Design an online news service that provides curated news content that can be trusted by the masses.

What is the goal of MS in this? Is it drive engagement? Is it another venue to gain considerable ad spend?
- Ad spend, engagement
Is it going to be an app or standalone website?
- App & Standalone website
What does the product do? Is it going to only curate (or) in the business of journalism and content creation?
- Curate
What will be the geo where it will be launched first?
- USA
Users
- Anybody with internet

User Personas

Morning reader in a commute
- Characteristics - Busy professionals, lack of time. Prefers a digest with important news
Avid news reader - Retirees, people with time in their hands spends time visiting many websites and reading news articles
Young, teens - sports and entertainment fans

I am going to pick the busy professionals personas

Pain points
- Lack of time, commute. Needs a site that can be trusted and saves time and effort.
- Wants to know what's going on and doesn’t have time to go in depth for every story. Picks and chooses stories that they want to get a deeper understanding
- Wants to be seen as knowledgeable and understand a topic in a multi-faceted way, diverse opinions
- Have interest in politics, sports, local, state, entertainment

Prioritization

Within the topics of curated news, some of the topics are by default trustworthy like game scores, sports results (to a certain extent)
Have other sources for entertainment news currently
Political news
Local news
State news

I am going to pick non entertainment news along with sports as first order of implementation

Implementation

Since we are in the creating a trustworthy curation of news , I am going to use a methodology similar to how ad/page ranking model works to score every news
Classifying the trustworthiness of news using a trust index/number which is a number that can be derived from some algorithm, composing of many pillars, few of them below
- Source of the news - website
- Journalists if mentioned
- Number of citations that have quoted this as a source - This can morph into a problem with virality. Care should be taken as when the virality goes up to do fact check
- References to sources of information quoted in the news article
- Automated Machine Learning/AI contextual analysis of the news
- Automated fact checker - highlighting facts and opinions in a news article, Calling out data as not verified. If flawed calling out using a color coding
- Tagging every article as the kind of story
  - Opinion/Point of View
  - Interview
  - News article
On a curation/digest level, diversity of news is also important to help understand other points of views
While sending curated digests, we will also send another link that might be offering a different point of view as identified by our AI model with tags like "you might like"
Crowd sourcing for verification of facts

Evaluation

Use Case	Implementation difficulty	Impact to User	Ranking
Building a news source database	Medium Based on its bing search index, MS has the expertise definitely. It is also the foundational element of the product	High	P1
Ranking by source	Difficult Giving scores to news sources	High	P1
Building a journalist score	Medium Building a journalist DB and scoring them	Medium	P4
# of citations	Medium Extension of its search index	High	P1
Automated reference creator	Difficult Good value add if people want to see the source of the news	High	P2
Automated Machine Learning/AI contextual analysis of the news	Difficult, self learning becomes better as users vote on the accuracy Medium For providing a succinct version of the news	High	P2
Automated fact checker - highlighting facts and opinions in a news article, Calling out data as not verified. If flawed calling out using a color coding	Difficult In the long run very helpful, builds credibility.Again self learning, needs input and previous items to get this to work	Medium	P3
Tagging every article as the kind of story	Medium Needs AI reinforcement model to understand the context and build the tagging.	Medium	P4
Crowdsourcing of fact verification	ROI analysis needs to be done. Easy Users taking effort to validate the facts, might also bring diversity of opinion to users. Users might have to incentivized to do so	Medium	P2
Things you might like	Low effort Requires curating and classifying different news sources of same topic/similar topic	Medium	P3

Metrics - Not focusing on ad metrics currently

App
- # of downloads
- % increase of # of downloads week, month
- # of updates to software
- # of profiles filled with preferences of news sections
- # of App opens and histogram of App opens over time (daily, week, month)
- Time spent in app (per user, cumulatively)
- Time spent per story
- # of shares of app
- # of shares per news article
- # of uninstalls
- # of Crowd sourced fact verification requests
Website
- # of daily visitors, weekly, monthly
- # of unique visitors, daily weekly monthly
- % increase in visitors unique
- # of accounts created
- Pages per session
- Session length
- Bounce rate
- Exit rate

Summarizing

Bulding a news ranking AI system using contextual analysis, crowd sourcing of validation

Chaitanya Gaur · Answer 2 · 2020-10-04T21:45:07+0000

Goal: Is to create a common news platform to provide latest content or updates.

Clarifying Questions: Is this an mobile app(Yes)

Are we designing for Global or US: US

Are we targeting user aquisition right? Yes

Different set of personas

1) Regular reader who has ample amount of time to read different topics

2) A reader who has limited time and wants to just get latest news

3) Reader who is focused on specific topics like Politics, Sports, tech etc

4) Content Moderator

I would go with person who has limited time in work day and just want to get latest updates about the day.

Use cases for this persona:

1) I want to know top news from US as I have limited time.

