What can the conversion slide really support?
Show an observed change without inventing its cause.
“A polished sentence isn’t the same thing as insight.” (opens the source)
The question
You are presenting a fictional retail analytics work sample in India to commercial and product leaders. Give a 90-second opening on whether to roll back a new layout. Conversion means transactions divided by sessions. The before/after records are observational, not a randomized layout test.
| Before | After | |
|---|---|---|
| Transactions | 100 | 64 |
| Sessions | 400 | 400 |
An AI draft says the layout caused a 36% conversion decline. Check the result and the causal claim separately.
“Choose the exhibit, give the opening and defend the recommendation under questions.”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Separate result and cause
What would you do here? Open the move you’d make.
Label the nine-point observed decline and leave cause open.
The strong move. A real observed change does not identify its cause.
Call the change nine percent.
Tempting, but it misses something. The absolute difference is nine percentage points; relative to 25%, the decline is 36%.
See how a strong candidate says it
“Conversion fell from 25% to 16%: nine percentage points, or 36% relative. I would show those observed rates without saying the layout caused the change. These before/after counts do not isolate a layout effect.”
Why it works: A real observed change does not identify its cause.
Step 2: Open without overclaiming
What would you do here? Open the move you’d make.
Ask for a comparable layout analysis before rollback.
The strong move. The panel gets a clear finding and an evidence-based next action.
Demand a full rollback from the pooled counts alone.
Tempting, but it misses something. No comparison in the opening packet isolates the layout from changes in audience or other conditions.
See how a strong candidate says it
“Observed conversion fell from 25% to 16%. I recommend checking comparable audiences before treating that fall as a layout effect. The slide shows a real change; it does not yet show what caused it.”
Why it works: The panel gets a clear finding and an evidence-based next action.
Step 3: Answer the channel split
The panel supplies a channel split. Before: online 90 transactions from 300 sessions; store 10 from 100. After: online 36 from 120; store 28 from 280. The pooled counts remain unchanged.
What would you do here? Open the move you’d make.
Show the split rates beside each period’s session mix.
The strong move. The exhibit must separate audience mix from within-channel rates.
Insist every channel deteriorated because the pooled rate fell.
Tempting, but it misses something. Both channel rates are unchanged in the new breakdown.
See how a strong candidate says it
“Online remains at 30% and store remains at 10% in both periods. Online’s share of sessions fell from 75% to 30%. The mix change reproduces the pooled 25% to 16% result without a within-channel rate decline.”
Why it works: The exhibit must separate audience mix from within-channel rates.
Step 4: Revise the exhibit
What would you do here? Open the move you’d make.
Seek a within-channel layout comparison.
The strong move. A better explanation removes an unsupported claim without inventing the opposite.
Declare the layout definitely works.
Tempting, but it misses something. Unchanged observational channel rates do not establish the layout’s causal benefit either.
See how a strong candidate says it
“I would replace the causal headline with the observed rates and channel mix, then test layout effects within comparable channels. Composition explains this pooled change. That does not prove the layout is harmless or beneficial.”
Why it works: A better explanation removes an unsupported claim without inventing the opposite.
The answer, in short
- Decision
- Use a channel comparison before deciding a layout rollback.
- Because
- The observed pooled decline is explained by the changed session mix.
- What would change it
- A credible within-channel layout test could support a causal decision.
Common traps
- Treating percentage points as percent.
- Replacing an unsupported negative causal claim with an unsupported positive one.
Check your answer
- I used the correct conversion unit.
- I separated result from cause.
- I showed the panel’s new channel evidence.
- I revised the next action without overstating certainty.
Last reviewed Nov 10, 2026