Dry goods sharing - taking over a new data team as a lead - Problem Inventory and insights findings

This summary is gradually sorted out from my work log , Recorded in 2015 What I did when I took over a data team in .

At that time, being appointed to take over a new data team needed to be in the previous Q  What to do , This Q How to do it ？

I will sort out the contents of the work log at that time and send them through official account . This article contains three parts ,  An inventory of work tasks 、 After checking the problems insights Discovery , Next move action;

Stage One  Problem summary ：

• Business caliber changes very frequently , It will be adjusted and changed every quarter , It needs to be rewritten metric Logic , Relevant data statistics , Reports need to be redeployed , The workload of each redeployment is increasing （ The dimensions that need to be disassembled are getting finer ）. Periodic weekly report and daily report , Abnormal fluctuation analysis , Main heel apps OKR（ Distribution ） of ; This part basically involves every week .Muce Report ① curing , Frequent for daily needs , And metric Data with little logical change , Cure to muce ①（ Such as application upgrade 、 Distribution volume of classification pages and CTR）.

• Temporary assign Of task, It belongs to thematic analysis , Usually with the core metric It's about fluctuations , The cause cannot be found directly from the routine monitoring daily and weekly reports , Constant investment 2-3 Days are even longer （ For example, after the Spring Festival DLU In the fall 、apps Penetration of per capita consumption ）.

• At present, the period of data early warning is relatively short , Japan 、 Zhou , In daily work, for metres Exception analysis of will take up most of the working time , The time for special analysis will be relatively less , There is not much business output with high value .

• Follow the version （ Every two weeks ） and feature, Buried point 、log test 、feature Effect evaluation 、 Put forward product and operation suggestions （ For example, the first two Q The most time invested is application upgrade 、 Details page 、 Classification revision ）. new feature Usually not log（ The difference in SP Strategy evaluation, direct cut-off flow view policy ID that will do ）, The above process must be followed up from beginning to end .

• At present, in addition to docking products , Also often docking operations , such as apps User's life cycle （ docking push operating ）, There are many temporary needs .

• Macro data prediction analysis and research , Important but not urgent .

• Underlying maintenance of basic data , Not much work , a week 1~2 Participate in the solution for the second time .

• stay Muce2 ① Implemented many algorithms ,ETL The process , Need to go muce3 transfer . Big workload , There is no time for this

• startpage Analysis of data periodic changes （ It is mainly the time when weekly reports and data fluctuate greatly ）. Find ways to drive data growth through data methods such as fixed investment .  Effect analysis and operation suggestions after the release of new features

• All kinds of temporary needs cover most of the working time , The time devoted to in-depth thematic analysis is limited .

• Due to the current takeover 2 A project , Irregular temporary needs will double, The urgency of these needs is not easy to assess , It will disrupt the original work plan ;

• Buffet Platform maintenance , Collect user feedback and solve problems in use . But there is no manpower to support the optimization of the project , The spirit is willing but the flesh is weak .

• Business analysis , No fixed docking object , Generally, I will analyze the project initiation and report output according to the business urgency . This part of human input is relatively controllable .

• Data organization and sorting . The general service object is ripple, Applied research ,mkt etc. . Mainly because they don't know how to solve the underlying data , Don't know how to get data . Generally, it is to provide assistance and guidance for their specific problems . The specific content of this part of work is ： The demand is scattered , More , But individual requirements are not very time-consuming .

• muce3 ① Replace muce2 ① It seems that the established intermediate table or result table and muce ① Statements have a great impact , It is likely that the previously established ones are not available .

• There are few cases of cooperating with business parties in data-driven products or operations , More data is used to evaluate product changes or features ,  It is of little value .

• Currently, there is less support , Mainly push Test the user package extraction and effect evaluation .

