Data Intelligence Platform
An integrated Data Intelligence solution, which enables business people to ingest data, build their own pipelines through a simple graphical UI, monitor changes, and wrangle tables through machine learning algorithms.
WHAT
Web app
WHERE
Oracle
WHEN
2017
​
Project duration
6 months
Project Overview
There were multiple reorganizations within the company and many products, which served a similar purpose or addressed the same target audience were about to merge into one. DFML (Data Flow Machine Learning), BDP (Big Data Preparation) and BDD (Big Data Discovery) all addressed business people with a purpose to explore patterns in big data. Data Intelligence Platform was a fresh start with a lot of existing pieces.
-
I was responsible to put together a high fidelity prototype with my team. I have orchestrated the effort as one of the most experienced Axure experts
-
I designed search tailored to this specific product framework and influenced many areas like explore, prepare and pipelines.
-
I worked together with the SaaS team from long before this project to bring this product to the Alta Standard platform
My Contribution
Challenges
-
I had no manager title, so one challenge was to bring all the designers together and orchestrate their work, without being titled to do that
-
because of the high expectations, multiple C-level people wanted to influence design into the deepest details, sometimes against each other, without understanding design disciplines
-
we had to design in high fidelity (which is inefficient and not really possible)​
Addressing those
The easiest part was my colleague I was sitting with. We already worked together and there were a strong trust and respect for each other's work. He focused on data ingest, I did for the rest. A bit more challenging was to work with other designers never worked with Axure and were not interested in knowing the whole picture and sitting on the other side of the continent. On that part, I had to take the extra mile to technically fit their work and explain the necessary parts of the picture to keep them up. By that time I had a nice working relationship with the product managers (PM), but they had competing concepts against each other (each and every product component had a PM). It took a lot of meetings to bring them to conclusions. My approach was to get involved with every conversation to represent a strong usability vision. I am not saying I haven't lost battles. For example, I had to implement kabob menus to each and every tile, even if that sounded ridiculous from the beginning. After not being able to convince one PM about that concept I just made that for him and work on a different area with another PM to implement the better solution in a different area. After I made both functioning I just had to let them play with each others area and see which one was better. This way we could mostly come to a consistent solution for both areas.
We did everything in high fidelity. I knew the pitfall of that from the beginning: it could take the conversations sideways and it did. I understood the real high-pressure C-level people had to deal with, but we were short on UI designers and simply pleasing every request they had on icons and labor-intensive details would take too much time from architectural and usability work. The method worked the most was to accept their point that there was a huge potential to make things better on UI but bring an actual architectural, or workflow challenge to them, to let them understand that the detail caught up their attention would not kill a concept, but a workflow glitch could. If that did not work the other trick I had to work with smart components. I built everything with repeaters, masters, and libraries, so if someone asked a change, which influenced multiple pages I just had to change in one place and that spread everywhere. I faced 2 big challenge with the smart components:
-
I had to redo the photoshopped artifacts and translate those to smart objects (which was extra work for me, but it worth)
-
In the end, I was the only one who could follow the complex system of variables, masters. Because of the fast pace, I could not bring every designer to the same page. It worked, but in the long run, it was not a sustainable solution.
I had no chance to hire designers or bring every involved designer to the same page or teach them, so in some areas of the prototype, I had to change the approach and have dumb pictures, to be able to leverage the workforce we had. Those areas could not follow the changing directions that fast but harnessing more designer's work was still more beneficial to me and motivating for them. I lost the flexibility and speed in some areas, but I could be part of the PM decision making progress to influence and leverage everyone who was available.
You can see some example pictures below from the work we did:
I want to highlight how extensive is this prototype. I have used a real dataset and some of the preparation functions really do their job, create new columns with real data. The filtering and search are also real, even if it does not support every word (in my future prototypes those work as well). One can pick data set, or pipeline or connection and the names will carry forward. If you are interested you can play with this. If you want to see the smart elements I recommend picking the "Prepare Data Set".
End Result
This project was a great challenge to work in high fidelity from the beginning. People loved it because it looked beautiful, but it was not efficient because of the hi-fi (people spent too much time on unimportant details, like icons, when major architectural decisions were in flux). On the other hand, it was a pretty successful lean approach to answer a business question. Developers made only POC code pieces, but no real dev work happened towards the product, which saved approximately 5M$ for the company.