Led the development of a no-code data analytics platform

Situation

At a global professional services consultancy, our data science team worked closely with the Strategic Communications practice in London, supporting fast-turnaround client requests. These ranged from understanding how people discussed e-cigarettes on social media to analysing conversations about ESG across online forums. Although the subject matter changed from project to project, the underlying analytical process did not: collect and clean text data, apply NLP methods to surface topics and sentiment, discover the influencers who shape the direction of discussions and produce outputs consultants could incorporate into client reports. Delivery depended on data science capacity for every request, however routine, which constrained how quickly the team could respond to clients.

Strategy

I proposed replacing project-by-project delivery with a self-service platform, designed the analytical workflow and front-end, and led a team of four data scientists through development. Our approach had three strands:

Requirements gathering with the Strategic Communications team to establish how consultants actually worked and where the bottlenecks sat
Analysis of past engagements to identify the methods that recurred most frequently across projects
Translation of that recurring workflow into reusable no-code modules, allowing consultants to run standard data analyses — including topic modelling, sentiment analysis, and social network analysis — and generate initial report outputs independently

Result

Recurring analysis projects that had taken days were completed in hours — a reduction of roughly 90% in delivery time. A repetitive analytical workflow became a reusable, self-service capability: consultants could answer routine client questions without a data science hand-off, while the analytical approach remained consistent across engagements. This freed the data science team to concentrate on bespoke problems that genuinely warranted custom work.

情境與背景

在一家跨國專業服務顧問公司，我們的資料科學團隊與倫敦策略傳播部門緊密合作，協助處理時效緊迫、需快速交付的客戶需求。這些需求涵蓋範圍廣泛，從解讀社群媒體上大眾對電子菸的討論，到分析網路論壇中有關ESG議題的對話。儘管每個專案的主題不同，背後的分析流程卻大致相同：蒐集並清理文字資料、運用自然語言處理（NLP）方法擷取主題與情緒、找出主導討論走向的關鍵意見領袖，並產出顧問團隊可直接運用於客戶報告的成果。無論需求多麼例行化，每次交付都仰賴資料科學團隊的產能，這也限制了團隊回應客戶的速度。

策略與執行

我提議以自助式平台取代逐案交付的模式，並負責設計分析流程與前端介面，同時帶領四人資料科學團隊完成開發。我們的做法包含三個面向：

與策略傳播團隊進行需求訪談，釐清顧問實際的工作方式，並找出流程中的瓶頸
分析過往專案，找出跨案件中最常重複使用的分析方法
將該重複性流程轉化為可重複使用的無程式碼模組，讓顧問得以自行執行主題建模、情緒分析與社群網絡分析等標準資料分析，並獨立產出初步報告內容

成果

原本需要數天才能完成的例行分析專案，如今僅需數小時即可完成，交付時間縮短約90%。在平台上線之後，高度重複性的分析流程，可由非資料科學家的專業顧問來主導進行，在極短時間內便可回應客戶的例行問題，同時各專案間的分析方法仍保持一致。如此一來，資料科學團隊得以將心力集中在真正需要客製化處理的複雜問題上。