How grades are made

Laying out the financial statements of major Korean and US companies one by one is tedious even for professionals. Jupick scores public financial data with the same yardstick for every company, so you can judge the current financial strength of a company without reading its statements. Each company is scored on three areas (earnings, financial health and dividends), and the scores are combined into a 0 to 100 overall score and a letter grade from S to D.

One thing first: a grade is not a forecast of the share price. It is a snapshot of past and current financial figures that have already been disclosed, and a high grade promises no rise in price and no investment return. Jupick gives no trading advice and no price estimates of any kind. This page explains how a grade is calculated. It says nothing about what to invest in. Background on the service is on the About page.

The overall score mixes the three area scores with different weights. The company’s ability to make money counts most, and the financial health that supports it and the returns to shareholders count too. The weights are earnings 40%, financial health 35% and dividends 25%.

Earnings (40%) measures the strength of the core business. ROE is the profit made on shareholders’ capital, operating margin is the share of sales left as operating profit, and revenue growth is the change in sales from a year earlier. When all three are high together, the company is running a profitable business that keeps getting bigger.

Financial health (35%) measures staying power. The debt ratio is debt relative to equity (lower is safer), the current ratio is assets that can be turned into cash relative to debt due within a year, and interest coverage is how many times operating profit covers interest costs. However much a company earns, a risky debt structure pulls this area down.

Dividends (25%) measures whether money comes back to shareholders steadily. Dividend yield is the annual dividend relative to the share price, the payout ratio is the share of profit paid out as dividends, and the dividend streak is how many years in a row dividends have been paid. Dividends are treated as a sign that the company has the room and the will to share the profit it actually earns.

The Korean site also has a one-page sheet with the meaning, formula and current Korean market median of every metric, including these nine.

Each metric is first converted into a score from 0 to 100. For every metric a few thresholds are set in advance, and values between two thresholds are scored on a straight line between them, so the score never jumps at a boundary. Values beyond the outer thresholds are clamped to the top or bottom score. However good a value is, it never scores above 100, and however bad, it never falls below the floor of that metric.

Metrics fall into three scoring directions.

• Higher is betterROE, operating margin, revenue growth, current ratio, interest coverage, dividend yield and dividend streak. The score rises with the value and stops at 100 once the value is high enough.
• Lower is betterDebt ratio. The lighter the debt, the higher the score. Only the direction is reversed and the calculation is the same. A negative value does not mean light debt but negative equity, so it gets the lowest score of this metric.
• Best within a rangePayout ratio. Paying out around a third of profit scores highest. Paying out too little (stingy with shareholders) or too much (paying more than the company earns may not last) both lower the score. There is a floor, so it never drops to 0.

The thresholds are listed below as they are. If only the method were described and the numbers hidden, you could not check the arithmetic yourself, and a score you cannot check is a matter of faith. Most grading services show the grade and hide the calculation. There is nothing to hide here. Join the values in the table below and you get exactly the metric scores shown on the stock pages.

How to read it: a company with an ROE of 12.4% sits between 10% (65) and 15% (85), so the straight line between them gives 74.6. On each stock page, a thin bar under every metric shows that score, and wide screens also show a note such as ‘the 10% to 15% range’. The left end of the bar is always the low score, and only for the payout ratio does the score rise as the value gets closer to 35%. The payout ratio is scored as 100 − |value − 35| ÷ 40 × 100 and never below 10.

This is an absolute scale based on fixed thresholds. The same thresholds apply to every company, so one company’s score never moves with other companies’ figures on the day. The thresholds themselves were set with reference to the strong and weak levels commonly seen in real companies, so the scores and grades stay close to common sense.

What happens when data is missing? The grader needs a number for all 9 metrics, so a metric cannot be left empty even when the source gives no value. When the source is silent, a neutral fill-in value is used. It is not just any ‘average company’ number but the single point on that metric’s scoring curve that gives exactly 50.

ROE

7.0%

Operating margin

7.0%

Revenue growth

3.0%

Debt ratio

110.0%

Current ratio

128.0%

Interest coverage

4.2x

Dividend yield

1.8%

Payout ratio

15.0%

Dividend streak

4.2 yrs

Neutral fill-in value of each metric. All nine score 50.

