Redesigning Grain's feedback experience for better qualitative and quantitative data
Product Design Lead
User Experience, User Interface, User Testing

Problems & opportunities

Collecting feedback from customers via emails primarily did not suffice the company's need for better segmented data.

Grain aspires to use data to identify food trends and taste profiles in order to cater the best cuisines to its customers. The first step in achieving this vision was to build a feedback collection mechanism that allows for more useful and actionable feedback for our food and operations teams. Essentially this means getting the fundamentals of operations right while collecting personal taste preferences in preparation for the development of food innovation and smart food recommendation in the next phase.

The old feedback experience was built on Typeform based on a 5-star rating system. An email would be sent out to the customer at the end of the day after they've made an order on the Grain's website.

Old email feedback that opens up a Typeform feedback form

Here are some issues we identified based on a) the quality and quantity of feedback email replies, b) customers' complaints about the feedback emails:

  • Feedback emails have a 46% open rate but only 20% click rate (to navigate to Typeform feedback form) and 15% form completion rate
  • Customers only give feedback for an order as a whole, we had no way of separating feedback for the individual dishes as well as for the delivery service
  • Regular customers complained about receiving too many feedback emails and unsubscribe from our newsletters which led to missed marketing and reactivation opportunities

We needed to find a way to encourage more responses and engagement from customers without spamming them with multiple emails that chip away at their attention.

Tasty, healthyish meal boxes

Target customers

Grain’s target customers are young professionals with spending power, digital natives with high expectations when it comes to digital products. They are more pro-active when it comes to interacting with digital brands, they identify with the brands' values and form a personal connection with the brands they choose.

For this particular project, the feedback collection targets existing Grain's customers who have made at least one order. These are the customers who are attracted to the brand's 'healthyish' value and have decided to pay Grain for their jobs i.e. good, 'heathyish', tasty food delivered to their doorstep.

Context & needs

Understanding the state of the email feedback system, how that fit into our customers' day-to-day life and the customers' behaviors and attitude towards giving feedback for a product was essential to the solution phase:

  • Previously, customers received feedback emails for their lunch orders at 2PM, after lunch serving time is over which is also during their work hours
  • Generally the majority of customers don't leave feedback unless their experience is a peak experience (peak negative or peak positive)
  • The 5-star rating system is ambiguous and introduces complexity when it comes to what each star/rating means to the individual customers
  • The feedback email experience was separated from the rest of the ordering experience, it happened outside of the Grain's website and apps when the customers were not in the Grain's 'state of mind' i.e. ordering food from Grain
  • Grain sends a host of different emails to its customers including weekly menu emails, and the more irregular marketing and promotional emails, new product feature emails. However for a regular customer who place orders frequently (2-3 times a week), the follow-up feedback emails caused the frustration of receiving many emails that clutter their inbox
  • For this project, we focused on the web and mobile web platforms and most popular browsers such as Safari and Chrome but also Microsoft Edge considering our customers access the Grain website via their work desktop
Old Typeform feedback form

High-level initiatives

The two main objectives of the project became apparent to us: we needed to make our feedback system as unobtrusive to our customers as possible and at the same time collect more data for the company.

Our initiatives for this project include:

  • Integrating the feedback experience into the website and apps to keep it in context
  • Simplifying the rating mechanism to encourage more and more accurate ratings
  • Allowing customers to give individual feedback for individual meals and then for the delivery service separately
  • Allowing us to track Order ID and User ID that goes with each feedback


We started with identifying the most important case in which our customers would want to leave a feedback for their order. As mentioned earlier, generally the majority of customers don't leave feedback unless their experience is a peak experience (peak negative or peak positive). This happens after they have made their order and eaten their meal. They then revisit their order confirmation email or the Grain's website and apps to leave their feedback for the specific order. After speaking to a few of our customers, we came to the conclusion that the feedback form should then be available to them upon landing on the Grain's website and apps after they've received their orders.

On the other hand, we also started looking into the different rating systems. For the new feedback experience, we wanted our customers to be able to as quickly as possible rate the individual dishes in their order, this could mean having an upward of 6 items (the average order size) to rate. Additionally, given the ambiguity introduced by the 5-star rating system, we hypothesized that a more 'human' approach would reduce cognitive load and yield better results. Smiley faces were taken into consideration. As we were picking the different happy and sad faces for our feedback form, we realized that the human emotion is also very complex and customers potentially would also infer a smiley face differently from each other. We went back to the drawing board and revisited the intent of the feedback system. We agreed that a binary rating system would be efficient, accurate, simple and get us the most straight-forward data set. The thumbs up proved to metaphorically make sense to our customers as a strong approval/disapproval signal but remained playful and friendly.

The myriad of human emotions

Looking through our customers' feedback provided us with some insights into the most frequent complaints that our customers have for our food and delivery service. We wanted to make it easier for our customers to quickly surface these negative experiences by adding quick key words into the feedback form. The data gathered from this would then become actionables for our internal teams. We worked on a quick mobile prototype to test these assumptions along with the interaction design of the feedback form.

Individual feedback prototype, created with Flinto

We ran this test internally and most of our testers found the keyword quick selections convenient and helpful.

Interfaces & more prototypes

We moved on to add more details to the feedback form to run more tests with larger orders i.e. orders with more items.

Mockup of feedback states

Concurrently, we wanted to also test the different ways to show customers the feedback form when they revisit the Grain's website or apps. We came up with 3 designs and created 3 Invision prototypes to test: a) a banner integrated on the menu experience, b) a toast and c) immediately showing the feedback form modal popup.

The banner and toast designs proved to be insufficient. Our testers did not notice the banner design while dismissing the toast message, assuming it was an unimportant marketing toast. To our surprise, the popup modal though more intrusive, was the most effective. Dismissing this popup was easy enough that the testers did not find it cumbersome.

Feedback flow usability test

Usability test

The final round of usability test on the feedback form brought about some findings that helped us make the last few tweaks to our designs:

  • Customers want an easy way to vote positive for all of the items in their order so we added an 'All is good' text button
  • Customers find it cumbersome to comment individually for each of the item in the form, instead they prefer to leave a summary comment at the end so we decided to remove the individual input fields for the first release
Feedback form on the app
Final details

Results and room for improvements

Rapid prototyping and testing helped us tremendously in quickly validating our assumptions for this project.

We launched our new feedback system and measured the results, both quantitatively and qualitatively. We found that the feedback completion rate has increased significantly, in fact, doubled the original number. More importantly, Grain can now build a meal and delivery score dashboard to monitor the quality of our food and delivery service. This will enable food innovation and smart food recommendation later on.

As the need to collect even more quality feedback on the individual dishes grew over the months that followed, we decided to add the individual comment input field back to the form and monitor whether it would affect the customer experience. We had to make some trade-offs. There were less qualitative feedback for the entire order in the beginning but as customers started to perceive the value of how their individual feedback helped improve the quality of the dishes on the menu, the numbers increased.

Individual meal and delivery score dashboard made a reality by the new feedback system

Balancing customers' needs and business' needs has led to some interesting results in this project. Our customers' first instinct rendered the individual input field feature redundant but as the business needs prove to bring customers values, the final result was extremely positive for all stakeholders.

A potential problem with the modal popup is that in order to dismiss the feedback modal quickly, some customers would rate all positive for their orders. By and large however, the amount of insights we gather from qualitative feedback outweighs the minor use case.