ROLE: UX Researcher, UX Designer, UI Designer
DURATION: Jan 2019 to Present
Rapid Prototyping of everyday consumer and enterprise level problems to build and refine design skills
As part of an academic coursework, we were given a design challenge of varying complexity and scope every week and we had 1 week to do the research, propose a solution, test the solution and also validate through artifacts. This is a work in progress.
The online grocery business is expected to be a $100 billion dollar industry by 2025. Amazon squired Whole Foods, HEB acquired Favor, Walmart is buying up startups such as Parcel, and Instacart and others are still continuing to expand.
The process of purchasing groceries via a website or app comes with tradeoffs. One challenge that has proven to be a difficult experience to duplicate in the digital space is that of purchasing fresh produce and meats. There are individual differences in how people select fresh products, compared to something like paper towels, that are proving difficult to design for from a Ux perspective and is costing these retailers a lot of lost revenue as the convenience of ordering online is beating their desire to buy fresh produce and meat.
Design an experience that improves the experience of ordering steaks online. Assume the scenario of a user rating to buy 4 steaks for a dinner party and they’d like the steaks to be similar across all dimensions possible including thickness, weight, marbling, and anything else you think could be relevant.
With limited time and scope, I focused only on the most essential areas of research and understanding the problem before delving into competitive analysis and rapid iterations and prototyping of the solutions.
This problem is a common one to all the major retail chains and here is how they attempt to solve it in their own methods.
Inferring from the competitive analysis, I came up with the following design pointers for the solution
Reviews help the shopper, and also the option to zoom over the images.
Everyone does the cost as per lb, which is secondary information, but given to the user as primary.
No one talks about marbling and thickness and other such experiential information.
No one talks about it like a meal - in a human centered way, it is spoken of as a product.
Secondary information is also stuff like where is is sourced from, how it is raised etc etc.
The solution has to begin from the place the meat and the cuts are stored. If the cuts are not made in the place that the meat is shipped from, then the meat has to be catalogued in a certain way. The primary thing is to take a picture of it when it is put into the display cases. This would take a picture of the steak in an angle, measure its weight as well.
The catalogue will be updated automatically when this happens.
Flow of the App
Reflection & Learnings
Understanding the problem as not merely a UX based or information problem, but as something that required an infrastructural change in the flow of how goods are brought and catalogued into a store’s inventory.
The tight schedule of less than a week did not allow me to actually interact with the users, the people who work at the supermarkets and conduct user research to the detail I had wanted to.
A UX problem can sometimes translate into a system or an infrastructural problem as well. Broadening our scope from the domain of the problem space helps to quickly find the solution.
Often times the problem is a source but rather a consequence of something else in an outer domain. Broadening the research would help us identify and rectify such issues.
As we further integrate conversational elements into our devices, Alexa in my Sonos, Google Assistant in my car, Siri in my wireless headphones, it becomes vital for UX designers to design for experiences that may live partially or fully outside of the screen.
Design a flow for setting up a work meeting in your car through Voice Interface. Signify any points of friction or anxiety, and the way to smoothen them. Also design and mention Fallback Strategies (the experience when the AI does not understand or is unable to carryout the user's request).
Understanding the steps involved in setting up a meeting so that this can be translated into the workflow with VUI - first into a flow and then into specific phrase maps and commands.
VUI input and output data
For VUIs, there can be different kinds of data just though the voice content itself. These can be cue words, silence, and although not yet, perhaps other kinds of sounds that indicate input into the system such as laughter, or groans etc. For this project, since there is a lack of input data types from the user into a VUI and the corresponding responses from the VUI to the user, I have made a glossary for the different data types in this interaction.
Understanding VUI flows
While designing VUI, the context of the conversation - which is driven by the intent of the user is the equivalent of a screen in a Visual UI. The user can ask questions which directly are relevant to the system’s previous question, or instead, are only relevant to the systems previous two questions. This can be called local intent and global intent - these need to be prepared for. Since the user is not looking at any kind of visual feedback, the feedback and input prompts need to be sound as well.
Since voice is something users associate with other sentient beings like humans or pets, The expectations are high, and this can be addressed by making the VUI set the rules for the conversation - “you can ask me to repeat, or go back etc”
Keep giving the user feedback about their “location” in the process and also design for pause and continue - most VUIs time out after sometime, and have to begin the whole conversation again. Instead, having a command like, “pause”,“or lets continue this conversation later”, might be useful in giving the user a good VUI experience.
Inputs, Respones and Fallback Strategies
Because of the high expectations of the user from VUI systems in general, the input can be either too long, to vague or too individualized such as use of slang etc.
Broadly, there are two kinds of errors in user input - insufficient or incomprehensible input, or irrelevant input. The VUI has to respond accordingly to each with error messages, asking the user to either repeat or reframe themselves in the case of incomprehensible or insufficient data, or ask them to stick the the conversation at hand in the case of irrelevant input (like a random “will you marry me?” Input in the middle of a conversation).
Initiating and Ending Conversations
All conversations with the VUI can be initiated with a trigger word, which is unique to each device - “hey google”, or “Alexa”, or “hey Siri” are for the google, amazon and apple VUI systems.
While there is no cue word to pausing the conversation with the VUI, inclusion of it might result in better user experience - A user who is sufficiently deep into a flow with a VUI can pause it, take a call or talk to someone else, and get back to the VUI to continue the conversation where they left it, without losing their input until that point. Ending cue words are usually trigger word followed by “stop”. This stops the conversation at any point and also the function.
Identifying keywords through Phrase Maps
in VUIs, users begin with a lot of expectations about the system since users usually associate voice with sentiment beings like people or pets. The same words can be said in numerous ways or in multiple forms in combinations with different words, and tones. For this reason, the cue words and input words are mapped out for alternate words.
The tools used for mapping out for multiple alternate words for cuing and specific input are called phrase maps. For this project, I have not had the time to design phrase maps for each input words and phrases.
Reflection & Learnings
The lack of existing systems and standards in VUIs (not including Alexa and Google's best practices) made it challenging to identify input and response data types from the VUIs and to group them into similar types.
With only one week to finish the project, I was unable to finish the phrase maps for the alternate cue words and input words for the VUIs