Why 50 and not 0. Missing values used to be set to 0, and this page once called that ‘conservative’. That was wrong, because the meaning of 0 flips with the direction of the metric. A lower debt ratio is better, so writing debt as 0 does not judge the company conservatively. It gives the top score of a debt-free company. Failing to find a filing turns into a bonus. Setting ROE to 0, on the other hand, makes the company look like one that made no profit. One rule becomes a reward for one metric and a penalty for another, and neither is something we actually learned about the company. 50 is the only point that is neither.

Why not the industry average. Filling gaps with the median of all companies is common. Jupick does not do it. The watchlist is a group of the largest companies by market cap, so its median already sits high. An ROE of 15% scores 85, and using a median near that would hand a high score to a company whose data never arrived. ‘Unknown’ would turn into ‘excellent’. Claiming a company does as well as the big names is the same invention as filling with 0, just in the other direction.

Metrics with fill-in values are not hidden. In the area scores of a stock page, such a metric shows a hyphen (-) instead of a number, with a one-line reason below it. There are three reasons. Not reported means the source answered but the item was not there. No data means the source did not answer in this update. Not applicable means the metric does not make sense for that kind of business. The first will not appear tomorrow either, the second may be filled in the next update, and the third has nothing to wait for. You would do something different in each case, so they are not lumped into one word.

Not applicable is used narrowly so far. It has been used since September 4, 2026, and only for one metric, the interest coverage of Korean companies. It is not decided from an industry list. It is decided from one fact in that update’s statements: interest expense was found but there was no operating profit line at all. Financial companies have no operating profit line in their income statement, so the division does not make sense rather than the company failing to report. US companies come from a different source that does not make this distinction yet, and the same case shows as not reported.

So some not-applicable cases still show as not reported. The current ratio of banks is one. Banks do not split current assets and current liabilities that way, so it is closer to not applicable, but the screen says not reported. That is because a table of which metrics do not apply to which industries has not been made yet. This is written down knowing it is not exact.

A hyphen does not always mean ‘a neutral fill-in was used’. Strictly, it means the value did not come from the source in this update. The number actually used is one of two. If a value was confirmed in the previous update, that value is kept. Only when there is none is the neutral fill-in from the table above used. Overwriting a known value with a fill-in would throw away what was already confirmed. So the score of a hyphenated metric may be 50, or it may come from a real value measured in an earlier update. Either way, it was not confirmed today, and that is all the hyphen says. A not-applicable mark says one more thing: the value will not be filled in the next update either.

A capped value is not missing. A dividend streak shown as ‘15 years or more’ does not mean there is no data. It means the record was checked back only that far. It is a confirmed lower bound, not a blank, so no hyphen is used.

Not every 0 in the dividend metrics is missing. If the dividend filings were read successfully and contained no dividends at all, that 0 is a confirmed fact: the company pays no dividend. So the dividend yield, the payout ratio and even the dividend streak are scored as 0, with no hyphen. Using fill-in values here would give a non-paying company a dividend score and put a 0% yield next to a 4.2-year streak in the same area, which makes no sense. On the other hand, an update where the filings could not be read at all is no evidence of no dividend. In that case 0 is not written and the value is treated as missing. A company that has paid dividends for 20 years should not drop to 0 years because one lookup failed. The two zeros look the same on screen, so unless they are told apart at the moment the source answers, there is no way to tell them apart later.

The payout ratio was chosen differently. The other eight have one-way curves, higher is better or lower is better, so there is only one point that scores exactly 50, and the fill-in value is calculated rather than chosen. The payout ratio is best within a range, around a third of profit, so both 15% at the low end and 55% at the high end score exactly 50. Neither could be chosen by taste, so for a long time this value was left empty and scored as 0%, and that was worse. 0% sits far down the curve at 12.5, so companies that pay dividends but lacked only this figure got lower dividend scores than they should. That goes directly against the premise of this chapter: silence should not be punished.

So instead of taste, numbers we had already counted decided it. As of September 3, 2026, 324 of the stored major companies (81%) had an actual payout ratio, with a median of 31.19%. More than half, 55.2%, were below the ideal point of 35%. Measured from the median, 15% is closer to the companies we cover than 55%. So the fill-in for this metric is 15%. As long as a gap is filled with a single constant, the choice cannot be entirely free of judgment, but at least it was chosen by our own count and not by a coin toss. A payout ratio confirmed in the previous update takes priority over this 15%. For US companies, though, it is restored only in updates where the source did not answer at all. If the answer was fine and only this line was empty, the next update will likely be empty too, so the old number is not restored to look like a new measurement.

