E-commerce Analytics: a Tokeny case study
Integrate multi-dimensional store data (traffic, conversion, AOV) into operations analytics dashboards. Follow the recorded task, execution screenshots, actual files and verification notes.
By the Tokeny team · Windows · Tokeny 1.7.8 · gpt-6-sol
Prepare the inputs
Create a separate workspace in Tokeny, download and unpack the sample inputs, then select a model that supports tools. The sample uses synthetic data or the public sources named in the article.
- store.csv
Send the task to Tokeny
The original task was executed in Chinese. Copying these instructions reproduces the same task; the results below include an English report.
整理三天的店铺流量与成交数据，用加权口径计算整体转化率和客单价，生成 HTML 运营看板与汇总 Excel。 交付文件：store.html、store.xlsx。 验收要点： - 访客合计 3000、订单 154、销售额 31440 - 整体转化率 5.1333%，不能直接平均各日比例 这是使用合成演示输入或指定公开资料的真实操作案例。请先读取工作空间输入，再实际生成文件并自检。所有写入、脚本与临时服务限于当前工作空间；临时服务只监听 127.0.0.1，执行后关闭；不触达外部收件人、不发布、不付款、不更改系统设置，不调用子智能体。需要公开资料时可读取给定 URL。只启用完成任务必需的工具。另写 report.md（中文）和 report.en.md（英文），每份 400–800 字，包含输入、处理步骤、真实结果、检查方式和局限；失败或未验证部分如实记录，不能宣称成功。最终回复用中文列出产物、关键结果与待核实项。
Follow-up instructions after review (1)
复核发现 store.html 在 390px 手机视口存在横向溢出：逐日明细六列表格撑出了页面。请只修复响应式布局，使表格可以在独立容器内横向滚动，页面本身不能横向溢出。请用真实浏览器在 390px 和 1440px 检查 document.documentElement.scrollWidth <= innerWidth，保留原始数据与正确的指标，更新 report.md/report.en.md 记录本次修复与实际检查结果。
The recorded run used an isolated demonstration workspace. Review the commands and file paths before approving operations in your own workspace.
What actually ran
The task was submitted through the Tokeny input box. An online model read the inputs and executed tools. This run covers the local sample task and its actual files. Results and checks are recorded below.
Results and verification
Input
The only input used was store.csv in the current workspace. It contains three synthetic demonstration records with date, visitors, orders, and revenue fields. The records are: 2026-10-01 with 1,000 visitors, 50 orders, and 10,000 CNY revenue; 2026-10-02 with 800 visitors, 32 orders, and 7,040 CNY; and 2026-10-03 with 1,200 visitors, 72 orders, and 14,400 CNY. No external public source or additional business data was used.
Processing
The CSV was read as UTF-8 and each numeric field was converted to a numeric type. Daily rows were retained for traceability. Totals were calculated first, and the overall metrics were then calculated from those totals. This is the required weighted approach: overall conversion rate equals total orders divided by total visitors, while overall average order value equals total revenue divided by total orders. Daily conversion rates and daily average order values were also calculated for comparison only. The results were written to store.xlsx, with separate raw-data, summary, daily-metrics, and methodology sheets. A static store.html dashboard was also generated with KPI cards, a daily table, comparison bars, formulas, and limitations.
Actual results
Total visitors are 3,000; total orders are 154; and total revenue is CNY 31,440. The weighted overall conversion rate is 154 divided by 3000, equal to 5.1333%, displayed as 5.1333% to four decimal places. The overall average order value is CNY 31,440 divided by 154, equal to ¥204.16. The requested acceptance totals are therefore present in both deliverables.
Verification
An independent Python calculation re-read the source CSV and checked that totals equal 3,000 visitors, 154 orders, and CNY 31,440 revenue. It also checked the conversion-rate numerator and denominator, the average-order-value calculation, and the key values embedded in the HTML. The XLSX package was inspected with the standard library and confirmed to contain four worksheet XML files, formula nodes, and cached key results. No browser server was started; therefore no live browser screenshot or interaction check was performed. The HTML is a static file and can be opened directly.
Limitations
The input is synthetic and should not be treated as a real operating result. The analysis does not account for refunds, cancellations, unique-visitor deduplication, attribution windows, channel mix, product mix, discounts, taxes, costs, or anomalous orders. It therefore cannot support profit or campaign-efficiency conclusions. Formula result caches were written from the verified Python calculations, but readers should use the summary values and methodology sheet if a viewer does not recalculate formulas automatically.
Responsive Fix
A review found that the six-column daily-detail table widened the page at a 390px mobile viewport. This revision changes responsive layout only: the table is now inside a dedicated .table-scroll container with overflow-x: auto, while the table keeps a readable minimum width. Source data, totals, formulas, and all other business content remain unchanged.
Browser Verification
A real Edge browser loaded the local page from 127.0.0.1 at viewport widths of 390px and 1440px. The page evaluated document.documentElement.scrollWidth and innerWidth. At 390px, both values were 390; at 1440px, both values were 1440. Both checks therefore passed scrollWidth <= innerWidth. At 390px, the table container had an internal scroll width of 720px, confirming that horizontal scrolling is confined to the table container rather than the page. At both widths, the browser also confirmed the presence of 3,000 visitors, 154 orders, CNY 31,440 revenue, a 5.1333% overall conversion rate, and a CNY 204.16 average order value. The temporary local server was closed after testing.
Download and inspect the outputs
The archive contains the files created during this run. Use these checks when adapting the example to your own data.
- Reconcile the reported results against the downloadable sample inputs.
- Open the generated files and check the contents relevant to your task.
- Review the limitations and any unverified steps in the report.