Review Flow Redesign

Restorando, Buenos Aires. 2017

The problem

We wanted to bring interesting, accurate and reliable data to diners about restaurants.

Reliable:

Users find information submitted by other diners trustworthy, as there is no interest from diners in hiding flaws or promoting a restaurant.

We were capturing data only for diners who reserved a table, which made the data reliable but less accurate since we got less reviews. We didn't get any reviews from prospect restaurants since they don't have the booking feature, and that made our site not the best tool to find out where to dine out if users were not interested in booking.

Accurate:

Also, restaurants change over time, how can we know something has changed? Should we consider a time lapse (maybe only last year reviews) or we could found discrepancies to know when something changed (which would make it even more reliable)?

Interesting

What information should we capture? What is the most relevant information that diners want to know before visiting a restaurant?

Addiotionaly, can we know what type of user someone is and highlight other diners with the same interest? For example, diners who often go out to dinner with the family will be more exposed to reviews and recommendations from similar diners.

This is how the review flow looked like before this project:

Research

Some of the problems we found in the review flow:

Review's content:

Users didn't even know the rating they were giving to the restaurant. We asked them to rate the food, atmosphere and service, and with that information we calculated the final rating. And, in a way, giving a score can be satisfying, we were missing out on giving users the opportunity to have their history of restaurants and how they felt in each one.
A restaurant with an excellent reputation had received a bad review, and we wanted to know why. I looked up details and found that a diner rated them poorly for their atmosphere. The comment said: "the atmosphere is too young". Perhaps asking for a score on the atmosphere didn't help us understand what the restaurant is like.

Users wanted to read comments while they were validating restaurants. Also, they wanted to read useful feedback, not things like "nice", "great", "good". We needed to get more descriptive feedback in the reviews.
Most users didn't know if they should gave us the price with or without discount (if any). Because of that, the price range we calculated was unreliable as there was a mix of discounted and non-discounted data.
We were asking for subjective categories such as “Ideal for children, for work, for the family, for the couple”. These categories were not helping users, we needed to get specific qualities.
We previously found that actual photos taken by diners were highly relevant to users. We needed to get more.

Wizard experience:

The steps that required the keyboard to be completed were not designed for that. When the user logged in, the keyboard was not open, and when they touched the text area, the keyboard covered most of the content, including the form's actions.
The last few months we had been conducting experiments with new steps. There were three steps that were very successful and we kept them in the flow: plate size (large, regular, or small), ambient sound (low, normal, or low), and menu photo uploader. However, the flow turned out very long and many users ended up abandoning it..
Users didn't have any reference on how long the reviewing flow was.

Therefore, we needed to know how many steps users were willing to complete. Are they more interested or willing to provide some kind of data?

In native Apps, the response rate was good in all the steps. We just noticed a small drop-off after the first step.

On mobile web platforms, the drop increased as users progressed further up the flow. And there were three big drop-offs: after the first, the fourth, and the last question.

In conclusion:

We concluded that users who only completed the first step only wanted to confirm that the reservation has been made correctly. We thought we had an opportunity:

To rephrase the review flow to be a way to contribute to the community: this way, they would be helping other diners.
To allow them to keep track of their own gastronomic experiences, providing them a history with their reviews.

To do this successfully, we needed to keep the flow as short as possible to reduce effort and increase gains. I considered four steps maximum because of the drops in that step. By shortening this flow, we would be asking users for less effort and we would be able to obtain more information from them.

Design

Allow users to score transparently.
Include a “breadcrumb” that shows the number of steps and the progress made.
Design keyboard pages to land the user there, with the field focused and keyboard displayed, with most of the content visible.
Change the empty state on the comment step. Instead of the “write a comment” that we where using, we replaced it with a text that would guide them, with specific information that users reported as relevant: "other users would like to know about the dishes and the ambient (terrace, privacy)"
Include a photo/menu uploader in the comments step. This reduced three steps to one.
Detect if they had photos taken in the area of the restaurant and on the date of the booking, and suggest those pictures. That would make the action effortless.
Clarify that the price reported should include the discount. This allowed us to calculate menu prices and show a reliable price range in the restaurant profile. We could also show how the price was reduced with discounts.
Ask for specific qualities about the food, the atmosphere or the service. Instead of obtaining a subjective rating, we would now obtain characteristics specific to these. For example:
- Food: tasteless, tasty, rare, abundant, vegan, vegetarian, spicy, authentic, suitable for coeliacs, little variety, normal.
- Service: Bad, slow, normal, fast, good.
- Ambient: juvenile, family, romantic, intimate, dirty, noisy, authentic, nice.

As a result, the new flow was shorter, effortless and cleaner. It also maintained restaurants profiles updated and with relevant and reliable information.

Thanks for reading, I hope you have enjoyed it!