Every credit appraisal makes assumptions. Revenue will grow at 15%. Margins will hold at 16%. Working capital will behave as projected. Capex will produce the expected returns. The borrower will generate sufficient cash to service the debt. The sanction is based on these assumptions. The assumptions are based on the borrower’s projections.
The back-test asks a question that most lending institutions never systematically ask: for loans sanctioned in the last three to five years, how did the borrower’s actual performance compare to the projections on which the sanction was based?
This is not a question about individual borrowers. It is a question about the institution’s credit process itself. If the process systematically accepts projections that are 20% too optimistic, the institution is systematically over-lending by the amount attributable to the optimism. The individual loan may or may not default. The portfolio carries a structural risk that no individual loan review can detect.
How the back-test works
Step 1: Retrieve the original appraisal. For each loan sanctioned in the review period (typically 50 to 100 loans, representing the material exposures), obtain the CMA data, the projected financials, the assumptions memo and the sanction note. These documents contain the assumptions on which the facility was approved.
Step 2: Compare to actuals. For each assumption, the projected value is compared to the actual result for each year since sanction:
Revenue: projected Rs 50 crore, actual Rs 38 crore, variance -24%. EBITDA margin: projected 16%, actual 12%, variance -4 percentage points. DSCR: projected 1.8x, actual 1.1x. Working capital days: projected 75, actual 95. Capex: projected Rs 8 crore, actual Rs 12 crore. Free cash flow: projected Rs 12 crore, actual Rs 3 crore.
Each variance is measured, not described. The measurement is the foundation of the analysis.
Step 3: Identify the pattern. Across the portfolio of 50 to 100 back-tested loans, is the variance random (both positive and negative, averaging close to zero) or systematic (consistently in one direction)?
Random variance suggests that the appraisal process is unbiased but uncertain. Some borrowers over-perform, some under-perform, the errors cancel. The institution is making honest bets that sometimes win and sometimes lose.
Systematic variance, particularly systematic optimism (revenue consistently over-projected, margins consistently over-estimated, working capital consistently under-estimated), suggests that the appraisal process is accepting projections that the evidence does not support. The institution is systematically over-lending.
In Northrop’s experience across mid-market lending portfolios, the back-test typically reveals: 55% to 65% of borrowers over-projected revenue by 15% or more. 40% to 50% missed their margin assumptions by more than 3 percentage points. 25% to 35% consumed more working capital than projected. The systematic optimism bias averages 18% to 22% on revenue projections.
Step 4: Determine why. The bias has identifiable causes. The borrower has an incentive to over-project (higher projections support a larger facility). The relationship manager has an incentive to present the borrower favourably (the sanction generates the business). The appraisal team may not have the sector expertise to challenge industry-specific projections. The process may not require the appraiser to back-test the borrower’s prior projections against actuals (which would immediately reveal whether this borrower has a track record of over-projecting).
Step 5: Apply the findings. If the back-test reveals systematic over-projection of 18%, the institution has two options:
Option A: Apply an 18% haircut to every new borrower’s revenue projection before calculating the debt-service capacity. The haircut is not scepticism. It is evidence-based calibration. A borrower projecting Rs 50 crore is evaluated at Rs 41 crore for facility sizing.
Option B: Require the appraisal team to back-test each new borrower’s prior projections before accepting the current projection. A borrower who over-projected by 25% in each of the last three years should not have their current projection accepted at face value.
What the back-test reveals about the institution
The back-test does not evaluate individual borrowers. It evaluates the institution’s credit process. If 60% of borrowers over-projected, the question is not “why did these borrowers miss?” The question is “why did our credit process accept projections that 60% of borrowers could not deliver?”
The answer typically involves: insufficient challenge of borrower projections during appraisal, absence of back-testing as a standard appraisal step, relationship pressure to sanction (the RM wants the deal done), sector-level optimism (the industry is growing, so the borrower must grow too) and a process that evaluates the projection’s internal consistency (do the numbers add up?) rather than its external credibility (has this borrower, or this industry, ever achieved these numbers?).
The institution that back-tests its appraisals gains something more valuable than any individual loan review: it gains knowledge of whether its credit process is producing reliable decisions. That knowledge is unavailable from any other source.
Northrop applies this test across lending portfolios. Business@NorthropIndia.com | +91 92899 25657.