• Business is changing fast , It will inevitably cause anxiety when looking at the data ; Sometimes business decisions cannot wait for data results ; There is a problem with the data analyzed in the small flow test , Require rapid improvement ; In product improvement, we discuss the indicators we want to see with analysts , Statistical caliber and   Define the burial point , Take a look after the implementation of the project ; In terms of business, the data is still scattered everywhere , There is no way to see it together

• CLM  It should be a long-term project , In the short term, it can't produce much , The longer you do it, the greater the output , People with different data experience do different outputs ; The existing business caliber is not unified ,weekly  Often encounter inconsistent caliber ; There is no full flow chart , Searchable part ; Currently in various PM All the data you want to get there may exist , Just not organized ; Sometimes the data source of daily newspapers cannot be found at all

• Data students   near 2 Work content of the last month , The conclusion is as follows ：

• There are too many temporary needs , It is easy to generate secondary demand , Three demands ; It takes a lot of time , Reduced data analysis 、 Time for thematic analysis ; Resulting in less direct and valuable output ;

• The current temporary demand of data analysts is approaching the limit of the team , At the same time, it also faces a lot of bottom construction 、 Caliber carding, etc ;

• Can temporary demand be diverted to products 、 operating ？ What kind of tools are suitable ？

• APP  The main new version involves a lot of page changes , Large impact area , Corresponding Log It also needs to be adjusted , Can this part be unified by the technology responsible for log management ？

Stage Two Insights：

• Business each Q There are changes , Businesses involving caliber change rapidly , The workload of each redeployment is increasing , There are many temporary needs , There is not much time for more valuable data analysis ,; There is a limit to the monthly demand of each analyst ; In the construction, we are faced with a large number of middle and low-level construction and development ; There is no time for anything else ; It has become a human flesh data acquisition machine

• There is no way to know how well data analysts master the business ; Usually with the core Mertics  It's about fluctuations , From daily monitoring daily 、 The reason cannot be found directly in the weekly report , Need to invest 2-3 Days or even longer for analysis , It has become a thematic analysis ;

• Version upgrades and featur, It takes a lot of energy to follow up the burying point 、log test 、feature assessment , new feature Usually not log; Need to follow up from beginning to end ; There is no unified path management ;

• If muce2 Switch muce3 It's a big impact , I don't know at the moment Muce3  Data accuracy , Available data Muce3  Cover muce2 percentage ; stay muce2  The implementation of daily reports is expected to be in 150+（apps pa） The processing logic of analyst landing is about 500+ strip ,CLM Machining width table used ; Go to Muce3 The transfer cost is large ;

• Buffet New requirements and maintenance of the platform , Because there is no data development 、 backstage 、 Front desk support project optimization , The heart has spare power but not enough ;

6 . CLM Application prospect of , Loss recovery strategy ; The output of data directly affects the business line ;

• The business line is not efficient in using data , By product 、 Operating human flesh to carry ; In the current situation of the company , How to achieve small input and large output , Leverage business ; How to help develop new business ;

Stage Three To Do List：

• Training oriented ： oriented Apps PA（ The future company ） Products 、 operating 、 engineering  , Guide everyone to understand the data ; Combined with classic business cases （ How to find the secret of market fluctuation , How to explore Mertics  The trend of ）, Data analysis method 、 Tool use 、 Acquisition of data source 、 Data acquisition process ;

-- Facing analysts' self-learning ： Everyone, everyone Q Want to do 2 Share business above the market , Business data exploration and discovery sharing ;

-- Improve the output quality of analysts in thematic analysis ; Find ways to divert temporary needs while increasing the time for thematic analysis , Assist in analyzing quality , Grasp the analysis angle ;

• Improve the data monitoring system 、 Data analysis system tools , Including the analysis process ;

-- New log printing and other processes and analysis standardization ;

-- Landing some businesses 、 Data analysis methods plus some data products integrated with functions （sample introduce olap, Introduction path , Funnel analysis methodology ）, Improve products 、 operating 、 Production efficiency of the project ;

• CLM Methodology in operation , Explore a good landing point ;

• --  around Apps PA Our core indicators explore some landing modes from the perspective of data operation ;

4. Temporary demand diversion ：

-- Cultivate new people , Those who are interested in entering the intern can cultivate the ability of data analysis , Simple diversion part simple demand

remarks

①（Muce 2、 MUCE 3（13 year ~16 year ）  Company log data analysis platform , Also assume the role of data warehouse ）

②  The picture is my cat

Pine nuts （ Li Boyuan ）,BI&  An old driver in the field of data products , Floating through several big factories .  Official account number ：  Pine nuts talk about data