Dataiku Reviews

My main use case for Dataiku is data science and AI projects.

We used Dataiku for a demand forecasting project where the objective is to forecast the demand for each product for the next three months.

The best features Dataiku offers include the ability for users to use the node without having to code and the functionality related to low-code/no-code.

Dataiku has positively impacted my organization by allowing non-technical users to adapt a data science project and to maintain a part of a data science project.

I think a pain point related to Dataiku is the visualization, which is not straightforward, and the integration, which is also not straightforward for non-technical users.

To improve Dataiku, the company could enhance the capabilities related to integration and visualization.

I have been using Dataiku for three years.

Dataiku is stable.

Dataiku's scalability can be better.

I have never tried Dataiku's customer support.

Before, we used a solution that I cannot mention, but the change is more related to using a more straightforward solution for non-technical users.

Before choosing Dataiku, I evaluated KNIME.

I have not seen any specific outcomes or metrics such as time saved, reduced costs, or improved project delivery.

I have not seen a return on investment with Dataiku in terms of time saved, money saved, or fewer employees needed.

I am not the person involved in the process regarding pricing, setup cost, and licensing.

My advice to others looking into using Dataiku is to use it principally to help and support non-technical users.

Dataiku is deployed in my organization on a public cloud on Amazon Web Services.

Amazon Web Services is our cloud provider.

I am not the person involved in the process of determining whether we purchased Dataiku through the AWS Marketplace.

My review rating for Dataiku is 7.

Public Cloud

Amazon Web Services (AWS)

We use the solution for data science and machine learning.

We were a team of six Dataiku scientists and one data engineer. We focused on fully leveraging Dataiku for all our data science-related tasks. This included data preparation, preprocessing, benchmarking machine learning algorithms, handling everything related to production, and making our algorithms available to stakeholders.

The advantage is that you can focus on machine learning while having access to what they call 'recipes.' These recipes allow me to preprocess and prepare data without writing any code. This saves a lot of time because I can quickly handle all the data preparation tasks and concentrate on building my machine learning algorithms.

One of the main challenges was collaboration. Developers typically use GitHub to push and manage code, but integrating GitHub with Dataiku was complicated. While it was theoretically possible to use GitHub with Dataiku, in practice, it was difficult to manage our code effectively and push it from Dataiku to GitHub.

Another limitation was its ability to handle different types of data. While Dataiku is powerful for working with structured data, like regular or geospatial data, it struggled with more complex data types such as text and image. In addition to the challenges with GitHub integration, the limited support for diverse data types was another feature lacking at that time.

I have been using Dataiku for over a year.

Since Dataiku relies on various open-source libraries and tools, updates or upgrades to these components can sometimes impact the stability of Dataiku's features. This can make it challenging to maintain consistent stability, as changes in the underlying open-source tools can affect how Dataiku functions.

I rate the stability as six out of ten.

There are some scalability issues.

I rate the scalability as seven out of ten.

Technical support was very good compared to other tools. We had access to chat and support.

Positive

The initial setup is very easy. It has many tutorials and many guidelines. After the initial deployment, it took about a week to manage all the setup and resolve various issues before we had a stable version of Dataiku that we could use consistently.

I rate it as eight out of ten, whereas ten is easy.

It is very expensive.

I wouldn't recommend using Dataiku if only one data scientist is on the team. However, having a larger team—let's say more than five data scientists—can be very helpful. Dataiku offers features that are especially useful when multiple people are working on the same project, and it also has tools that make it easier to move from the proof of concept stage to production.

Overall, I rate the solution as seven out of ten.

On-premises

My current client has Dataiku. We do sentiment analysis and some small large language models right now. We use Dataiku as a Jupyter Notebook. 

We use it a lot for marketing and analytics. The marketing and sales team uses Dataiku.

It's got good feature selection and creation of feature stores, and it also gives different aspects of modeling approaches. There are a lot of similarities with DataRobot. 

So feature selection, different modeling, and financial metrics are good aspects. 

The no-code/low-code aspect, where DataRobot doesn't need much coding at all. 

Dataiku still needs some coding, and that could be a difference where business data scientists would go for DataRobot more than Dataiku because you still have to code and use either Python or R, or Scala. However, with DataRobot, you don't have to do that at all.

I've used Dataiku for about four years.

The company is based in France. But they're more and more in America as well.

So, the support was okay.

Positive

I used DataRobot. Dataiku has a different kind of structure to it. It's not financially heavy like DataRobot, which caters more to financial companies, like banks. Dataiku doesn't have that yet. I think they are also working on that area. But yeah, there are some key differences between the two products.

DataRobot has an additional feature with financial firms that it creates all these financial metrics when you run a time series analysis. Those things I have not seen in Dataiku.

