SkillsSkill 1 of 5
Show your own thinking, not AI polish
Check every claim on your résumé against its baseline and your own part, and drop anything you can’t explain.
Worked example
Your résumé says you drove a 40% increase. Can you defend it?
The bullet reads well. Check it against the numbers before an interviewer does.
The question
Larkhaven is a fictional regional grocery-delivery company. As a summer intern you helped email 2,000 customers who had stopped ordering; 1,000 similar lapsed customers, picked at random, got no email. An AI tool rewrote your résumé bullet from your notes.
Drove a 40% increase in customer reactivations by designing a targeted win-back email campaign.
| Customers | Ordered again | Rate | |
|---|---|---|---|
| Same month last year, no email | 2000 | 50 | 2.5% |
| This year, emailed | 2000 | 70 | 3.5% |
| This year, random holdout, no email | 1000 | 30 | 3.0% |
Invented data. Last year’s row is a different group of lapsed customers. The holdout was drawn at random from this year’s list.
- You, at the timeManager chose the offer. I built the send list, wrote two subject lines and set up the holdout and the tracking sheet.
“The interviewer reads your bullet aloud: “Drove a 40% increase in customer reactivations.” Walk me through it. What did you do, and what can you claim?”
Your turn
Say your answer out loud, or jot notes below. Both work. Or skip to the worked answer.
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: Say what you actually did
What would you do here? Open the move you’d make.
Walk through it as a team win: we designed it, we sent it, we got 40%.
Tempting, but it misses something. It dodges the question. Your notes say your manager chose the offer and you built the send list, the subject lines and the holdout. “We” hides exactly what the interviewer asked about.
Say you designed the campaign, since the bullet already says so.
Tempting, but it misses something. The bullet says “designing,” but your manager chose the offer. One follow-up about the offer and the claim falls apart.
Name my part: my manager chose the offer; I built the send list, wrote the subject lines and set up the random holdout.
The strong move. It matches your own notes and separates your work from the team’s before the interviewer has to ask.
See how a strong candidate says it
“My manager chose the offer. I built the send list, wrote the subject lines and set up a random holdout so we could measure the email properly.”
Why it works: It separates your work from the team’s before anyone has to ask.
Step 2: Recompute the claim
What would you do here? Open the move you’d make.
Recompute it: 40% is 70 against 50 from a different group last year. Against the holdout it’s 3.5% vs 3.0%, about 10 extra orders.
The strong move. The holdout was drawn at random from this year’s list, so it shows what would have happened anyway: 3.0% of 2,000 is 60, and 70 − 60 = 10.
Defend the 40%: 70 is 40% more than 50, so the math is right.
Tempting, but it misses something. The arithmetic is right; the comparison isn’t. Last year’s 50 came from a different group of lapsed customers, so it can’t show what the email caused.
Switch to the lift against the holdout: 3.5% over 3.0% is about a 17% increase.
Tempting, but it misses something. Closer, but it still leads with the biggest-sounding number. On 2,000 emails the gap is half a point, about 10 orders; that’s the honest scale.
See how a strong candidate says it
“The 40% compares 70 reactivations with 50 from a different group last year. The fair comparison is the holdout: 3.5% of emailed customers ordered again against 3.0% who got nothing. That’s half a percentage point, about 10 extra orders out of 2,000.”
Why it works: It replaces a flattering baseline with the comparison the test was built for.
Step 3: Say what the number can’t show
What would you do here? Open the move you’d make.
Call it a 10-order win: the holdout proves the email worked.
Tempting, but it misses something. Only 30 holdout customers ordered again. A gap of 10 orders on numbers that small could easily be chance, so “proves” overstates it.
Say the lift is small and uncertain: 30 holdout orders are too few to rule out chance.
The strong move. It shows you know how much evidence 70 against an expected 60 is, not just which way it points.
Say the test failed because the lift wasn’t significant, so the email didn’t work.
Tempting, but it misses something. Too far the other way. Uncertain isn’t zero: 3.5% against 3.0% is a small lift that a bigger test could confirm or rule out.
See how a strong candidate says it
“Only 30 holdout customers ordered again, so a gap of 10 orders could easily be chance. I’d call it a small, uncertain lift, not a result.”
Why it works: It shows you understand how much evidence you have, not just which way it points.
Step 4: Claim what you can defend
What would you do here? Open the move you’d make.
Rewrite it: set up a win-back test with a random holdout across 3,000 lapsed customers; small lift, 3.5% vs 3.0%; recommended a larger test.
The strong move. Every number survives a follow-up: 2,000 emailed plus 1,000 in the holdout is 3,000, and the rates come straight from the table.
Keep the 40% on the résumé and explain the caveats if asked.
Tempting, but it misses something. If the bullet needs caveats to survive, it isn’t defensible. The interviewer read it aloud; the 40% is the claim being tested.
Take the project off the résumé; a small result looks weak.
Tempting, but it misses something. Setting up a random holdout across 3,000 customers is real work, and recommending a larger test is good judgment. Hiding it loses your best evidence.
See how a strong candidate says it
“I’d rewrite the bullet: “Set up a win-back email test with a random holdout across 3,000 lapsed customers; found a small lift, 3.5% against 3.0%, and recommended a larger test.” If a bigger test held the gap, I’d claim the lift.”
Why it works: A smaller, true claim you can explain beats a big one that falls apart at the first follow-up.
The answer, in short
- Decision
- Drop the 40% claim. Describe your part, the holdout comparison and what you recommended.
- Because
- The 40% compares two different groups; against the holdout, the email added about 10 orders out of 2,000, which may be chance.
- What would change it
- A larger or repeated test that keeps a clear gap over a random holdout.
A real interviewer will ask about your own projects. Use your real numbers and your real part in them. Don’t memorise an invented story.
Common traps
- Letting AI polish a claim you can’t explain.
- Comparing against last year instead of the holdout.
- Claiming the team’s result as yours.
- Hiding a small result instead of saying what you learned.
Check your answer
- Did I say what I did, separately from the team?
- Did I compare against the right baseline?
- Did I say how sure the number is?
- Can I explain every word of my résumé without help?
Want feedback on how you’d answer? Practice “Show your own thinking, not AI polish” with an AI interviewer on Socratify.
Last reviewed Nov 9, 2026