HIVERY Curate is a
web app that uses ML and automation for manual processes; innovative solutions in B2B Category Management & Retail industries.
Curate is HIVERY's software tool used in the retail and category management sector in big superstores like Walmart and Target in the USA. HIVERY aims to change the norm in this industry by automating what is traditionally a very manual process to an automated platform achieved a platform/software tool that is led machine learning algorithms.
UX's involvement in Curate is complicated as it is heavily dependant on algorithmic development. That said the behavioural change that comes along with this innovative platform makes the UX portion of Curate particularly interesting. As a designer, working at Hivery required solving various constraints and challenges.
The challenges
User friction
The first and major problem to overcome was that retail management has always been a manual process with tools that have been existing in the industry for the long term.
The first and major problem to overcome was that retail management has always been a manual process with tools that have been existing in the industry for the long term.
Through Curate we needed to both address the industry norms and standards as well as tackle issues that pre-existed in the industry to have a field for innovation i.e. machine learning experience for decision making processes.
01
But machine learning is no magic. You need to provide information (data) and constraints (input for where you want to limit the algorithm's behaviour) for the algorithm to work in the desired behaviour.
The majority of those inputs were decisions that our users no longer needed to think about as they were, as they were used so often, they were a "given" that didn't require additional thought.
Let's say Coca Cola was to organise their new shelf structure for 2022. In the manual process, they would just place the product on the shelves that the category manager thought would be the best position.
Let's say Coca Cola was to organise their new shelf structure for 2022. In the manual process, they would just place the product on the shelves that the category manager thought would be the best position.
Let's say Coca Cola was to organise their new shelf structure for 2022. In the manual process, they would just place the product on the shelves that the category manager thought would be the best position.
But in the machine learning world, you need to specify why you think this place is the best position for Coca Cola, as well as specify that the brand needs to face the front - a detail that a category manager would do by default in a manual process.
But in the machine learning world, you need to specify why you think this place is the best position for Coca Cola, as well as specify that the brand needs to face the front - a detail that a category manager would do by default in a manual process.
But in the machine learning world, you need to specify why you think this place is the best position for Coca Cola, as well as specify that the brand needs to face the front - a detail that a category manager would do by default in a manual process.
This level of detail "rule creation" caused the users to have friction with the product under 3 segments:
e.g.
01
The users ended up having a surplus number of rules per project that often conflicted with each other due to their quantities.
02
The users didn't understand what rules had errors and problems they were facing as per the result of the algorithm
03
The only medium of "output" the algorithm provided for the users was the planograms generated per store as a result of optimisation.
The challenges
The internal process
Another major challenge design had to face was the lack of process that made user research and problem analysis quite impossible.
Hivery is a start-up company going to a scale-up. And hence there are a lot of projects to deliver in a very short time frame. Due to this, the nature of project delivery is often consistent with being reactive and delivering the core functionality without necessarily understanding the user needs and problem statement.
02
a need for a new process...
Where a medium for designers to connect with the clients or client representatives to understand the user issues and pinpoint the right solution areas.
The +$XX way of displaying add on pricing further elaborates that anything other than Essentials is an add on that is additional pricing.
And the area of validation that would allow the designers to assure the business that the solution they are developing is heading in the right direction.
The challenges
Hivery is a B2B company
The last challenge was a generic challenge that related directly to Hivery's business model as a B2B company. Hivery worked with clients, and as an innovative product in an industry that was very category-specific (like Coca Cola, RedBull, Kellog's) the product needed to be tailored to individual businesses needs specifically.
In a particular example, what that meant from a design point of view was that Coca Cola's requirements and experience with Curate App were going to be entirely different to, let's say Kellog's.
That level of customisation made the delivery of a holistic solution particularly challenging.
03
The solution space
started with the process iteration...
What is needed to understand the user problem?
Journey map the requirement
Generate
user profiles
Brainstorm the risks and metrics of projects
Defining stakeholders
Competitor analysis /
existing solutions
Deadlines of
deliverability
User pain-points
Solution spaces for the projects
What is done today? Problems...
?
?
Deep Dives are the session I've introduced to the product team as the kick-off session that begins the "Product Discovery" phase in the Hivery Product Team.
Involving the users, the design team, the product team and the sales team, these sessions are aimed to:
Identify
user problems and pain-points related to the project
point out
solution areas and success metrics of the project
discover
risks and impact of the project to the future of the product
These 1.5 hour workshops consist of 3 major activities that answer the
what?
why?
how?
deep dives
The solution space
"Optimisation Report" Tab
A tab that provides what went wrong in the optimisation session...
Prior to this project, there was no way the users knew what went wrong with the Optimisation session. They were often confused about the results the "Optimiser" (the algorithm) would provide for the users as our platform offered no context as to why certain things failed and certain rules were discarded.
By introducing the Optimisation Report, the users would get a more succinct idea of why certain rules were not working with each other, by viewing the infeasibility report.
!
Elements to interact...
The infeasibility container:
The naming of the issue that went wrong with the optimisation session
The affected number of products, stores, points of distribution and rules by that error
Affected rule that is automatically turned off by the system.
Ability to export a CSV file that includes all the affect points of distribution (which products in which stores)
Ability to auto-remove the affected PODs from the selected rules in the infeasibility container.
Remove PODs Modal:
Upon engaging with the "Remove PODs" button, this modal is presented to the user to select from which rules should the system auto removes the affected PODs.
By removing the PODs from a selected rule, the Optimiser will no longer find conflict between the same rules and would be able to run the rules without any other clashes.
The Auto Remove process will enable our users to fast troubleshoot the issues that are existing within rules but allow the clean-up process to be handled from the machine front.
#
01
Hivery ended up deploying a very useful tool for its users that have evolved their interaction with the product. Through further iterations, this journey will improve further on
02
The Product Team discovered the weakness in their process and allowed an entire process change. By introducing the Deep Dive Sessions, the process will ensure to understand user problems, the depth of the project and the essential solution spaces better than ever.
03
Users had a more cohesive understanding of the algorithm's behaviour as the Infeasibility Report provided insight to the users as to where our machine learning algorithm has fallen short and would provide automated troubleshooting options for users to remedy and better the overall experience with the Optimiser.
