PromptBase
PromptBase SelectGet Select
Close icon
General
Home
Marketplace
Create
Hire
Login
Chat
Sell
Explore

Prompt Details

Model
(claude-5-5-sonnet)
Token size
2,520
Example input
[WHAT LOOKS WRONG OR IS MISSING, AND HOW THE DATA WAS COLLECTED IF KNOWN]: Delivery status is missing for two orders. [WHAT YOU WANT TO MEASURE OR FIND OUT]: Find the possible late-delivery rate for all 100 orders. [ANONYMIZED ROWS, COLUMN NAMES, COUNTS, OR SUMMARY STATISTICS]: 100 eligible orders, 98 recorded statuses, eight confirmed late. [DOMAIN, DECISION OR PROTOTYPE, SUCCESS CRITERION IF KNOWN, AND TOOLS YOU USE]: Dispatch operations. Use Excel to decide whether at most 10% of orders were late.
Example output
**What you can conclude: Bound only (which happens to settle the decision)** Of 100 orders, 98 have a recorded status and 8 are confirmed late. The 2 unknown orders could each be late or not late, so the late rate for all 100 orders has to lie between two extremes: - **Best case** (both unknown orders were on time): 8 / 100 = **8%** - **Worst case** (both unknown orders were late): 10 / 100 = **10%** Your target is "at most 10% late," so the test is late count ≤ 10. Every possible outcome (8, 9, or 10 late orders) passes. **The target is met, even in the worst case**, and no assumption about why the statuses are missing is needed. This is a direct count of possibilities, not an estimate. One thing to note: the worst case lands exactly on the 10% line. That passes under "at most," but there is no margin. If the real rule is "under 10%," the worst case would fail and the result would be unresolved. **Recommended method** Use a **bounding calculation** (count the unknowns first as "no," then as "yes"). It fits because the total is known (100), only 2 orders are unknown, and the question is a threshold check. Imputation or modeling would add assumptions you don't need and could only land inside this 8–10% range anyway. **What must be true** - The 100 orders are the complete set you care about, with no orders missing from the list entirely. You said "100 eligible orders," which supports this, but I can't verify it from the numbers alone. - The 8 "confirmed late" orders really are late under your definition, and the 98 recorded statuses are accurate. - The 2 unknown orders are counted among the 100. If they are not, the denominator changes. **How to apply and check it in Excel** Assume statuses are in column B (values like "Late", "On time", blank for unknown). 1. Late count: `=COUNTIF(B2:B101,"Late")` → 8 2. Unknown count: `=COUNTBLANK(B2:B101)` → 2 3. Total: `=COUNTA(A2:A101)` (using an order ID column) → 100 4. Lower bound: `=late/total` → 8% 5. Upper bound: `=(late+unknown)/total` → 10% 6. Verdict: `=IF((late+unknown)/total<=0.1,"Target met in all cases",IF(late/total>0.1,"Target missed in all cases","Unresolved"))` **Check that could change the recommendation:** confirm that 8 + 2 + on-time count = 100 (so 90 on time). If the on-time count is 89 or 91, a row is duplicated or lost, and the bounds need recalculating. **What to do next** Find the status of the 2 unknown orders from the source records (dispatch logs, driver confirmations, or customer system). Once either is known to be on time, the upper bound drops to 9% and you have margin. Also confirm whether the policy is "at most 10%" or "below 10%," since that is the only thing that could flip the verdict. Do you know whether the target is "at most 10%" or "below 10%"?
Claude

Analytics Method Finder

Add to Cart
Instant accessInstant access
Usage rightsCommercial use
Money-back guaranteeMoney‑back
By purchasing this prompt, you agree to our terms of service
CLAUDE-5-5-SONNET
Tested icon
Guide icon
4 examples icon
Free credits icon
Can your incomplete data support a decision? Get a plain-English analysis of what you can conclude, which statistical method fits, and what to verify next. Use it for missing records, capped measurements, uncertain ranges, or conflicting sources. Describe the goal and business context, with an anonymized sample if available. Get a worked calculation when the data allow it and a practical next step when they don't.
...more
Added 15 hours ago
Report
Browse Marketplace