1. Eppo makes experiment setup easy for engineers with metric collections
2. Eppo is powered off of our Snowflake Data Warehouse, making it easy to integrate complex metrics that require joins across multiple data types, or events that happen signficantly after experiment diversion.
3. Eppo made it easy to create new fact tables for diversion criteria, simplifying metric creation (we were able to key data off of both messages and tenants in minutes, something we struggled to get working with a competing product for weeks).
4. Eppo slices (split by) make it easy to investigate unexpected experiment impact or results
5. The UI is simpler and easier to understand than competing products, while also containing a more powerful feature set.
I've actively recommended Eppo to other startups and one of our VCs. Review collected by and hosted on G2.com.
Eppo has a few experiment measurement features we'd really like support for, and I'm sure they'll get to these soon:
1. Pre-post experiments (useful for binary rollouts, runtime environment ugprades, etc that can't be experiment flagged). We've creatively worked around this limitation for now by altering the data we feed Eppo.
2. Automatic experiment updates on flag value ramps. It would be great to see Eppo understand that an Experiment has been ramped (e.g. 10->20% diversion) and for it to automatically update the expected diversion % after that point. Eppo automatically detects variation distribution and messages in the UI that the results are no longer valid, it would be nice to have a mode where it automatically adjusted it's expected variation when it detects stable traffic variations. Review collected by and hosted on G2.com.
39 out of 40 Total Reviews for Eppo
Overall Review Sentiment for Eppo
Log in to view review sentiment.
Eppo has an excellent interface for showcasing experiment progress and results. Visually clean and easy to understand.
Also love their plug-and-play integration with Snowflake. We finished POC in 2 weeks!
They also have one of the best support teams. Always ready to answer questions and educate various team members from experimentation basics to advanced statistical methods.
Their tool supports companies just starting out with experimentation as well as more advanced methodologies. Review collected by and hosted on G2.com.
Nothing yet! We are lucky that we have a solid Data Engineering and BI Engineering team to help setup the metrics and database tables, otherwise, this could have taken longer. Review collected by and hosted on G2.com.
- Strong product: best in class stats engine
- Growth-minded and helpful team: providing timely support for questions and constantly incorporating customer feedback into the product
- Great documentation: good learning materials for both experimentation practioners and anyone new to ab tests
- Easy to grasp information architecture on the platform Review collected by and hosted on G2.com.
Nitpicks alert: Would love to
- Get adoption metrics for free with count metrics
- Have deep link into explore charts with slice and dice
- Be able to do pairwise comparison between variants more easily on the fly Review collected by and hosted on G2.com.
Makes complicated analysis simple and understandable, which uplevels our entire organization. Review collected by and hosted on G2.com.
Not all product work can be an experiment, so there's still additional work I need to do outside the product. Review collected by and hosted on G2.com.
Dedicated customer support team, responsive team, always be customer oriented and address customer needs. Review collected by and hosted on G2.com.
immature software development tool, reliable engineering support. Review collected by and hosted on G2.com.
We use Eppo at Delivery Hero for warehouse-native expriment evalutions (in tandem with our own feature flag management tool and sdk). With some analytics engineering investment in setting up a reliable metrics catalog, we have been able to dramatically reduce the overhead in experiment evaluations for our large product organization. Eppo offers an intuitive experiment setup and summary experience that has saved Product Analysts a lot of time and mundane overhead. An additional benefit is that it comes with all the latest statistical evaluations (and corrections, e.g. - when the randomisation and evaluation units do not match in an experiment) that allow us to move past fixed-horizon testing when it makes sense.
In addition, the team, starting right with the leadership, have a strong customer success mindset and are highly responsive on slack to all operational concerns. Review collected by and hosted on G2.com.
I don't have a whole lot to complain about, given we were very clear with what we looked Eppo to solve. So there has been no misaligned expectations. Review collected by and hosted on G2.com.
Eppo provides us with robust statistical analysis of A/B tests using modern methods that allow us to learn as fast as possible given the the effect size. It was easy to integrate. We were up and going in about 3 days Review collected by and hosted on G2.com.
Because it's warehouse native, it makes assumptions about the underlying data models in your warehouse. If they don't match those assumptions, then you need to do work to make everything play nicely together. Fortunately, they are reasonable assumptions and not hard to match if you don't already. Review collected by and hosted on G2.com.
Easy to add metrics and experiment.
Centralized management.
Support CUPED. Review collected by and hosted on G2.com.
Some functions not available.
For example, not able to compare between variants (can only compare variant with control).
Not able to run difference in difference. Review collected by and hosted on G2.com.
It's great to have the daily reporting as a result of Eppo - it has led to finding problems faster and getting more buy-in from stakeholders. The advanced analytics is also critical to making better decisions across our business. Review collected by and hosted on G2.com.
The updates are only daily, rather than real-time, given computing cost constraints, which limits some of the value. Also the fact that CUPED analysis isn't available (due to compute costs) on the explore graphs means that we sometimes present inconsistent data. Finally, three-way experiments could be better visualized (a vs. b vs. c, rather than a vs. b and a vs. c). Review collected by and hosted on G2.com.
Eppo has the ability to act as both as a data visualization of our test results but also as a record of truth of historical wins and losses Review collected by and hosted on G2.com.
Being able to only update results once a day, though that might be more of a limitation on our company's database's end Review collected by and hosted on G2.com.
I like the most that Eppo handles all of the calculations around the facts, so this makes the experimentation process available for all the company. And also brings consistency in the way metrics are computed Review collected by and hosted on G2.com.
There are a lot of definitions to implement in our warehouse, so i would expect to have cleaer guidelines for scaling the projects: good practices about table materialization, duplication of models (facts-entry points), naming conventions, etc.
Also best practices to integrate feature flags with the experiments, as usually feature flags are implemented by different teams that the ones actually following up on the experiments Review collected by and hosted on G2.com.
