A different kind of analytics, search arbitrage! The Analysis team focuses on:- Answering questions from upper management (CDO, CTO, VP Product, etc)- Building reporting assets that express useful truth to the media buyers, relationship managers, etc.- Dreaming up Better Ways To Do The Business (sic)Something that I'm proud of: we've never understood our data/business match better than we do today, both mechanically and dynamically.Some projects:- Sorting Hat, where we seek "better" term bundles to use in Google SERP requests: ~5% revenue lift, ~25pp reduction in Google intervention, with some banditing I think we can get this to ~8%, on top of some other possible improvements. What I love about this is there's no - count'em, zero - NLP involved. Just dynamics!- we call it Shapes: a tool that lets relationship managers better understand their postures, through contextualising new publishers' traffic ("aha, newpub, you're shaped like oldpub32"), expressing publisher similarities, and (somewhat reductively) explaining key characteristics of given groups. More formally, it's a reduction (UMAP!) of a 66D featureset down to 2, with some domain knowledge/context applied. Special thanks: Komodo Technologies!- Assessment of 3rd party risk scoring solutions, and whether we can use them: an unbounded question if there ever were one.- Research documentation! Confluence has no respect: you write for a whole day and at the end? "7 minute read".Currently working on: rejection models identifying what traffic we shouldn't buy and definitely shouldn't sell. Somewhat tricky, not least of which because we may fork it into an online learning model too.And of course, spreading the good word of SQL. Unfortunately the scale of data is so much larger, n(stern calls from dba) has gone up to 2, with an option on 3, 4, even 5. I'm sorry.

As the Structural Data Analyst, I was the Analytics team's SME on data retrieval and delivery - especially in the context of understanding our warehouse (Redshift) and our BI layer (Tableau). I constructed canonical datasets, pulling together Single Sources of Truth (SSOT) and patching semantic gaps. If I couldn’t get it, it didn't exist.Working with Studios managers, I constructed a canonical dataset from interdependent project management records with a (intentionally unstaged!) query, built Tableau reporting on top of this Common Dataset, and generated correct reports in 30 minutes that used to consume 4 manager-days weekly - a 72X improvement!For Scout, I designed and built a basic reporting backbone in Tableau, providing a centralised alternative to dozens of bespoke Tableau reports running off custom SQL.I designed and constructed the SQL pipeline generating canonical data on sales staff performance. This involved also defining semantic bridges across the 3 business models (Staffing, Studios, Scout) and accommodating business changes. From these datasets, we calculated outputs like commission and utilisation.To make sanity checks and deep dives more accessible while preventing wheel reinvention, I constructed a general SSOT workbook for the Analytics team expressing the current state of knowledge of our data at a granular level. Complementing this, I wrote explainers describing the Staffing and Studios business models, and the data generated. This simplified onboarding and increased understanding of the business in the Analytics team.I also wrote explainers for Analytics in how to understand and apply the basic concepts of SQL, as well as good practices to use and antipatterns to avoid.Additionally, I provided analysis for C-level projects, such as: Estimating potential financial impact of extending sick time policy to all US talent Assessing the forward momentum of remote work Analyzing impact and adoption of the Redeployment product

At Aquent, I was the data warehouse SQL monkey, quickly becoming the SME for data retrieval and delivery (n(stern phone calls from DBA) = 1).Throughout my time at Aquent, I constructed, refactored, and maintained the Marketing Effectiveness reports, improving them from a daily 4 hour error-prone manual task in interlinked Excel spreadsheets to a 10 minute SQL-based Yellowfin report, and then to an established Tableau datasource: cleaner, faster, less manual, and more explicable.Similarly, Reporting & Insights freed up managerial resources by simplifying the weekly spreadsheet-based collation of Staff Action Forms into a simpler 3 hour ETL-supported process.Reporting & Insights also supported C-suite ad hoc analysis and hypotheticals. This included the Coreflex project, where we identified the "saturation limits" at which non-Core tasks should be offloaded from Client Service Managers to Flex staff.I was also attached to the Product team, making available data and analysis on products such as Expert Interviews and ChequedReferences to Product Managers.

As student guides, we represented the university to international and domestic visitors, both electronically and in person; we also led large public (30+) and private (15+) tour groups. Depending on the group, these tours were conducted in either English or Mandarin Chinese.While at the Information Centre, I streamlined existing work processes by creating an automated spreadsheet for the daily cash-out procedure. This saved time and decreased likelihood of human error. I also created a spreadsheet system to consolidate staff responses for shift scheduling, reducing necessary managerial oversight and effort.

I translated English essays and blogs assigned to me by the China Personified producers into Chinese (Simplified) for the site. It was a fun way to practice my mother tongue professionally, during term!

While at InSegment, I researched online application procedures for onboarding recommendations to the client (Blue Hills Bank). I also wrote tier-targeted copy and designed layouts from stock for client (HUECU), online banner ads promoting a new product, and outlined notes for internal use on extant SEO techniques.

In a 3 AA team, we completed the assembly and delivery of rebranded staff uniform packages to the entire hospital (14 wards, 10 outpatient clinics) in 5 business days, half the time estimated by our manager, through applying assembly-line methods.I also culled data from vendor quotes and scrutinised drafts of vendor contracts, answering to the department administrative manager.

I answered for 14 servicemen (recruits to corporals inclusive) in SIW HQ Wing.As the Operations Specialist, I introduced automated spreadsheets to the Course Tracking process, reducing human error. I also streamlined the daily Parade State compilation by ~90%, from 4 hours to half an hour, by setting up a checklist process.As the Aide to the Chief Instructor, I also prepared documents and analysed course reports for presentation in the monthly command conferences, in which I served in a secretarial position.