A simple formula to master project management metrics—data reviews no longer a worry
For non-technical project managers, there's an unspoken pain
You may have brilliant data insights in mind, yet still fall short when confronted with complex metric formulas.
Metric measurement is good—it pleases bosses and is essential for reviews
But when you enthusiastically open the system, ready to take action, reality often throws cold water on your plans—
As a PM or business owner, all you want is to quickly check: "How many requirements from the last iteration were delivered late by the product team?" Yet faced with complicated calculation logic, you can only close the page silently and turn to internal metrics experts or data operations colleagues: "Hey, could you please generate a report when you're free?" Then comes a long wait in the scheduling queue.
When configuring it yourself, you may need to check SUM, COUNTIF and nested parentheses repeatedly, and clicking "Run" can still return a syntax error.
By the time the dashboard is finally configured, the golden moment has passed. When configuration requires specialist help, a data request may enter a queue and arrive later than the decision it was meant to support.
With a more complex configuration process, metric functions may end up being maintained by a small number of formula specialists. But you shouldn't waste your energy writing formulas—your time should be spent uncovering the real issues behind the data!
AI-generated metric formulas make speaking natural language possible
To address this inefficiency, Feishu Projects offers an AI-generated metric formulas feature. Just speak naturally—the tedious work of drafting formulas is handled by AI. Confirm with your account manager which editions and plans include this feature.
When creating a metric chart and needing a calculated field, enter the formula editor. On the right side of the editor is the AI chat panel.
Confirm edition availability with your account manager.
Start by describing the metric in plain language
All you need to do is enter everyday language into the chat box—for example: "Count the number of items where *@Requirement Type* equals standard process." AI drafts a suggested formula from your stated intent; verify the underlying logic matches your team's definition before adopting it. Before clicking "Adopt," review the suggested formula against how your team actually defines the metric—does "delivered late" here match your iteration's cut-off rules?
If you need more granular data, there's no need to struggle locating field names—just use the @ function as if chatting in a group. For instance, type: "Count the total estimated points of requirements assigned to @Wang Xiaoming in @Product Team." AI is designed to interpret @-mentioned teams, people, and task fields and draft a formula from them—check the draft references the fields you intended. Spot-check the first result against a manual count before using the chart in a review.
When encountering unclear metric concepts, don't hesitate to ask AI. It is designed to support multi-turn conversations, so you can ask follow-up questions when a metric concept is unclear or request step-by-step help refining a formula; within a single editing session it should retain the conversation context. Confirm context retention in your current product version before relying on it.
For extremely complex data requests, don’t try to overwhelm AI in one sentence! Break the problem into smaller steps and guide it through each part—then compare the intermediate results with figures you already trust before building further.
Start with one plain-language metric request, then check every referenced field and compare the result with figures you already trust.
Click "Read More" at the bottom left to contact a consultant and ask about availability and pricing of the AI metrics feature before committing.
What Hong Kong decision-makers should weigh before adopting AI metric tools
For a Hong Kong firm running cross-department projects, the real procurement question is not whether AI can draft formulas, but who owns metric definitions once they are generated in natural language. If two project managers phrase the same request differently and get different formulas, the numbers presented at your monthly review lose credibility. Before signing, ask the vendor how formula drafts are versioned, reviewed and locked so that a metric like "requirements delivered late" means the same thing in every report.
A practical evaluation step is to run a two-week pilot with controlled, non-sensitive or approved historical sample data. Have one team define three core metrics manually and another team generate them with AI, then reconcile the outputs. Also confirm data handling locations, edition availability and any add-on costs in writing for the plan you are evaluating. For teams also comparing broader collaboration and OA options, our guide to online office and OA platforms in Hong Kong and Macau covers the selection questions that matter most.