2) I should be able to comment, share or like the content

3) If I miss reading, I should get daily digest or something

4) Provide notifications around natural calamity, disaster or high alert.

Features                                    Importance to Persona                      Implementation

1. Sign Up/Register                Low/medium                                     Use FB SSO

2. Provide top news                          High                                      Use reading Score

3. Send email if u miss login           medium                                    easy email digest

4. content to be share/vote/comment medium                                   easy

Prioritizng features based on importance to Persona: 2,4,3,1

Reading Score: How many people clicked on each content+ no of votes+ no of comments+ no of times shared

Metrics: 1) no of mobile downloads

2) no of activations

3) DAU/MAU

4) Avg.no of content clicks in a single session per user

5) Avg. session time

6) % of users clicked at least 1 page

From above 6 metrics, I will be focusing user aquisistion so my primary metrics would be downloads/activations and % of users clicked at least 1 page. And based on the value, i will dig secondary metrics.

shu · Answer 3 · 2022-03-19T03:01:11+0000

Clarification (Assume interviewers replied: you can define all by yourself)

What is the definition of "curated news"? (Relevant, high quality)
The online service is a web app or a native app? (Native app)
How to define the "masses"? (a majority group of users or represents the general behaviors)

Problems-why do users want to see curated news

Users feel overwhelmed when accessing tons of news
Users need to find relevant and high-quality news, to better understand the world around them
Users want to hear a range of different views for the same news, to know the full picture

Users Segmentation-several options

By news consumption styles:

"Fast-food" style: no time to read details, only scan the headline stories
"Snack" style: have enough time to explore and deep dive into news storyline.
"Michelin" style: focus on several key categories of news, but need expertise level of info, as it matters to their business/work

2. By working status (or age)

Retired people (> 60+)
Students (>15+)
Work people (in between)

I will select the #1 option, and esp focus on the "Fast-food" and "Snack" style: as it represents the "masses";

Primary: "Snack" style- as it is with good engagement, growth potential, ads target audiences
Secondary: "Fast-food"

(note: if there is no "masses" limitation, will pick up the "Michelin" style, as in my POV it can be more creative)

Why build it

Value to users (see the above "problem" part)
Impact
- Market size: build a news platform (even community) to gather high-end elites, with an estimated XX DAU
- Revenue: (1) brand advertising (2) pay and subscription fee, etc. with estimated XX rev

CUJs & Solutions

*Snack Style (SK) *Fast-Food (FF) Style

P	User Groups	CUJs	Solutions	MVP or Not
P0	SK & FF	I want to read fresh and high-quality news	1. Index and understand the massive sources of publishers 2. Design a "scoring mechanism" to measure publishers' news "quality score" based on data signal-e.g., Impressions, visits, clicks, reading time, news time stamp, head/torso/tail publishers, etc. 3. Track users' data to backfill and update the quality score mechanism	-MVP with selected 5-8 publishers
P0	SK & FF	I want to read top, breaking news	1. Identify the hot topic and breaking news: -publishers side defined -data signal: impression, clicks in the last 24-48 hours -search: top search queries 2. Trending -leaderboard to recommend top viewed news articles based on users' data	-MVP: #1 -Later: trending
P0	SK & FF	I want to read relevant news	1. Design an algorithm to recommend relevant news to users: Type A: Local relevancy -data signal - Geo, users preference settings, etc. Type B: Interests relevancy- historical read data, search queries, users preference settings etc	-MVP: allow users to do preference settings -Long term: enable personalization recommendation
P0	SK & FF	I want to read news on the go	1. Build mobile apps	MVP
P0	SK	I want to read news from diverse publishers	1. Index multiple publishers via API 2. Partnership: allow publishers to nominate partnership via tool/process	MVP with selected 5-8 publishers
P0	FF	I want to read news from authority publishers	1. Customization: allow users to define only see news from publishers they selected	MVP
P1	SK	I want to digest the storylines of news	1. Build a storyboard to arrange news with a timeline
P1	FF	I want to scan the headline/TLDR of news before digesting	. Build topic panel or bulletins: only show news bullets	MVP
P1	SK & FF	I want to share the news with friends	1. Enable to share news links on social apps or email apps: Whatsapp, Facebook, Gmail, etc.	MVP
P1	SK & FF	I want to search and find interesting news	1. Enable news search	Not in MVP
P2	SK & FF	I want to subscribe to the news I am interested in	1. Design subscribe features, users can define only receive news fall into their interested categories: e.g., sports, entertainment	Not in MVP
P2	SK	I want to understand other users' opinions of the news	1. Allow users to make comments on the news page	Not in MVP
P2	SK	I want to share feedback on the publisher's quality	1. Allow users to give "thumbs up, down" or other signals-" don't recommend this publisher"	Not in MVP

Success metrics

Adoption: App downloads, User registers, DAU, paid subscribers
Engagement: Likes, share, comments, feedback, average news articles/user
Retention: DAU/7DAU
Monetization: subscription fee, ads revenue (in future)