When a metric goes blank in only one market: the hole an average hides

On the evening of August 30, 2026, Jupick counted the blanks among its major stocks metric by metric. Interest coverage alone was blank for 230 stocks (57.5%). That number itself had been known for several days.

The only new step that day was splitting the 230 into Korean and overseas stocks. It came to 196 (98%) of the major Korean stocks and 34 (17%) of the major overseas stocks.

Once split, the two numbers turned out to be completely different in nature. The overseas 34 is the usual level. There are that many cases where this metric does not hold at financial companies, and that is normal. The Korean 196 is not that kind of number. Only four Korean stocks had a value for this metric.

This article records the method used to find that hole, and the more uncomfortable fact learned after finding it.

Jupick's collection process has a safeguard meant to catch a metric vanishing wholesale. It raises a warning when an item is blank in every stock updated. The threshold is set at every stock to avoid noise. A single stock going blank happens all the time, and sounding a warning for it would mean nobody reads the warnings.

But this safeguard counted Korean and overseas stocks in one bucket. Even if the Korean side died entirely, overseas stocks kept the count up, so the every-stock condition could never be reached. Whether half the Korean stocks died or all of them, the total stayed positive.

If the data for both markets came from the same place, counting this way would not be a big problem. The problem is that the data sources are entirely different. Korean data comes from the disclosure system, overseas data from a separate financial data provider. One side failing and the other side failing are unrelated events, yet they were counted on a single counter.

This is usually how averages and totals hide failures. Summarize two different populations in one number, and an anomaly in one is diluted by normality in the other. That is why, before building a summary, it pays to see first what is being combined with what.

When Jupick first saw that many stocks had blank interest coverage, it read this as a sector issue. The metric is operating income divided by interest expense, and at financial companies such as banks and insurers, interest expense is not a cost but the body of the business, which blurs the meaning of the division. So the assumption was that the blanks clustered among financials.

The explanation was plausible. In fact, the 34 overseas stocks broadly fit it. Applied to Korean stocks, the same logic does not fit.

Of the 196 blank Korean stocks, 44 fell under a financial sector. The other 152 are neither banks nor insurers. Counting again by sector makes it clearer. In semiconductor manufacturing all five relevant stocks were blank, in pharmaceutical manufacturing all five, in electronic components manufacturing all nine, and in special-purpose machinery manufacturing all fourteen. These companies clearly have interest expense.

This is where a plausible explanation becomes dangerous. When it fits some cases well, the rest goes unchecked. To test an explanation, you look for places where the result should differ from what it predicts. If sector were the cause, non-financial sectors should have been filled in, and they were not.

To tell whether collection had failed or the data source simply had no value, a comparison is needed. A good control is a value that went through the same path but could produce a different result.

Here the current ratio fit that role. For Korean stocks, these two metrics are calculated together from a single fetch of the same disclosure data. If the data had not been received, both should be blank.

The measurement showed the current ratio blank for only 30 (15%) of the major Korean stocks. In other words, the data was arriving normally. That day, Jupick read this to mean that after receiving the financial statements it was reading current assets and current liabilities fine, but failing to find only the interest expense item in the same document.

This one comparison shrank the list of possibilities considerably. It was clear that the cause was neither a data source outage nor a sector trait. What remained, Jupick wrote that day, was a problem in its own rules for extracting a specific item from the data. As described later, the narrowing had to go one step further, but it still holds that a control set up in advance greatly cuts the time needed to narrow down the cause when a failure occurs.

So how were the grades of these 196 stocks being calculated? Jupick does not drop a stock from grading because a metric is blank. It puts a neutral fill-in value in place of the missing metric and calculates. That value is set so that the scoring curve turns it into exactly 50 points, the midpoint.

This approach is intentional. Giving zero points for a missing value would turn a company with thin data into an unsound one, and calculating from the remaining metrics without it would measure each stock by a different yardstick. The neutral value is a compromise chosen to avoid both traps.

The problem is the scale of the compromise. The financial stability area is made up of three metrics, debt ratio, current ratio and interest coverage, and for most Korean stocks one of them is a neutral value rather than a measured value. The grade still comes out, but the floor it stands on is thin.

One thing is worth making clear. Discovering this is not a reason to change the grading method right away. Fixing the collection could have moved the financial grades of what were then 196 stocks all at once. A change like that has to be made on its own, in a state where the reason for the movement can be explained.

What was fixed that night was the counter. The warning that checks whether a supplementary item has gone entirely blank was split by market. Now, if an item goes blank across all Korean stocks alone, the warning appears as a sentence saying how many Korean stocks are entirely blank. The reverse works the same way.

The same work split one more counter. The safeguard that catches rounds where the update half failed was also counting Korean and overseas stocks in one bucket. If every Korean stock died but the overseas ones were fine, the number updated would be exactly half of the total, just barely meeting the half threshold, and the round would end with no failure mark while stamping a fresh update time. This, too, is now checked per market.

