Hey! When using percentile capping on ratio metrics, it's not the value metric that is capped but rather the numerator (leaving the denominator untouched). This is not what I expected and I don't think this makes sense because for ratio metrics we care about the value of the metric and not the value of the numerator in isolation. My concrete example is looking at gross margins where I'm experiencing weird results which seems to be related to outliers. However, adding a percentile cap doesn't help since that just lowers the numerator making the % gross margin into a unrealistically low number which in turn will make it difficult to explain the data to others. Hope this made sense and let me know if I should clarify anything!

Hi y'all, I'm new to 4.2 and had a question about retention metrics. I noticed when calculating a >=7 day retention metric for an experiment, the numerator correctly calculates the users who return after 7 days, but the denominator is still including all users. Is there a way I can limit the denominator to only include users with at least 7 days in the test? This also holds true for row filters, as I would like to apply row filters to both the numerator and denominator. Thanks!

Hey folks, I'm trying to look into experiment result with certain dimension - however, upon looking at the helper doc, I could not find this UI on my end from the result page. May I get some help for pointers? Thanks so much!

What configuration should I use to build the Dockerfile locally and on the server pod? Specifically, how much memory and how many CPU cores are recommended? I want to build the GrowthBook image on the pod.

Hello, in the main results page of an experiment, we used to be able to filter by any dimension (even unit dimensions) and we would be able to see the distribution of the dimension value by the variants. This is very useful to check that the 2 variants are equally distributed. We were also able to check that in the dashboards, now it's not available in either places. Any way to obtain this info?

Question: Does the (python) SDK support async hooks for tracking experiments? • the python SDK docs say an async method is supported (

_async_ _def_ on_experiment_viewed(...)

) • the python source code doesn't seem to have any async behavior or types, seems to just run the method (and the README itself doesn't mention the

async def

option). Is this a docs discrepancy?

Hey team - quick question: Is there a way to make a custom report the default view on the Results tab (instead of the default goal metrics)? If not, is there a recommended workaround?

Hey team, I have a question regarding API keys. We have old API keys in our account per environment (created more than a year ago). API keys are only visible when looking at the organization response but not in the API keys section. I’m trying to create a new API key for a new environment I created and it seems that’s no longer an option, as it wouldn’t let me choose an environment when setting up the API key. I can see that the new API key is created in the list of apiKeys and that indeed the environment field is empty. I wonder if this is a result of a old functionality that was removed from the platform? Screenshot attached

hi, i sent security bug to security@growthbook.io and got email response, not sure when it will be processed.

Hey team, I have my dev environment connected to GrowthBook API, however the prod account is not connected. Confirming the prod environmental variables are set. Any idea on how to resolve the issue?

Hi all, I have a question about the metrics information. When I view a metric (Metrics and Data → Metrics), the data only goes up to the last time I manually clicked “Run analysis.” Is this the intended behavior, or should the analysis run automatically—for example, once per day? Also, does this impact experiments that rely on these metrics? Will an experiment continue to receive metric data even if I don’t run the analysis manually every day? I can see that the data in BigQuery is updated daily.

anyone? 🙂 https://growthbookusers.slack.com/archives/C01T6Q1SVFV/p1765099495467029

Hello, is someone else facing this issue on a self-hosted GB 4.2?

Hi! I would like to discuss how can I buy pro features such as cuped etc for my team, but i need to self host it. Now we host GB using our hardware and we can't move in a cloud

Is there docs for: • the roadmap for Growthbook development and direction? • guidance for contributing changes • issues tracked beyond GitHub? (More specifically: we had minor friction with the Python SDK that I was able to workaround; I'm curious if it would be worth contributing upstream, and how to initiate that conversation.)

I noticed Pro plan is missing on self-host. Is there any indicative pricing of Pro features mentioned under cloud but for self-host?

Hello everyone, I’m trying to contact sales team via emailing but nobody has replied yet (almost 2 weeks). Can anyone help?

Hi folks, I have a self-hosted enterprise version of Growthbook and for some reason all images we attach to experiments and comments are not displaying. They look alright when we first attach them, but a few moments later they get unavailable. Any idea of what might be happening?