Rough edges that remain. Metrics with neutral fill-in values still count in the area average. An area where one of three metrics is a fill-in looks just as solid on screen as an area where all three were measured. That is why each stock page shows, under the overall grade and under each area, how many metrics have a value.

The metric scores add up to area scores. An area score is the average of its three metric scores. Multiplying the three area scores by 40%, 35% and 25% and adding them gives the overall score from 0 to 100. Finally, the score shown on screen is rounded down. For example, 84.6 is shown as “84”, never rounded up to 85 (the S boundary). This keeps a displayed number from being read as one grade higher than it really is.

The overall score becomes one of five letter grades by the cutoffs below. Going from the top, the first grade whose condition is met is given.

The short phrase next to each grade is just a name for that score range. The gaps between cutoffs are not even because the thresholds are tighter at the top (S and A), where truly strong finances are picked out. Two A grades at 71 and 84 can feel different, so Jupick always shows the letter grade and the number together.

Valuation metrics such as PER (price to earnings) and PBR (price to book) are not part of the overall grade. They are shown separately on the stock pages. The grade and the valuation answer different questions.

The grade asks “is this a financially strong company?” (financial strength). PER and PBR ask “is the share price high or low compared with the company’s profit and assets?” (price). A strong company is not always cheap, and a cheap share is not always a strong company. Mixing the two into one score lets each signal hide the other, so “strong finances, expensive shares” and “ordinary finances, cheap shares” could end up with similar grades. So Jupick grades financial strength only and shows the price figures as they are, for you to judge.

Reading this chapter as “the share price never touches the grade” would not be accurate, though. There is one exception, described right below.

Exception: Dividend yield includes the share price

What was left out of the grade is PER and PBR, not the share price altogether. As noted in chapter 2, Dividend yield is the dividend relative to the share price, and it is one of the nine scored metrics. So one metric whose definition includes the share price counts toward the grade. Its weight is about 8.3%, because the dividend area weight from chapter 2 is shared by the 3 metrics of that area.

Whether the share price actually moves this metric depends on the market. US companies use the dividend divided by the share price, so the ratio is recalculated when the price changes even if the announced dividend stays the same. Korean companies use the cash dividend yield stated in the dividend report on DART as it is, so it is not recalculated when the price moves and stays the same until the next filing.

So if the Dividend yield score of a US company went up, the company may have raised its dividend or the share price may have fallen. To tell which, look at the dividend and the share price together on the stock page. Each stock page also shows how many of its metrics use the share price, in the section where the share price enters the grade.

This metric stays in the grade because dividing by the share price is the standard way to measure how much dividend an investor actually receives. Changing the definition would make the number impossible to compare with other sources. So instead of removing it, this note explains it.

This method is not perfect. Please keep the following limits in mind when reading a grade.

• Delays and errors in public dataJupick does not pay for data and uses public sources (exchange prices, electronic disclosures and public financial data). If the source is late or wrong, the grade is affected too. Financials are refreshed on the disclosure cycle, not every day.
• A grade is not a trading signalAn S grade is not a signal to get in, and a D grade is not a signal to get out. A grade only summarizes the current financial state and is separate from where the share price goes. A high grade promises no investment return.
• Industry differencesThe same yardstick applies to every company, so industries where heavy debt is normal (such as finance and utilities) cannot be compared in exactly the same way as others. Traits of an industry that do not show in the numbers are not captured by the grade.
• Based on the past and presentA grade rests on past results and the current financial state. Future factors that are not yet in the numbers, such as new products, a turn in the industry or management changes, are not reflected.
• The effect of missing dataAs explained in chapter 4, a metric the source did not provide is scored with the value confirmed in the previous update if there is one, or with the neutral fill-in otherwise, and it is shown as - on screen. The fill-in does not push the grade up or down, but it does mean there is less confirmed evidence about that company. A grade with many fill-ins is less certain at the same score. If a figure looks different from the real one, please report it as described on the About page and we will check and fix it.