How to Build, Productize, and Iterate an Analytical Data Product

How to create an analytical data product

1. First clarify the business scenario and the specific business problem (e.g., user acquisition, frequency increase, revenue growth, activation).

2. Choose an analysis framework to evaluate the scenario’s strengths and weaknesses.

3. Abstract the framework and turn it into a reusable product.

4. Continuously iterate the product because business contexts evolve and a single analysis model cannot solve all future problems.

01 – Understanding a Business Scenario

To understand a scenario, identify the core KPI (e.g., GMV, DAU, playback volume), the teams involved (product, strategy, operations, R&D, etc.), and the metrics that reflect performance. Use the OSM model (Objective‑Strategy‑Measurement) to structure the analysis.

Key steps include defining the KPI, mapping responsible teams, quantifying performance, and selecting common analysis dimensions such as user, scene, channel, content, and promotion.

02 – How to Understand a Business Scenario

Clarify the KPI (e.g., GMV = active users × playback conversion × payment conversion × ARPU) and identify the teams that influence it: functional product, strategy product, content operations, traffic operations, and supporting teams like R&D or asset management.

After defining KPI and teams, build a basic metric system covering core effect indicators and process indicators, and consider common analysis dimensions (user, scene, channel, content, promotion).

Typical dimensions include:

User: DAU, retention, GMV, MAU, playback count, playback duration.

Scene: user stickiness, retention.

Channel: efficiency metrics such as CTR, CVR.

Content: paid vs. free consumption depth.

Promotion: acquisition and activation metrics.

Use systematic reporting to visualize how each dimension contributes to the overall KPI.

03 – Decomposing an Analysis Framework

Methodology: goal decomposition → benchmark comparison → multi‑dimensional analysis → actionable recommendations.

Steps:

Decompose the goal (e.g., break GMV into user count, conversion rates, ARPU) to locate the problematic factor.

Find benchmarks (historical averages, rating‑based distributions, peer content) to assess whether a factor’s performance is abnormal.

Conduct multi‑dimensional analysis (user segmentation, channel, content) to pinpoint the root cause.

Quantify contributions using intra‑group (within‑group) and inter‑group variance calculations.

Finally, propose targeted actions for each identified segment (high‑play‑high‑pay, high‑play‑low‑pay, low‑play‑high‑pay, low‑play‑low‑pay).

04 – Productizing an Analysis Framework

Core: problem → solution → todo.

Two principles:

Clear data‑view order: specify what users should see first, next, default, and personalized views.

Explicit functional themes: separate daily monitoring, weekly review, and monthly/quarterly analysis into distinct modules.

Typical modules:

Module 1 – Overall KPI status.

Module 2 – KPI decomposition (e.g., GMV breakdown with benchmark comparison).

Module 3 – Dimensional drill‑down (user, channel, content) with concise, relevant metrics.

Each module should provide a summary conclusion and reference points for comparison.

05 – Iterating and Operating the Product

Product iteration follows three phases:

Cold‑start: validate logic and strategy coverage (e.g., A/B testing) with seed users.

Promotion: expand usage, collect new scenario feedback, and continuously refine the analysis framework with analysts.

Maintenance: after the strategy set stabilizes, focus on user‑experience improvements and incremental feature upgrades. Evaluation criteria center on whether the product’s strategic recommendations positively impact business metrics.