We just started using GrowthBook for Features/ab tests, and our having trouble splitting users. We are using Segment's anonymous id to identify our users. We created an A/B experiment under a feature and split control and variant by 50/50. However, users are getting tagged more with the control almost 88% compared to 11% for the variant. Has anyone run into this problem before? Any help or advice would be most appreciated, thanks!

that is not right

Sure! Here is the code we have for loading growthbook and experiments

const growthbookInstance = new GrowthBook({
    apiHost: process.env.NUXT_ENV_GROWTHBOOK_URL,
    clientKey: process.env.NUXT_ENV_GROWTHBOOK_KEY,
    attributes: {
      id: segmentData, // Segment anon id 
    },
    enableDevMode: isDev,
  });

  await growthbookInstance.loadFeatures({ autoRefresh: true });

  const expVariant = growthbookInstance.getFeatureValue('exp_example', '0');

you are not using the trackingCallback?

We are not using trackingCallback. Instead after the experiment loads we pass that to our analytics since we have other experiments running with Optimize still. Thank you for sharing

evalFeature

. I will add that and see if that helps to debug.

I have a feeling that without using the trackingCallback, you’re getting users who are getting a specific value without being included in the experiment

Thank you! I refactored to use

trackingCallback

to send to our analytics and it seems to be splitting more evenly now.