SHARPECPA

Sharpe CPA  /  For CPAs  /  Analytic review

Analytic review that actually finds things.

Comparing this year to last year is not an analytical procedure. An expectation is. This guide covers what the standards require in reviews and audits, how to build expectations precise enough to catch a material misstatement, a worksheet that does the arithmetic, and the ratios that matter for fuel distributors and contractors.

The free tools are used at your own risk. They are provided as is, without warranty, and are not accounting, legal or tax advice. All of the work a CPA performs is subject to that person’s or firm’s judgment. Do your own research and verify that the checklist or protocol you are using is valid and up to date. Each tool is built from public sources — see the sources and the terms of use.

What is required

Reviews and audits use the same four steps.

In both, the accountant judges whether the procedure suits the account, considers how reliable the data is, develops an expectation precise enough to reveal a material misstatement, sets the difference that needs no further work, and then compares. A review cannot be performed without analytical procedures; in an audit they are required at planning and at the end, and may be used as substantive procedures.

Review (AR-C 90)Audit (AU-C 315, 520)
WhereEvery material item, including disclosures, and every area of increased risk (.25)Planning, as a risk assessment procedure (315.14b); as substantive procedures where chosen (520.05); near the end, for an overall conclusion (520.06)
Required comparisonsPrior period, adjusted for known changes; plausible financial and nonfinancial relationships; expectations from knowledge of the entity and industry; disaggregated revenue (.26)Chosen to address the assessed risks and assertions (520.05a)
Designing the procedureSuitability; reliability of the data; an expectation precise enough for limited assurance; the acceptable difference set before comparing (.27)The same, with precision enough to identify a misstatement that could be material, alone or with tests of details (520.05)
DifferencesInquire of management; perform other procedures if needed. Corroboration is not required, but responses must be reasonable and consistent with what else is known (.28, .30)Inquire and obtain evidence for management's responses; other procedures as needed (520.07)
DocumentationEnough for an experienced accountant to follow the work: the expectation and its basis, the comparison, inquiries, responses and the accountant's judgment of them (.137)The expectation and factors, the comparison, and the investigation and its results (520.08)

Paragraph numbers are from the current AICPA codifications. AR-C 90.26–.28 were paragraphs .19–.21 in SSARS No. 21 as issued.

The process

Four phases. The first one decides the rest.

The AICPA's practice aid for review engagements describes analytic review as four phases. Almost every weak analytic fails in phase 1: the expectation was last year's number, so any misstatement that looked like a normal change went straight through.

PHASE 1Form the expectationPredict the balance from what you know: drivers, rates, volumes, events. Decide the threshold now, before you look at the recorded amount.
PHASE 2IdentifyCompare recorded to expected. Below the threshold: done. Above it: first ask whether a more precise expectation is available cheaply.
PHASE 3InquireAsk management, and others where risk is higher (project managers, the dispatcher). Look for minutes, board reports and variance analyses that support the answer.
PHASE 4EvaluateQuantify how much of the difference the explanation accounts for. What remains unexplained is a possible misstatement: accumulate it, or do more work.

Precision

The method sets the precision.

Precision is how close the expectation is likely to be to the right answer. The closer it is, the more likely a difference means a misstatement rather than noise. Four things drive it: the method, how disaggregated the data is, how reliable the data is, and how predictable the account is.

MethodWhat it isPrecisionUse it for
TrendThis year against last year or several yearsLowest, and biased toward accepting the recorded amountStable, small accounts; only after adjusting for known changes
RatioRelationships between accounts, or to nonfinancial data or industry figuresLow to moderate; evaluate the dollar effect, not the change in the ratioReceivable days, inventory days, gross margin, expense to sales
Reasonableness testAn explicit prediction from drivers: volume × rate, headcount × pay, debt × interest rateModerate to highPayroll, interest, depreciation, rent, fuel gross profit, excise tax, commissions
RegressionA statistical model of the balance from several predictorsHighest, and measurableLarge, data-rich revenue and cost accounts, mostly in audits

Raising precision

  • Disaggregate. Monthly beats annual; by location or product beats consolidated. A misstatement is small next to an annual total and large next to one month.
  • Adjust the prior year. Prior-year payroll plus headcount, raises and bonus changes is a reasonableness test; prior-year payroll alone is not.
  • Use data the entity cannot easily change. Terminal gallons, bank statements, payroll provider reports, contracts, published rates.
  • Prefer income statement accounts. Their relationships are more predictable than balance sheet accounts.

Data reliability

  • Independent, outside sources are more reliable than the entity's own.
  • Data produced by people who don't keep the books is more reliable than data from the accounting system being tested.
  • Nonfinancial data (gallons, square feet, employees, units) improves precision when it is reliable.
  • Testing for both overstatement and understatement needs data reliable in both directions.
  • Budgets are management's numbers: compare to them with caution.

Tools

Build the expectation, set the threshold.

Pick a reasonableness test, enter the drivers, and send the result to the worksheet. The threshold helper uses the firm's default, the lower of 10 percent of the expectation and half of materiality. That is firm guidance, not a standard: set your own if the risk calls for it.

Reasonableness test

—
Expected amount

Threshold helper

—
Threshold: the difference that needs no further work

Lower the threshold for areas of increased risk. A threshold above materiality cannot catch a material misstatement.

Worksheet

Expectation, threshold, difference, explanation.

One line for each account or relationship. Differences over the threshold need an explanation and the amount it explains; the unexplained remainder is the possible misstatement. Entries stay in this browser only; export to CSV for the file.

Account or relationshipMethodRecordedExpectedThresholdDifferenceStatusExplanationExplainedUnexplained

Ratio library

The relationships worth computing.

Ratios find accounts that have moved out of line; evaluate what they find in dollars, not in the change in the ratio.

Ratio or testHowWhat a change can mean

Explanations

An explanation has to account for the dollars.

Management's answer is not the end of the analytic. It has to be plausible, consistent with everything else you know, and large enough to account for the difference. Quantify it; whatever it does not explain is a possible misstatement.

WeakStrong
“Sales were up.”“Diesel gallons rose 9 percent with the new Colusa contract (contract and terminal volumes seen); at the average margin that adds $61,000 of the $70,000 difference.”
“Timing.”“A $48,000 load billed January 3 was delivered December 30 and recorded in December; bill of lading seen.”
“Prices went up.”“Average rack price rose 11 cents (published rack data); on 1.9 million gallons that is $209,000 of the $230,000 increase.”
“We had a lot of repairs.”“Two tank-wagon engines replaced, $38,000 (invoices seen); considered whether they should be capitalized.”

Common failures

  • Last year's balance used as the expectation, with no adjustment for what changed.
  • Threshold chosen after seeing the difference, or set above materiality.
  • Annual, consolidated figures where monthly or by-location figures were available.
  • Expectation built from the same system and people whose output is being tested.
  • Explanations accepted without being quantified, or inconsistent with other work.
  • Disclosures, and balance sheet accounts, left without any analytic.

What the file should show

  • The expectation and how it was built: method, drivers, sources of data and why they are reliable.
  • The threshold, and when and how it was set.
  • The recorded amount and the difference.
  • Who was asked, when, and what they said; what was seen to support it.
  • How much of the difference the explanation accounts for, and the conclusion on the rest.
  • Any further procedures and their results.

Sharpe CPA portal subscribers record the same work on the Analytical procedures workpaper inside each review plan.

Sources

Where this comes from.

The standards control. The practice aid is nonauthoritative guidance. Thresholds, warning signs and examples on this page are the firm's own and are illustrative.