Data Analysis

Anomaly Diagnosis

Analysis Intent Routing

AE analysis intent routing: automatically recognize analysis intent and route it to the right ae-cli command or MCP tool, prefer reusing existing reports and dashboards, and make Builder mandatory so QP is never hand-written.

IndustriesAll industries

Agent · Skills Live

skill · loadedindustry+100

Run this Skill on my data

User path

step to step

Analysis readyevidence attached

The AE Analysis Intent Skill is an analysis-intent routing capability in ThinkingAI's industry Skill library. Starting from a natural-language analysis request, it helps data and operations teams route automatically to the right ae-cli command or MCP tool, with project-gating confirmation, priority reuse of existing reports and dashboards, mandatory Builder QP generation, an absolute-contribution attribution algorithm, and recommended anomaly-investigation paths, moving from "not knowing which command to use" to a precise mode where intent routes automatically to the right tool. This is ThinkingAI's decade of industry experience, packaged into ready-to-use Skills.

Live walkthrough

AgentSave as AgentExport session

ae-analysis-intent

Payment revenue is up about 10% over the past two weeks. The UA team says new channel D drove it (up more than 3x) and wants me to add budget to D. Something feels off. Help me confirm who actually drove this lift.

Intent parsing: decompose the question into three sub-intents: trend confirmation + dimensional attribution + decision support

Project gate: verify a sample card game DEMO (project_id=5)

+list_reports searches existing payment reports

Reused "Daily Payment Amount," no duplicate ad-hoc built

Builder generates an event analysis QP, grouped by channel, pulls last 14 days vs. prior 14 days

+10.1 ($10K) QoQ (+9.2%)

Absolute-contribution attribution: rank by absolute change, not relative growth

Channel B contributed 50%, channel D contributed 41%

First, the premise behind whether to add budget: the main driver of this lift is not channel D, it's the established channel B: of the +10.1 ($10K) lift, B accounts for 50% (+5.1) and D for 41%. The "up 3x" is the visual effect of a small base. The way your loosely worded question got decomposed and executed is below too.

1. How your question was parsed

Your words	Parsed intent	Routed to	Metric / dimension
"Payment revenue is up about 10%"	Metric trend confirmation	Reuse existing report "Daily Payment Amount"	Last 14 days vs. prior 14 days
"Is new channel D driving it?"	Dimensional attribution	Builder-generated event analysis QP → absolute-contribution attribution	Group by `channel`, rank by absolute change
"Should I add budget?"	Decision support	Attribution conclusion + channel-quality cross-check	Lift structure × quality metrics

← Scroll to see more

2. Attribution result: the main driver is channel B

Channel	Prior 14 days ($10K)	Last 14 days ($10K)	Absolute contribution ($10K)	Relative growth
Channel B	50.2	55.3	+5.1 (50% of lift)	+10.2%
Channel D	1.2	5.3	+4.1 (41% of lift)	+342%
Other channels combined	58.0	58.9	+0.9 (9% of lift)	+1.6%

← Scroll to see more

By relative growth, D's +342% is certainly eye-catching; but attribution has to be ranked by absolute contribution: D's growth came off a 1.2 ($10K) base, and in absolute terms it's still less than one routine volume push from B. This is the base-rate fallacy: a high growth rate on a small base is the easiest thing to mislead a budget decision.

3. Three traps this routing helped you avoid

Didn't build an ad-hoc from scratch: searched existing assets first, reused "Daily Payment Amount," so the metrics match what your team looks at day to day

Didn't hand-write the QP: event analysis is forced through Builder up front, so event names and fields are all validated against metadata

Didn't attribute by relative percentage: absolute-contribution ranking corrected the gut conclusion of "D is the driver" to "B is the driver"

Conclusion

Budget recommendation: first protect channel B's steady spend, since it's the backbone of the lift; D can take a small incremental add for validation, but before scaling it, confirm whether its next-day retention and payment rate hit the bar (I'd suggest running the "Channel Quality Analysis" Skill for a six-dimension assessment). Don't set budget off a +342%.

On your data

That was a simulated run

Leave your work email and we will run a live walkthrough on your real business data.

The problem

Data analysis is a high-frequency daily task, yet more than 50% of teams are inefficient on the path from intent to tool: they know they want the DAU trend but aren't sure whether to use a report query or Ad-hoc, and they know they want to analyze a payment drop but aren't sure whether to use event analysis or property analysis. More subtle is the base-rate fallacy in contribution attribution, where judging the main driver by relative percentage leads to wrong conclusions. Existing reports and dashboards are also often ignored, with every analysis starting from a fresh Ad-hoc query.

What it does

Automatic intent routing: project_id gating, prioritized report/dashboard search, Builder QP generation, then Ad-hoc execution, automatically choosing the optimal path by intent

Existing assets first: search existing reports and dashboards first and query a matching asset directly, without wasting resources on duplicates

Builder mandatory first: the four model types (event/retention/funnel/prop_analysis) must go through Builder first, with no hand-written QP and no guessed parameters

Absolute-contribution attribution: rank dimension attribution by absolute contribution (not by relative percentage), eliminating wrong attribution from the base-rate fallacy

Two-tool priority: ae-cli first, with te-mcp only as a fallback when ae-cli lacks the capability or fails repeatedly for non-parameter reasons

When to use it

You need to query a metric's trend or anomalous change

You need dimension attribution for a metric change

You need to run event analysis, retention analysis, funnel analysis, or property analysis

You need to view data from an existing report or dashboard

You need to investigate a data-anomaly root cause (drilling down by channel, version, time, and segment)

You need to confirm the project_id before running project-level queries

In the field

Case

A game operations team · payment-revenue drop attribution

The team found payment revenue down 15% month over month and assumed the main driver was new channel D's 400% growth. After searching existing payment reports, the Skill computed absolute-contribution attribution: channel B went from 50 to 55, an absolute contribution of +5 or 50%, while channel D went from 1 to 5, an absolute contribution of +4 or 40%. The conclusion was corrected to channel B as the main driver, and the strategy shifted from increasing D spend to maintaining B's stability.

FAQ

How does this differ from ae-analysis?

ae-analysis is the analysis-platform operations manual, while ae-analysis-intent is the routing rules from analysis intent to tool. The former knows how to use a command, the latter knows which command to use.

What is absolute-contribution attribution?

Dimension attribution ranked by absolute change rather than relative percentage, avoiding decisions misled by a high growth rate on a small base.

What is te-mcp?

te-mcp is AE's MCP-protocol tool, used as a fallback for ae-cli only when ae-cli lacks the capability or fails repeatedly for confirmed non-parameter reasons.