What drilldown is good for#
The AI Advisor in its default analysis summarizes the situation broadly — e.g., "Your margins dropped by 5 %, mainly in the Cosmetics category."
If you want to know why exactly in Cosmetics, you'll find the "Continue analysis" section under the response with several buttons. Each button is a follow-up question that the Advisor generated specifically for your response (e.g., "Which products in Cosmetics are dragging margins down?"). Clicking the button starts a deeper analysis of the chosen topic.
The buttons are distinguished by a small tag:
- "drilldown" — the Advisor will go deeper into the topic you are currently solving (from Cosmetics margin to specific products)
- "new direction" — the Advisor redirects the analysis elsewhere (e.g., from margins to customers), but with awareness of what it already found
These follow-up questions can be chained — under each new response there are more buttons:
Analysis: Margins and profit
└─ Drilldown: Which products in Cosmetics are dragging margins down?
└─ Drilldown: Which customers stopped buying these products?
└─ Drilldown: What are they buying instead?
The specific button texts change from analysis to analysis — the Advisor generates them based on the response content and your data.
Drilldown and the monthly limit#
Each drilldown counts toward the monthly analysis limit the same as launching the main topic (limits are listed in Plans). Plan therefore how deep you want to go — the Advisor does build on the existing analysis, but each button is billed separately as an AI request.
Margly also limits the number of follow-up drilldowns per one main analysis — in one chain you can go up to 4 levels deeper. This is a technical limit to maintain context quality (the same for all tiers). When you exhaust it, the "drilldown" buttons go gray and Margly shows the message "You have reached the maximum drilldown depth" — you can continue via a "new direction" button (redirect to another topic) or by launching a completely new main analysis.
What drilldown typically does#
The specific depth of analysis is driven by the AI Advisor based on your data and the context of the previous response. From what real responses most often contain:
- Breakdown into segments — categories, products, customers, campaigns, carriers.
- Time breakdown — development in weeks / days / months across the tracked period.
- Comparison — top vs. weak items (typically top 5–10).
- Connections — the Advisor highlights correlations it sees in your data (e.g., "A revenue drop correlates with a decrease in ad budget.").
What drilldown cannot do#
- External reasons — the Advisor doesn't know there were spring holidays or that a competitor's ad affected brand sentiment. It only sees your data.
- Forecasting the future — drilldown analyzes what happened, not what will happen (Margly does not provide statistically valid forecasts).
- Custom questions — you can't phrase a question in your own words. The AI Advisor offers 2–5 context-generated buttons under each response, and you pick from those. Same as with main analyses.
Example: breakdown of a margin drop#
Analysis: Margins and profit
Response: Your net margin is 8 %, which is 3 % less than last month.
The main drop is in the Cosmetics category.
Drilldown: Which specific products?
→ Nike Air Max shoes (in top 3 last year) — stock level 0 for 5 days.
Black XL skirt — you deactivated it in Shoptet on April 12.
Drilldown: Should I return these products to the offering?
→ Nike Air Max shoes: YES — historically generated ~12,000 profit per month (in your store's currency).
Black XL skirt: consider — conversion dropped already 30 days before deactivation.
Create a task from a drilldown#
Just like with the main analysis, every recommended action in a response after a drilldown has an "I want to work on this" button. Clicking creates an action task — Margly records the baseline of key metrics and after 30 days evaluates whether the situation improved. Detail in My Tasks.
What's next#
- Using the AI Advisor — how to launch the main analysis
- My Tasks — turn a drilldown response into a task