If any financial company is choosing between DataRobot and Dataiku, they will definitely go for DataRobot because it creates all these financial metrics. It creates deltas, time series, time difference fields, and things like that. So, that is an added feature that DataRobot has.

Pricing is pretty steep. Dataiku is also not that cheap. It depends on the client and how much they want to spend towards a tool.

Overall, I would rate it an eight out of ten, except for some coding things that are there, which some people may not want to do, like certain business data scientists.

Dataiku is good for multiple teams' collaboration. If many teams are collaborating and sharing Jupyter notebooks, it's very useful. It has a good data processing structure and includes most of the models. I haven't checked the large language models in it yet, but it's a pretty good tool. It does well with analytics and has a sound structure on the back end. 

Some coding aspects are necessary, but it generates SQL code, which is an added feature. A lot of data engineers like Dataiku because it generates SQL code on the right side.

The use case is data science, and we've deployed Data Science Studio in multiple regions for four environments: dev, preset, pre-production, and production.

The solution responds to the requirements of the business team.

Data Science Studio's data science model is very useful.

I think it would help if Data Science Studio added some more features and improved the data model.

We've been using Data Science Studio for three years.

Data Science Studio is stable.

Data Science Studio is horizontally scalable.

Setting up Data Science Studio was simple, and deployment took about three months. You only need around three people to deploy and maintain.

I rate Dataiku Data Science Studio nine out of 10. 

On-premises

The best feature is the user interface. It allows us to see the visual flows. 

The intuitiveness of the software itself is very good. 

It has both click or code flexibility, which is quite useful. I already have coding knowledge, so I go for coding as my go-to method of interaction. However, for a normal person that may not have a background in coding, the visual lists are really great. It makes it easy for non-IT individuals to navigate everything.

The pricing is very good.

The solution is quite stable.

You can scale the solution well.

When the flows get complex, there are too many data sets on them. Normal users will get confused by the influx of information. This was a problem up until recently, however, they have since resolved it. Now, there's a feature called Zones(introduced in version 8) that allows users to collapse multiple flows into a single flow. It allows everyone to easily carry on with their work.

Other than that issue, that has now been resolved, there isn't anything lacking by way of features.

The interface for the web app can be a bit difficult. It needs to have better capabilities, at least for developers who like to code. This is due to the fact that everything is enabled in a single window with different tabs. For them to actually develop and do the concurrent testing that needs to be done, it takes a bit of time. That is one improvement that I would like to see - from a web app developer perspective.

I've been using the solution for more than a year at this point.

The solution is pretty stable. Since ours is a shared platform, the solution does freeze sometimes when a complex model is training or if you keep too many notebooks open the in-memory after editing. Though the usage has been optimised by admin end using resource control. Clearing notebook memory and job logs through in-built macros keep the solution stable.

The solution is scalable. It has the power to integrate the cloud platform. This makes scalability both possible and easy. 

We have ~50 developers/users working with the solution right now.

Technical support is really, really good. It's excellent, actually. If you click on the "get help" button and you mark it as urgent, you get a reply from the CTO himself. I find that most of my conversations are direct with the CTO, and that's due to the fact that most of the time my questions are quite complex.

They're very knowledgeable and are able to assist us very well. They always reply very promptly. We're quite satisfied with their level of support.

I'd rate it overall four out of five from a solutions perspective.

I haven't used another solution. This is my first as we are just growing this space within our organization.

I didn't handle the initial setup of the solution. Therefore, I can't speak to how difficult of complex the setup was.

This solution is cheaper than other competitors. It's less expensive, for example, than Alteryx or any other product.

While I started using version 5.1 of the solution, I've currently updated it to 7.0.1.

I would recommend the solution. It's affordable and user-friendly.

Overall, I'd rate it nine out of ten. I'd rate it higher, however, I don't have enough experience with other similar solutions, therefore, it's hard to compare.

On-premises

Our primary uses for this solution are data preparation and data modeling.

We have a testing environment, a production environment, and two API nodes.

Using Dataiku has meant that we spend less time on preparing and cleaning data, and we spend less time on blending models together. Ultimately, it means that we can spend more time modeling. 

The most valuable feature is the set of visual data preparation tools. 

The solution supports code from different languages including Python and R. Whatever code you want to use, it works well.

This solution allows us to store and retrieve data directly into BigQuery.

The documentation and tutorials are quite good.

From an administrative point of view, I would like to be able to communicate with the users who are logged into the system. For example, I would like to be able to send a broadcast message that says "I am shutting down the system."

I would like to see more organization and better cohesion within the tool.