These changes alter nothing on screen by themselves. The numbers and grades on stock pages are unchanged. The one thing that changed is that the next time the same failure happens, it shows up as a warning that day instead of passing quietly for months.

What is still unfixed is bigger. The rule that finds the interest expense item was left as it was that day. Fixing it would move the financial grades of what were then 196 Korean stocks all at once, so it was postponed as separate work. Writing down that it was postponed was the best that could be done at the time.

The conclusion that day was this: a blank interest coverage value for a Korean stock is information about Jupick's collection, not about that company. The October 8 remeasurement described below showed this conclusion was only half right. Even so, the part saying a company's condition cannot be read from the absence of a value still stands. Writing these two down so they are not confused is the purpose of this article.

A caveat was added to this paragraph. From September 4, 2026, a stock whose interest expense was read in a given round but which has no operating income is marked “Not applicable” next to the blank. A financial company's income statement has no operating income line, so the division does not hold. A “Not applicable” mark is therefore information about the company's sector. A “Not reported” mark was read that day as information about Jupick's collection, but as the next section describes, most of those also turned out to be information that the disclosure format has no such item.

Jupick returned to that postponed work on October 8, 2026. First it counted the current figures. Interest coverage is blank for 167 (83.5%) of the major Korean stocks. That is 29 fewer than the 196 of August 30, and the number of Korean stocks with a value rose from four to 33. This is the result of adding, on September 3, one more path that also reads the interest expense add-back adjustment in the cash flow statement.

The remaining 167 stocks were split again. Of these, 40 fall under a financial sector, and the other 127 are neither banks nor insurers. The shape seen in August remained, only smaller in scale.

So this time Jupick opened the data source directly. From the 127 still-blank stocks it chose twelve large non-financial companies, each in a different sector, fetched their full financial statements and went through the name and identifier of every item. None of the twelve had an interest expense item on the income statement. None of the twelve had the interest expense add-back item in the cash flow statement that was added in September either. Instead, all of them carried an item called finance costs, and their cash flow statements had an item for interest paid.

These two items are not the same quantity as interest expense. Finance costs is where interest is bundled together with losses from exchange rate differences and valuation losses on derivatives. Interest paid is the cash that actually went out during the period, so it excludes interest not yet paid and includes interest deferred last year and paid this year. Both are close to interest, but neither is incurred interest expense.

This is where it became clear that the August conclusion was only half right. The part about the data arriving was right. The part about Jupick's rules missing an item that was in the data was wrong. What the accounting standards require on the income statement is finance costs; they do not require a separate item for interest incurred alone. So many companies do not report that item at all. No rule can do a better job of finding an item that does not exist.

That leaves only one move: changing the bottom of the division to a different quantity. One could divide by finance costs, or by interest paid during the period. Even so, not all 167 stocks would fill in. For the 40 financial-sector stocks, the numerator, the operating income line, is often missing from the income statement, so the division still does not hold even with a different denominator, and the two stocks whose data source did not respond in this update stay as they are. At most, the new values would come from within the 127 non-financial stocks. And those values would be calculated differently from the values now held by 33 stocks, so two calculations would be mixed under the same name.

So again nothing was changed. Changing the denominator is not fixing collection but changing the meaning of the metric, and that requires the stock page to state what the value was divided by. Filling in numbers without explanation would move the financial grades of stocks on the order of a hundred overnight, with no record anywhere of why they moved. Writing down here that it was postponed, and why, is what was done this time.

A metric can go entirely blank in only one market. In the August 30, 2026 measurement, 196 of the 230 stocks missing interest coverage were Korean, and only four Korean stocks had a value.

Counting by total hides this kind of failure. Because two groups with different data sources were summarized in one number, even when one side was wiped out the other kept the count up and the alarm condition was never reached.

The sector explanation fit overseas stocks but not Korean ones. In the August measurement, 44 of the 196 Korean stocks were in financial sectors, and in the October 8 remeasurement, 40 of the 167 Korean stocks are. The remaining 127 include sectors that clearly have interest expense, such as semiconductors, autos and construction.

The contrast that the current ratio, calculated from the same data, was blank for only 30 Korean stocks split the cause once. The data was arriving. But reading the data source directly on October 8 showed that the item was not one Jupick was failing to read but one absent from the disclosures in the first place. What the accounting standards require on the income statement is finance costs, not a separate item for interest incurred.

So the one remaining option is changing the bottom of the division, and that is not fixing collection but changing the meaning of the metric. It has been postponed until there is first a place on screen to state what the value is divided by.

The habit of asking first what has been combined into a number is what finds holes like this. That is not a story about financial data alone.

This content is for information only and is not investment advice. Jupick gives no trading advice on any stock and does not forecast the future direction of prices, and the sectors used as examples serve only to explain how missing values are distributed, not as an assessment of any kind. The figures cited were measured on August 30 and October 8, 2026, and change with updates. Investment decisions and responsibility for them rest with the user.