Ellie

Ideation and Research

We began by analysing the existing landscape and competitors. Back in 2016, none of the competitors had a strong online presence, other than few Instagram influencers dominating this area.

We tried to redefine what “online presence” means, and tried not to jump into the action of revamping hellobali existing website.

Ideally, a magazine’s engagement rate depends highly on the content’s relevance to the reader’s preference, and we assumed that boosting engagement rate requires content personalisation. So as any self-respecting product development team would do, we tested this assumption.

We started by asking the question: How and why do tourists read travel magazines?

We knew hellobali readers comprised mostly of international tourists vacationing in Bali, and they got access to the magazine through free copies available in hotel lobbies and restaurant lounges.

But to answer the how and why question was tricky, as we were unable to obtain user’s behavioural data while reading the magazine since offline tracker had never existed in the product.

In our quest to answer the question, together with the magazine’s sales and marketing team, we conducted the following:

1. Interviewing magazine’s client (e.g. hotels, restaurants, spa)

2. Surveying Bali tourists from various nationalities.

3. Creating segmentation based on various travel parameters, and conducting in-depth interview with each segment.

Finding: Travel companion influence your travelling behaviour.

And below is the detail of each segment:

Behavioural findings:

1. Other than their personality, a traveller’s companion highly influence their travelling behaviour.

2. Most users chat with their friends for travel references.

3. The behaviour is recurring across nationalities.

Other than those findings, we discovered several problem during our research

1. Difficult to find a good food recommendation platform in Bali. (remember, this was early 2016 – might not be the case now)

2. Readers found interesting content in the magazine, but ended up browsing online for details.

3. Instead of buying, readers would only read the magazine at places where it’s freely available.

Synthesise

Based on the research findings, segmentation, and problems we discovered, we tried to ideate solutions that can serve as hellobali’s potential product blueprint.

So with all those ideas, we came up with:

Ellie..

An intelligent chatbot in the form of an Elephant, which is an homage to Ganesh, the travel and knowledge god in Hinduism.

Ellie is fun, helpful and knowledgeable – it recommends the most suitable itinerary for the user. It also tells jokes when it can’t please the user with its recommendation.

With the help of an external design agency, we also revamp the physical magazine and utilise the chatbot as a content extension. All the access to the digital content extension is guided by the character portrayed by the chatbot.

Production

We transform the magazine as a bridge to access Ellie the chatbot. Ellie loads as a PWA (Progressive Web App), therefore readers are not required to download memory-consuming mobile app.

So we developed basic NLP (Natural Languange Processing) module in python on top of Amazon Comprehend API, and the PWA front was developed in react.

This transformation meant a complete revamp to the magazine layout and design. It also forced us to rethink the way our readers consume our content via a printed magazine.

So here’s the new revamp of the magazine.

Each month, all of the chatbot’s content is updated based on magazine’s edition. Ellie becomes the central character of the magazine, telling fun facts and trivia on each articles.

The NFC chip is embedded on the magazine cover where user can tap their phone on it to access Ellie.

We give Ellie a quirky, humorous personality, in an attempt to experiment with people’s preference when talking to a chat bot.

Testing the Water

To put the product into test, we conducted guerrilla user testing at various cafes and co-working spaces across Bali. Within 3 days, we gathered 9 participants with healthy mix of cultural background and nationalities.

Participants followed scenarios while being recorded to avoid bias caused by our presence.

What People Think

1. Interacting and chatting with a phone while reading a magazine is unnatural and cumbersome.
2. User should be encouraged to interact with either one of the platform without enforcing them to shift around.
3. To several users, the recommendation engine is still buggy.
4. Users are delighted with Ellie’s personality.

Moving Forward and Retrospective

On the first week of the new magazine revamp, Ellie was accessed by 150 readers. This number gradually increased to around 3000 users on the 6th month edition of the magazine.

Most of the conversations were of casual nature, revolving around restaurant and holiday recommendation. Users were also keen on playing quizzes and trivia with Ellie.

Sadly after the 6th month, hellobali had a major issue that resulted in complete management change. The new management decided not to pursue experimental projects like Ellie and instead refocused themselves on selling traditional magazine and online ads.

We learned the hard way that any experimental projects, no matter how promising, can get shut down due to reasons beyond our control. But hey, life goes on.

Update 2020: We are in the process of resuscitating our NLP model for a contextual chatbot used as wellness assistant. Stay tune to see how Ellie incarnates to be something else 😉