Naoi 2025: Stage 3 Rose, Are the Collateral Buffers Really Enough?
Naoi's Stage 3 balance jumped to NIS 145.3 million, yet the total allowance ratio stayed at just 1.00%. At the book level there is a real collateral cushion, but the report does not map it directly to the impaired balances, so the question has moved from the headline to the structure of the model and the recovery assumptions.
The Gap That Stayed Open
The main article already pointed to the core contradiction in Naoi's report: the total allowance ratio looked almost flat, while the troubled part of the book no longer looked like background noise. This follow-up isolates the question that remained unresolved after the first read: do the collateral layers really absorb the jump in Stage 3, or is the report simply delaying part of the pain through an ECL model that still gives heavy weight to collateral and to the macro backdrop.
The raw numbers are sharp. Gross customer credit before allowance grew 33.2% to NIS 5.253 billion, and the allowance balance rose almost in line to NIS 52.48 million. That leaves a calm headline: 1.00% versus 0.99% a year earlier. But inside that same book, Stage 3 rose to NIS 145.33 million from NIS 32.35 million, and its share of the book rose to 2.77% from 0.82%.
The decisive point is that Stage 3 coverage collapsed to 12.89% from 38.71%. In other words, Naoi now recognizes a much larger stock of impaired credit, yet holds a much thinner accounting allowance against each impaired shekel. To judge whether that is justified, two separate questions have to be unpacked: what the collateral stack actually proves, and how the ECL model keeps the overall allowance headline from moving much at all.
What Actually Changed Inside Note 4
The misleading number on first glance is the total allowance ratio. It barely moved, so it is easy to think book quality barely moved either. But Note 4 tells a different story:
| Metric | 2024 | 2025 | What it says |
|---|---|---|---|
| Gross customer credit before allowance | NIS 3.944 billion | NIS 5.253 billion | The book grew 33.2% |
| Stage 2 gross | NIS 73.1 million | NIS 31.7 million | The watchlist bucket shrank 56.7% |
| Stage 3 gross | NIS 32.4 million | NIS 145.3 million | Impaired balances jumped 4.5x |
| Total allowance | NIS 39.2 million | NIS 52.5 million | Up broadly with book growth |
| Total allowance ratio | 0.99% | 1.00% | The headline still looks stable |
| Stage 3 allowance | NIS 12.5 million | NIS 18.7 million | Up only 49.6% |
| Stage 3 coverage ratio | 38.71% | 12.89% | Much less accounting cover per impaired shekel |
What matters most is not only the jump in Stage 3, but the opposite move in Stage 2. Stage 2 fell to NIS 31.65 million from NIS 73.14 million, and its allowance fell to NIS 412 thousand from NIS 853 thousand. This is not a picture of mild stress spreading evenly across the book. It is a picture of pressure moving into a smaller and harder bucket while the middle layer actually contracts.
The ageing buckets sharpen the point further. In 2025 Stage 3 already includes about NIS 142.2 million that is more than 90 days past due, versus only NIS 32.35 million a year earlier. Stage 2, by contrast, fell to about NIS 26.9 million above 90 days, from about NIS 54.9 million in 2024. The note explicitly says some balances can stay in Stage 2 even when they are more than 90 days overdue if management still believes collection is likely, based in part on repayment arrangements, third-party guarantees, sufficient collateral, legal enforceability, and borrower cooperation. That means classification is not mechanical. It already embeds judgment about recoverability.
What the Collateral Really Covers
At the total-book level, Naoi does have a meaningful collateral layer. NIS 3.853 billion of exposures are secured, against NIS 6.366 billion of gross collateral and roughly NIS 4.733 billion after haircuts. Put simply, on the secured slice that is about 1.65x gross coverage and 1.23x after haircuts. If the question is whether there is any real collateral cushion in the book, the answer is yes.
But that is no longer the relevant question for 2025. The relevant question is how much of the NIS 145.33 million Stage 3 balance actually sits behind the strongest part of that collateral stack. Here the disclosure stops. The report breaks down collateral at book level, but it does not map Stage 3 by collateral type, lien rank, LTV, or absorption-capacity band. So the NIS 4.733 billion figure after haircuts cannot be used as a clean pass mark for the impaired balances.