Using SKD is it possible to know if the flag is a "force" or "percentage rollout" ? I am looking on golang sdk, and I can see in both cases result source is set to

ForceResultSource

Would it be possible to create a PR to add a

percentage rollout

type? Use case - I would like to track how many users were evaluated by given rule - but I want to measure it only for percentage rollout

Hi, I'm trying to setup GCS for File Uploads. Added env variables, then tried to upload: •

/upload/signed-url-for-upload

returns 200 with 4 keys: signedUrl, fileUrl, filePath, expiresAt • then, request

<https://storage.googleapis.com/$MY_BUCKET/$MY_ORG/$MY_MONTH/$IMAGE?GoogleAccessId=$MY_ACCOUNT&Expires=$MY_EXP&Signature=$MY_SIGNATURE>

fails due to CORS error I tried to execute 2nd request via curl, and it uploaded empty file (it's ok), so the credentials are correct

Hello has anyone setup shopify with growthbook while using a custom pixel? For some reason I can't seem to get experiment_viewed to fire. Any help would be appreciated. I've tried the shopify SDK, the script SDK, I've tried sending the experiement_viewed trigger from multiple different places (It is firing from one in the datalayer but doesn't seem to be firing the event in GTM - Thinking it might be too early) any help would be appreciated.

hello, is there an API endpoint we can hit to get an experiment's results and data, like how many experiment views each variation has, or performance with metrics and stuff?

Hi, Need some understanding regarding a issue we faced recently had. We have pro plan and recently we integrated growthbook SDK and it seems to be working fine in the beginning but later worker started throwing 499 for some request. Upon investigating we found that the wall time of worker increased drastically. Which correlates with 499 error count.

Hi! Anyone here using Athena as a Data Source and getting the "Region is missing" error message? It was working correctly, we setup a fact table, some metrics and a new experiment and all queries were returning correctly. Then suddenly we started getting this error.

Hey guys. Has anyone here tried to integrate growth book sdk in Spring + Kotlin server? The example on official documentation does not look thread safe https://docs.growthbook.io/lib/kotlin-jvm#spring-boot

@Configuration
class GrowthBookConfig {

    @Bean
    suspend fun growthBook(): GrowthBookSDK {
        val growthBook = GBSDKBuilder(
            apiKey = System.getenv("GROWTHBOOK_API_KEY"),
            hostURL = "<https://cdn.growthbook.io/>",
            attributes = mapOf("environment" to "production"),
            networkDispatcher = NetworkDispatcherOkHttp()
        ).initialize()

        growthBook.refreshCache()
        return growthBook
    }
}

@RestController
@RequestMapping("/api")
class FeatureController(private val growthBook: GrowthBookSDK) {

    @GetMapping("/features/{userId}")
    suspend fun getUserFeatures(@PathVariable userId: String): Map<String, Any?> {
        // Set user-specific attributes
        growthBook.setAttributes(mapOf(
            "id" to userId,
            "environment" to "production"
        ))

        return mapOf(
            "newDashboard" to growthBook.feature("new-dashboard").on,
            "maxItems" to growthBook.feature("max-items").value,
            "premiumFeatures" to growthBook.feature("premium-features").on
        )
    }

Hello there. We've been noticing some peak in our p99 latency. We've been trying to identify the cause, but initial checks (CPU, memory, MongoDB latency, etc.) haven't revealed anything. After reading the self-hosted best practices, I suspect the stats engine might be having an impact. The documentation suggests splitting the engine if an impact is observed. However, I'd like to confirm if the stats engine is indeed the source of the issue. Is it possible to monitor it in isolation or something?

Hi, I'm just starting usage of growthbook and we have single mobile app but 2 different websites (separate brands, that in mobile app live together but on web have their own domains as eCommerce shops). Currently I assume having single project in GB and using user attribute "brand" is best solution but maybe somebody already went through such challange?

Hi, I have setup a growthbook experiment that is linked to feature

X

and also has attribute targeting setup, say attribute

a1=value1

. During the onboarding of a user into our app, the user is presented with a series of screens. One of the screens is where we set the users attribute

a1=value1

by calling

await growthbook.setAttributes(growthbookAttributes);

. Then, after a few more screens, we access the growthbook feature

X

. At what point would the experiment be started for this user? Would it be as soon as we have set their attributes or would it be only once we have accessed feature

X

in our code?

Hi everybody! I’m looking for someone who can do our Growthbook implementation for us. If you offer this as a service, please do not hesitate to reach out😁🙌

Quick question about whether it's possible to convert the Bayesian ChanceToBeat to a frequentist p-value. The robots are saying "you can informally think of Chance to Beat Control as roughly equivalent to 1 minus the p-value." I'm somewhat skeptical of that conversion. But is it true-ish? 🙂 cc @helpful-application-7107 @strong-mouse-55694 ps. I'm building an API based tool to scan past experiments, looking for replication worthy signals. https://www.linkedin.com/posts/jlinowski_experimentation-replication-optimization-activity-7405330721030041600-KeYL/ if anyone is interested in giving it a spin, ping me.