https://www.growthbook.io/ logo
Join Slack
Powered by
Hi Everyone! I was just evaluating GrowthBook and ...
# ask-questions
c
Hi Everyone! I was just evaluating GrowthBook and have a couple of questions: 1. Is there any support for edge experimentation i.e. integration with Fastly Compute@Edge? 2. Is there an option to configure some JSON or other payload for individual buckets? Thanks very much!
f
1. Our Javascript SDK should work for edge functions 2. We support both boolean (on/off) features as well as number/string/json data types. So you could make the value of a feature a JSON object and have different variations have different JSON objects
c
Thanks Jeremy! I meant something like this: https://github.com/optimizely/fastly-compute-starter-kit
Our use case requires that we don't make a call from Fastly to Experimentation Platform in order to get the assignment. We need to make this assignment at the edge to save some latency.
f
Our SDKs don't make network requests so that should be fine. Similar to Optimizely, you would fetch and cache the feature definitions from GrowthBook and pass those into the SDK. Then, all evaluations would happen locally.
c
Aah, that's great to know. Thanks a lot! 🙂
f
Thanks for that Optimizely link. We don't have a starter template like that, but seems like it would be a good idea to add
c
My pleasure! Thanks again
We support both boolean (on/off) features as well as number/string/json data types. So you could make the value of a feature a JSON object and have different variations have different JSON objects
Sorry, I meant for running an experiment, can we add a JSON payload to each variants?
f
It's a little confusing, but the experiments section in our app is just for analyzing results. The experiments are actually implemented and run through feature flags and our SDKs.
c
Aah, I see. That's why I was wondering why I can't adjust the allocation to buckets here. Thanks! I will go over the doc more thoroughly now.
Sorry to bother again, I have one more question. Is there a way to prioritize different experiments i.e. if you are running more than one experiment on the same web page, is there a way to ensure which experiment takes priority? Also, I wasn't able to find out docs about support for MVT.
f
The way we handle mutually exclusive experiments is with Namespaces. https://docs.growthbook.io/app/features#namespaces It's not quite the same as setting a priority, but is safer and cleaner when it comes to data quality and the statistical analysis
c
Awesome! Thanks a lot! 🙂
f
For MVT, if your feature has a number, string, or JSON value type instead of just ON/OFF, we let you add as many variations as you want. We don't have built-in support yet to look at interaction effects between different variables though if that's what you're looking for.
For example, you could make a 4-way test with the variations: • Blue / Top • Blue / Bottom • Green / Top • Green / Bottom And we can tell you if any of those combinations beat the control. But we can't tell you just the effect that color on it's own had for example.
c
Got it, very good to know!
I also read that since the assignment happens on the client side, there is no need to store user data on database, which is very cool! however, how does the analytics handle it, if there is no assignment/user data available?
f
That's maybe slightly misleading. You do still need to track the assignment event in your analytics system or data warehouse in order to analyze results
c
cool, got it. thanks!
I am so sorry, please don't kill me. What if we want to do a server-side experiment? Is there any sdk for that? Or do we just use the API endpoints?
51 Views
Open in Slack
PreviousNext
Powered by