# Gotchas Failure modes that produce a report that looks right and is quietly wrong. Those are the dangerous ones — owners forward these reports to banks, boards, and accountants. --- ## Gotcha: partial period compared against a full one **Why it matters:** On July 27, month-to-date sales versus last full month looks like a 20% collapse. It isn't. Owners have made staffing decisions off this mistake. ### Bad ``` Sales — USD 21,400, down 19% vs. June ``` June was a full month. July is three days short. The comparison is meaningless and the owner reads it as a crisis. ### Good ``` Sales — USD 21,400 month-to-date through Jul 27 (June through the 27th: USD 19,800, up 8%) ``` Same data, honest comparison, opposite conclusion. --- ## Gotcha: the same sale counted twice **Why it matters:** A Shopify order settled through Stripe and recorded in QuickBooks exists in three systems. Summing across connected sources inflates revenue, sometimes by double. ### Bad ``` Pull revenue from QuickBooks + Shopify + Stripe, sum all three. Report shows USD 129,000. Actual revenue was USD 43,000. ``` ### Good ``` Pick one system as the revenue source. Say which: Revenue — USD 43,000 (source: QuickBooks; Shopify and Stripe used for fulfillment and fee detail only) ``` Rule of thumb: the books are the revenue source when connected. Processors supply fees and timing. Storefronts supply product-level splits. --- ## Gotcha: a group disappears instead of showing zero **Why it matters:** If Portland sold nothing this week and the row vanishes, the owner assumes the data is broken and goes looking. A zero is information; a missing row is noise. ### Bad ``` Sales by location: Salem USD 18,200 Eugene USD 19,000 ``` Portland existed last week. Where did it go? ### Good ``` Sales by location: Salem USD 18,200 Eugene USD 19,000 Portland USD 0 (no sales recorded this period) ``` --- ## Gotcha: grouped rows don't tie to the total **Why it matters:** An owner who adds your columns and gets a different number stops trusting every report you produce after that. ### Bad Publish the total and the breakdown without checking they agree. ### Good Check the sum before delivering. If they genuinely don't tie, say why: ``` Sales by location sums to USD 41,900. Total sales is USD 43,200. The USD 1,300 gap is unallocated invoices with no location set. ``` Naming the gap builds more trust than hiding it. --- ## Gotcha: re-interviewing about a saved report **Why it matters:** The whole promise of this skill is "describe it once." Asking the same five questions every month breaks that promise and the owner stops using it. ### Bad ``` Owner: "Run the Monday report." Claude: "Sure — which metrics would you like, and what period should it cover?" ``` ### Good ``` Owner: "Run the Monday report." Claude: [reads saved_reports.md, finds it, runs it] "Monday operating pack — week of Jul 20..." ``` Always read `saved_reports.md` first. --- ## Gotcha: a plausible number filled in for a missing one **Why it matters:** This is the most serious failure in the skill. A report is a document owners act on and forward. An invented figure that looks reasonable can survive several hands before anyone catches it. ### Bad ``` Labor — approximately 30% of revenue ``` Where did that come from? Nowhere. Payroll accounts didn't return. ### Good ``` Labor — n/a (QuickBooks payroll accounts unavailable this run) ``` Never estimate, never round toward a guess, never carry last period's figure forward silently. --- ## Gotcha: leading with the table **Why it matters:** Owners asked for a report because they want a decision. A wall of numbers hands the analysis back to them, which is the job they were trying to hand off. ### Bad Open with a 40-row table, then a summary underneath it. ### Good Open with the one sentence that matters, three headline metrics, three findings. Everything else goes in the workbook.