In the next release of this solution, I would like to see deep learning better integrated into the tool and not simply an extension or plugin.

I would like to have a better way to manage images and sound.

The error messages are not self explanatory and can sometimes be difficult to understand.

I have been using this solution for one year.

The stability is quite good.

This is a scalable solution where we integrate with BigQuery for storing and retrieving our data. There are only two of us in the company who use this solution, although we would like to increase our usage.

The technical support is quite good. We have had to open a few tickets and they replied in just a few minutes. They are quick and supportive.

I still use several other solutions for data science including RapidMiner 9 and Weka. I have also been using Octave and MATLAB for modeling.

I use the Community Editions for these other products, so I have restrictions when it comes to things like the size of the dataset. When I need to be free of restrictions then I use Dataiku Data Science Studio.

The initial setup is very, very simple.

To deploy the entire platform will take one or two days.

We handled the deployment ourselves. We can work totally independently.

The annual licensing fees are approximately €20 ($22 USD) per key for the basic version and €40 ($44 USD) per key for the version with everything.

At the moment, we haven't had any need to use containers or Spark because everything is included in BigQuery.

My advice for anybody who is implementing this solution is to start with having somebody who can mentor you. Whether this is the case or not, the tutorial and documentation are quite good, so I would suggest going through the whole tutorial and academy material.

This solution does have a learning curve, although it is not steep.

I would rate this solution an eight out of ten.

On-premises

I just use the product for data migration. Once a month, I push data from one Redshift data warehouse to another Redshift data warehouse. I do that by using a simple SQL query.

I like the interface, which is probably my favorite part of the solution. It is really user-friendly. I might think that it is user-friendly because I'm in IT and it seems familiar to me. It is colorful and I think it is really beautiful and well-designed.

I think the interface is very nice, but for somebody who is not as familiar with IT as I am, it may be much more difficult for them. It is nice for me because I'm familiar with this type of software that falls in the realm of the data science platform. I can see how a client who really doesn't know anything about IT or computers might try to use it and find that it would be a little difficult to access some features. That type of user may really need training in order to work with Dataiku. So, in the next release of Dataiku DSS (Data Science Studio), they should make it more friendly for everybody to use, not just IT people. 

For me, I find that it is a little slow during use. When I use Dataiku to run my script to transfer data, it takes more time than I would expect for the operation to complete.

I have been using this solution for a year.

I have not done a lot of exploration into Dataiku and its other features as I only use it for a particular task. But from my experiences and from the review I'm seeing online, it is a stable solution.

I think there is some room for improvement in scalability as I already find it performs a little slowly. In my company, there are three of us. There is an IT manager, there is the data warehouse manager, and there is me. In the company that we use it for, there are more than 800 employees.

I have contacted customer service before. They respond quickly, so that is a plus.

We did not use other solutions before switching so much as we are using a few different solutions together. We use Knime and Alteryx for data science and analytics. I just use Dataiku for data migration for now. I just started using Knime, but I'm most familiar with Alteryx because I have used it for one-and-a-half-years to practice ATL (Active Template Library) on my data. Alteryx is somewhat the same as Knime, but it is more user-friendly.

The company I work for did the setup for me. It was already set up when I came.

We are consultants so we do the implementations ourselves. Another consultant introduced Dataiku to us initially. I guess they really appreciated the solution.

Dataiku is a very broad solution that offers many possibilities. If you want to use it you must be fully committed to it. 

The biggest lesson I learned from using the product is that you can do many things with it. But you must commit the time to discover the tool.

On a scale from one to ten where one is the worst and ten is the best, I would rate Dataiku as a seven. It is a little bit of a conservative rating because it is a nice solution and I just use it for a particular task.

Private Cloud

Amazon Web Services (AWS)

I use this solution to make predictions from time-series data. I am a consultant and operate this solution for my clients.

The most valuable feature of this solution is that it is one tool that can do everything, and you have the ability to very easily push your design to prediction. 

I would like to have better exclusion of data capability.

The ability to have charts right from the explorer would be an improvement.

I would like to see additions to the architecture for specific business use cases.

I have been using this solution for two years.

This is a stable solution. I have not seen any bugs or glitches.

This platform is easy to scale. You can take full advantage of your office architecture.

My client has more than twenty people using this solution.

I have been in contact with technical support through the website.  They have chatbots that dispatch tickets to specific people, depending on the problem. They respond quickly and I am satisfied with the support.

The initial setup of this solution is straightforward.

I performed the implementation myself. By reading the documentation, it is simple.

My advice to anybody who is implementing this solution is to use the tutorial first. There are lots of tutorials available that help to quickly explain the solution.

This is a product that I recommend.

I would rate this solution an eight out of ten.

On-premises