There is another non-obvious point here. The largest layer after haircuts is not first-ranking mortgages, but residential project surplus rights at NIS 2.08 billion, versus NIS 1.556 billion in first-lien mortgages. That is still real protection, but it is not the same thing as a first-ranking mortgage on an existing asset. By the company's own definition, residential surplus rights are tied to the required equity in a project plus expected profits, and financing against them depends in part on execution progress and sales absorption. In other words, the biggest collateral bucket is itself tied to project economics.
| Collateral type | Secured loans | Gross collateral | Collateral after haircuts | Coverage after haircuts |
|---|---|---|---|---|
| First-lien mortgage | 1,342 | 1,830 | 1,556 | 1.16 |
| Second-lien mortgage | 540 | 816 | 540 | 1.00 |
| Residential surplus rights | 1,414 | 2,746 | 2,080 | 1.47 |
| Tradable securities | 295 | 597 | 295 | 1.00 |
| Vehicles | 23 | 35 | 23 | 1.00 |
| Third-party guarantees | 239 | 342 | 239 | 1.00 |
| Total | 3,853 | 6,366 | 4,733 | 1.23 |
The implication is not that the collateral is weak. The implication is that the protection is not uniform. First-lien mortgages, second-lien mortgages, project surplus rights, marketable securities, vehicles, and third-party guarantees do not have the same recovery quality or the same realization timeline. So the aggregate collateral stack gives comfort at book level, but it does not prove which part of the impaired balances sits behind the strongest recovery layer.
Why the ECL Model Smooths the Picture
This is where the issue becomes structural. Naoi's general allowance is not only a function of what has already happened in delinquency. It is built through an ECL model that, according to the report, was externally validated in the fourth quarter of 2025, yet did not materially change in methodology. The model rests on three main anchors: the company's own loss history, a peer group of comparable instruments, and the macro outlook.
| Model driver | What the report says | Why it matters for the 2025 read |
|---|---|---|
| Internal loss history | The validated model is mainly based on the company's own default history | Several years of low specific provisions push the general allowance lower |
| Peer group | The peer group reflects sector provisions at the large Israeli banks | Higher bank provisioning actually pushed the general allowance up in 2025 |
| Macro outlook | The model uses interest rates, inflation, unemployment, GDP growth, and the pace of new-apartment sales | Apartment-sales pace gets heavier weight because of the book's real-estate concentration |
| Collateral structure | Instruments are grouped by collateral type and common sector-risk characteristics | A more collateral-heavy book can soften the model result |
One additional disclosure matters a great deal. The company says that in 2024 it added the value of collateral, after a quick-sale haircut, into the allowance calculation for exposures with a significant increase in credit risk. It also explains that because about 83% of those balances were secured, that change reduced allowance on those balances. In 2025 the company continues with a validated model that did not materially change in methodology and still groups instruments by collateral type. The reasonable inference from that disclosure is that collateral does not only protect the economics of the book. It also slows the rise in the general accounting allowance.
That is exactly why the risk has moved from the headline into the structure. The sharpest deterioration sits in Stage 3, but the figure that stabilizes the overall picture is still the general allowance ratio, which depends on low past defaults, collateral grouping, peer-group inputs, and macro assumptions in which apartment-sales pace receives extra weight. If housing-sales conditions weaken, the pressure does not have to show up first through fresh defaults. It can show up first through the model.
This is not just a technical note, either. The external auditor flagged the allowance for credit losses as a key audit matter precisely because classification and measurement rely on significant estimates, uncertainty, and challenging judgment. That does not prove the allowance is insufficient. It does mean the debate is no longer about one simple number, but about the quality of the assumptions that keep that number in place.
What Has to Be Seen Next
If the goal is to believe that the collateral really is enough, it is not enough to point to 73% secured exposure and NIS 4.733 billion of post-haircut collateral. The next developments have to support that read at the impaired-balance level as well:
- Stage 3 has to stabilize or decline, not only the total allowance ratio.
- The sharp compression in Stage 3 coverage cannot keep extending into 2026 without stronger evidence of recovery.
- Better disclosure would be needed on Stage 3 by collateral type, lien rank, or LTV range. Without it, the reader still has a book-level picture, not a focused risk picture.
- If apartment-sales pace weakens or bank provisions in comparable sectors continue rising, the general allowance should react. If it does not, the structural doubt only grows.
Bottom Line
At the total-book level, Naoi does have a real collateral cushion. But at the level of the question that became material in 2025, that is no longer enough. Stage 3 has become too large, its accounting cover has become too thin, and the report does not provide the mapping needed to connect the impaired balances directly to the strongest part of the collateral stack.
So the more accurate thesis is not that the collateral does not exist. It is that the collateral still does not resolve the risk question. Right now it explains why the total allowance ratio still looks calm. It does not yet prove that actual recoveries will be strong enough to justify the collapse in Stage 3 coverage. That is where the story has moved from the top-line comfort of the report into the structure of the book and of the allowance model.
Disclosure: Deep TASE analyses are general informational, research, and commentary content only. They do not constitute investment advice, investment marketing, a recommendation, or an offer to buy, sell, or hold any security, and are not tailored to any reader's personal circumstances.
The author, site owner, or related parties may hold, buy, sell, or otherwise trade securities or financial instruments related to the companies discussed, before or after publication, without prior notice and without any obligation to update the analysis. Publication of an analysis should not be read as a statement that any position does or does not exist.
The analysis may contain errors, omissions, or information that changes after publication. Readers should review official filings and primary sources before making decisions.