Sebagai seorang software engineer yang jarang berhubungan dengan sisi bisnis, LLM bisa membantu dalam mengasah intuisi bisnis kita. LLM bisa diminta untuk menganalisa skema pada tabel dari database dan membuatkan rekomendasi analisa yang bermanfaat untuk bisnis. Lakukan ini berkali-kali untuk melatih intuisi bisnis kita.

Sebagai Non-Engineer, kita bisa mulai meminta skema tabel dari teman-teman engineer kita. Sehingga kita bisa memperoleh gambaran umum analisa apa saja yang bisa kita dapatkan.

đź’ˇ Tulisan ini adalah SLOP, alias Spammy, Low-quality, Over-produced, Predictable. Sebuah konten tentang AI yang mudah untuk ditiru. Tapi saya percaya setiap SLOP ada manfaatnya jika dibagikan untuk dipelajari bersama.

Prompt

Misalnya ada sebuah tabel bernama users_summary , maka prompt yang kita masukkan adalah:

The following is a schema from table users_summary:

Name, Type
id, integer
registration_platform, string
most_used_platform, string
last_used_platform, string
registration_country, string
most_known_country, string
last_known_country, string


Analyze this table and then think about ways to use this table to generate meaningful information.

Contoh Jawaban dari LLM

The users_summary table contains valuable aggregated data about user behavior and geographical usage patterns. Here’s a breakdown of the schema and how it could be leveraged to generate meaningful insights:

Schema Analysis

Column NameTypeDescriptionidintegerUnique identifier for the user.registration_platformstringPlatform used to register (e.g., iOS, Android, Web).most_used_platformstringPlatform the user uses most frequently.last_used_platformstringThe most recent platform the user accessed.registration_countrystringCountry where the user registered.most_known_countrystringCountry where the user is most often active.last_known_countrystringMost recent known country of user activity.

Potential Uses and Insights

  • Cross-Platform Behavior: Identify shifts from registration to most-used or last-used platforms (e.g., users registering on Web but switching to mobile apps).
  • Platform Loyalty: Analyze the percentage of users whose registration_platform = most_used_platform to measure platform stickiness.
  • Platform Trends Over Time: Track platform changes across users to identify trends in user experience preferences.

2. Geographical Movement and Localization

  • User Mobility: Compare registration_country, most_known_country, and last_known_country to infer migration or travel trends.
  • Localization Strategy: Focus product development or marketing efforts based on countries with the highest shift in most_known_country.
  • Geolocation Accuracy: Evaluate how consistent user location data is over time.

3. Churn and Activity Monitoring

  • Dormancy Detection: If last_used_platform or last_known_country is outdated or rare, flag users as potentially inactive.
  • Platform Reactivation: Identify users who return to different platforms than they registered with, possibly for targeted outreach.

4. Segmentation and Cohort Analysis

  • Cohort by Registration Source: Analyze behavior differences between users based on how and where they registered.
  • User Segmentation by Platform Loyalty or Location Stability: Group users into segments such as “stable users” (no changes in platform or location) and “dynamic users” (frequent switchers).

5. Product and Marketing Insights

  • Market Penetration: Determine which platforms and countries are growing most rapidly in terms of usage vs. registration.
  • Campaign Effectiveness: Correlate country/platform shifts with specific campaigns or product launches.

Kesimpulan

Dengan bantuan LLM, kita bisa melatih kemampuan analitikal kita untuk memberikan informasi yang berdampak pada performa bisnis. Sebagai software engineer tentunya kita sudah bosan mengerjakan sesuatu yang tak berdampak, bukan? Saatnya software engineer peduli dengan bisnis.

Hasil di atas masih bisa diperdalam kembali dengan beberapa contoh berikut:

  • Could you focus on insights that may relate to… (e.g. revenues, retention)
  • Please make me an example SQL statement for…