0 that compatible Flutter 3.0.0 when released. I just create PR to downgrade dart version to 2.17.0 to make compatibility at Flutter 3.0.0 as minimum sdk.

Hey! I had an idea. I was wondering if there would be a way to have (our own) standard template automatically appear in the results description section when ending an experiment? Was thinking something like this (see picture) and that we would be able to edit the template if we see a need for a change. Not sure how easy something like that would be to implement though. This would also be super to have in the experiment description section, so we can advice our experimenters on what information they should include also there 😊

Hi GrowthBook team, hope you're doing great 🙂 We wanted to start monitoring usage for our GB users and it would be really helpful to have the date creation of the account and the date creation of the metrics. Any chance this is something you plan to add (in the team and metrics pages)? Thank you very much! CC: @stocky-nest-87172

Just to flag, there's some issues with how growthbook-react works, and how you document its use - due to it being mutable Running

growthbook.setFeatures

,

setAttributes

, etc do not trigger a re-render (because the context provider's value doesn't change), and there's nothing in the hooks etc to compensate for that Currently the only safe way to use growthbook with react is to treat the GrowthBook class as being immutable - any other way, it's fundamentally broken

Hey all, just a quick feedback, the URL targeting in the visual editor is so strict for no reason. e.g. putting the url as

<https://example.com/>

won't work if you access

<https://www.example.com/>

one with the

www.

same with

http

and

https

I think it's pretty unnecessary and lost me 2 hours debugging 😅

Hey all, another potential bug, the URL Targeting in the visual editor is causing issues, e.g. if you set the URL Targeting to

<https://www.example.com/>

accessing the

<https://www.example.com/abc123>

sub page will also still trigger the experiment. & it is not an SPA page, it's a next.js SSR page! so the

window.location.href

is accurate

Hey! Loving that we now can create ratio metrics where the analysis unit <> randomization unit. We have a lot of use cases for this, so thanks! 🙌 🙌 I have created a couple of them, and they seem to calculate the rate ok, but the violin plot and Chance to beat control calculation is off (see picture). Is this the case for others too, or is it just with us? Both the nominator and denominator work well by itself.

Love GrowthBook, here with our first issue! On a data-heavy experiment with two primary metrics and three guardrail metrics, we have two failing queries:

500: Code: 241. DB::Exception: Memory limit (total) exceeded: would use 7.46 GiB (attempt to allocate chunk of 0 bytes), maximum: 7.00 GiB: While executing AggregatingTransform. (MEMORY_LIMIT_EXCEEDED) (version 22.3.8.40.altinitystable (altinity build))

These failing queries run quickly in ClickHouse itself. The difference appears to be that GrowthBook runs the queries in parallel. Solutions to our problem (beyond increasing RAM) are something like: 1. Not running GrowthBook queries in parallel 2. Allowing some way to manually update guardrail metrics (so we can run them one at a time ourselves) Happy to provide more info if needed!

Hey Growthbook team! I found a UI bug in the experiments tool where the chart seems to have a misalignment between the highlight-able data points and the actual graph being drawn

Hello! Would love to have a semvar rule in targeting conditions.

is there a way to get access to the experiment names set in UI via one of the tracking callbacks? in

onFeatureUsage

or

trackingCallback

we can access the ids, but not the names

I see that recently an PR was merged into GrowthBook which adds the bones of a REST API (

/api/v1/features

). Would it be possible to make this API OpenAPI compatible so that SDKs are easy to build with tools like https://github.com/openapi-generators/openapi-python-client?

This would then make it easier to write ChatOps tooling which wraps these SDKs

Yeah, we definitely want to support OpenAPI going forward. I added a GitHub issue to track this. https://github.com/growthbook/growthbook/issues/675

Not sure if it’s been out for a while but I just wanted to thank you guys for the saved groups feature. Has already saved me time. Really appreciate the work you’re doing!