[內推文件分享] 紐約時報產品分析師：產品分析架構試寫
如同我在去年彼時第十一週裡面寫到的，在請學長內推紐時的產品分析師的時候，他建議我可以撰寫針對New York Times Cooking App，如果我作為一個分析師，會怎麼分析這個產品，他會協助將這篇分析短文轉給招募的主管，如果主管覺得我邏輯清晰，可以安排我進入面試環節。
因此，我研究了一下New York Times Cooking這個產品後，按照我對於產品分析的基本認知，撰寫了以下的文章。
在這邊我想要表達的是，作為一個美國求職市場的求職者，有前輩協助內推不一定足夠，如果有這樣的機會可以讓你表現你自己的思路，讓自己在眾多履歷中突出，一定要把握機會，多做一點。
分隔線後，就是我去年針對這個產品寫的文章囉！
Product Analytics Basic Structure for NYT Cooking App
Yun-Han Feng
In this brief article about the NY Times Cooking. I’d like to walk through my ideal thinking process and analytics structure as I were the product analyst in charge of the Cooking product.
I will break down my structure in the below metrics, based on the devices and each user funnel.
As for the major business value and priority for analytics, the main focus in device analytics will be NYT Cooking App and followed by PC website, tablet app and mobile web. I put down this order based on the business model of this product.
NYT Cooking is built on the monthly/yearly subscription model. I believe that for subscribers, the most commonly used device is the app. PC and tablet might be around the same weight. Mobile web will be the minimal choice for subscribers since mobile app is well developed and subscribers are able to create his/her own customized recipes in apps.
As for the user funnel, I will suggest breaking down into these three main phases, and build related product features and optimize the user interface experience for users at different phases.
In the next part, I will mainly use the NYT Cooking app as my example and run through some desired analytics procedure and recommendation for each user phase.
User Acquisition Stage
I will not focus much on the user acquisition stage. For my understanding, the acquisition is the main task for the product marketing team. I will just summarize my thinking process here and decide not to go too deep.
I think the main source of user acquisition of NYT Cooking is affiliated with NYT App. Google search/organic traffic might not be the main source for the website since the main focus for this product is mobile application based.
Another traffic source for website visit and further download, I think is from social media, including Facebook and YouTube. As I observed, the engagement for Cooking fan page and YouTube channel is not quite active. I think the product marketing team needs to think about the way to increase follower’s engagement.
The piece of suggestion I will give here is to use more short clips for boosting engagement and shares. For algorism and user preference viewpoints, a fascinating video can create more tags, comments, and sharing.
User Free Trial Stage
This stage is really important for NYT Cooking apps. Within 28 days, if users grow founder and find the app useful, they might subscribe to the app and get looped in the business model of this app. The main goal at this stage for a product analyst is to find the key metrics that drive users to stay and to keep testing the product feature that might draw users attention and engagement.
Predictive Modeling
I will try to build the predictive model that indicates if a user will churn or not after the 28-day trial. It is a time-based model using cohort to observe the user is coming back or not, which will be the most important feature in this model. Other features I think of are different events number in the app, such as numbers of saving in the recipe box, the average percentage of scroll-down, number of searches, number of rating, number of comments and so on.
I will use the decision tree or random forest model to build the prediction. In the end, not only can I predict the user churn, but also I can identify the key features highly correlated with churn. Based on those features, product teams can further test and optimize those product features.
Suggested Experimentation
The purpose of the trial is to keep the user active toward the app. The possible test I recommend is the experimentation on the notification system. Either the sending time or the notification number can be tested, and it can be expanded to text length, format, and content. The result is based on the conversion/click-through-rate from notification to app. Once we identify the optimal combination, it can be used to effectively bring users back to the app.
User Subscription Stage
At this stage, the goal of the product team is to maintain the expectation and ensure a smooth user experience. By doing so, the continuous subscription is guaranteed.
For me, the good way to retain the user is to provide more delicate personalization. I will list down some possible and simple features and how it can be tested in the analytics perspective.
Customized Search
As far as I know, the search function in the app doesn’t memorize the previous search. It now contains SUGGESTION, MEAL TYPES, DIETS sections. I will recommend building a historical record as the first step and further replaced the suggestion based on the historical records. Using LDA and cosine similarity, the minimal viable product of suggested search can be built.
Furthermore, we can test on different groups of subscribers to see if they are more likely to press the original suggestions or press the historical result and its recommended keywords.
This is purely based on my intuition and search behavior. I make the assumption the choice of recipe is serial-like and cohesive. A person who is fond of Taiwanese cuisine might want to try different dishes under the same type for a while.
Test on the unlimited scroll or the longer scroll-down
This is another feature which I think is worth testing. Forum or media like Reddit and Quora apply the unlimited scroll on its main page. This will enable users to browse through the contents longer.
The risk for this feature is that NYT Cooking is not an UGC app. Editor and content team curate the content, which might limit the number of recipes. If that is the case, the product team still can test how long the main page should be for higher time spent on the test and control group of users.
Descriptive analysis of Recipe Box
This is the part I will be curious about and worth investigating. I’d like to explore the usage of each function in Recipe Box, and based on that, I will suggest putting the most-used features in the first screen or layer. The more users play around with this recipe storage section, the less likely they will churn out the subscription. It creates a personalize stickiness between the user and the app. Cooking and recipe are usually the symbols of memory and emotional moments. While the recipes are piling up in the recipe box, the user might cherish this space more and keep using it.
Summary
In short, I try to plot the analytics path on NYT Cooking as an aspiring product analyst. Focusing on the mobile app, the path is closely aligned with the user journey. By satisfying user’s need at each phase, I believe these proposed recommendations can better user experience and extend the subscription.
這篇文章最終有成功讓我拿到面試的機會，雖然經學長轉述，主管說我大部分的內容是錯的，但觀念邏輯有打到點 (笑)，而最後我也有順利走完紐約時報的面試流程，拿到offer，整體來說是一個很棒的面試體驗，紐約時報產品組也非常精實，也在過程中不斷深化我想要做產品的渴望。
把這篇文章分享出來，也希望在美國商業分析師市場求職的你們，可以不間斷的懷揣著對於分析的熱情，在乏味的求職過程裡面，能找到一些破口，玩玩data、做做自己喜歡的分析專案，並且用這股熱起，一鼓作氣找到好工作。
