Alarum: NRR Broke Just as AI Accelerated the Top Line
Alarum finished 2025 with 28% growth, but in that same year NRR fell to 0.83 and two customers rose to 27% of revenue. AI pushed the top line higher, but it did so on a narrower and less proven revenue base.
What This Follow-Up Is Isolating
The main article argued that Alarum’s new demand wave, especially the AI-linked demand, grew revenue faster than the company could protect margin or turn that growth into cash. This follow-up is not revisiting cash conversion or the broader cost debate. It is isolating one tighter issue that explains why 2025 revenue quality looks weaker than the headline: NRR broke below 1 at exactly the same time that large-customer concentration stepped up.
That matters because the company itself attributes the $8.9 million revenue increase to strong demand from a large-scale customer building foundational AI models and to higher sales of new products. At the same time, net revenue retention, or NRR, fell from 1.27 at the end of 2024 to 0.83 at the end of 2025. This is not a technical footnote. It is the metric that is supposed to answer whether the customer base that already existed a year earlier is expanding, shrinking or staying stable without relying on new logos.
Concentration also moved sharply in the same year. In 2024, no single customer exceeded 10% of revenue. In 2025, two main customers reached 15% and 12% of revenue, together $10.835 million. In addition, six customers purchasing more than $1 million each generated about 49% of data collection business revenue. That is the core of this follow-up. Alarum’s growth did not only become more dependent on new large customers, it also began to rest on a steeper revenue pyramid.
| Revenue-quality signal | 2024 | 2025 | Why it matters |
|---|---|---|---|
| Year-end NRR | 1.27 | 0.83 | The older cohort moved from expansion to contraction |
| Single customer above 10% of revenue | None | Two customers, 15% and 12% | Dependence on a few large accounts jumped |
| Data collection solutions and datasets revenue | $0.6 million | $11.1 million | The newer engines took center stage |
| IPPN revenue | $30.3 million | $29.2 million | The legacy engine no longer carried growth on its own |
NRR Did Not Just Fall, It Changed the Meaning of Growth
The key detail in Alarum’s NRR definition is not the number itself, but what sits inside it and what stays outside. The company defines the metric based on annual recurring revenue, or ARR, from the same customer cohort that existed 12 months earlier. Expansion within that cohort is included, contraction and attrition are netted out, but ARR from new customers is excluded. In other words, all of the new AI-related wins are not allowed to improve NRR. They sit outside the metric.
That is exactly why the move from 1.27 to 0.83 is more serious than it may look on a quick read. Management explicitly says the company is experiencing a notable shift in customer segments, with strong growth from strategic customers in the AI vertical offset by declines in others. The numbers already imply the interpretation: the new growth came from above, through new or rapidly growing large accounts, while the older base weakened at the same time.
There is another layer here that is easy to miss. The disclosed NRR is not a single-quarter reading. It is the average of the trailing four quarterly point-in-time NRRs. That means the move to 0.83 by year-end already comes after statistical smoothing, not because of one noisy quarter. When the metric moves through 2025 from 1.13 to 0.98, then to 0.92 and finally to 0.83, it is difficult to dismiss the deterioration as timing noise. This is already a base-structure shift.
The economic meaning is simpler than the jargon. Alarum did grow revenue in 2025, but not because the customers that had already been with the company for a year were spending more. The new large customers offset erosion elsewhere. That makes 2025 a proof year for winning large new accounts, not a proof year for broad, healthy deepening across the installed base.
That also explains why the revenue-quality question comes before the cash question in this follow-up. If the older base is weakening, the company has to keep running forward just to avoid looking flat. In that setup, any slowdown in new-logo additions or any pause in spending by a large account becomes a top-line risk much faster, not just a sentiment risk.
Concentration Did Not Stay a Hint, It Became Dependence
The second warning sign is concentration. In the 2024 filing, no single customer crossed 10% of revenue. In 2025, two did, together accounting for 27% of annual revenue. That is too large a move to treat as ordinary noise. It means the newer growth engine did not only accelerate revenue. It also made the revenue curve much steeper.
The broader disclosure supports the same reading. Six customers purchasing more than $1 million each generated about 49% of data collection business revenue. By contrast, 53 customers purchasing between $100 thousand and $1 million generated about 34% of NetNut revenue, and another 176 customers in the $10 thousand to $100 thousand band generated about 15% of revenue. This does not mean Alarum lacks a middle tier or a tail of smaller customers. It has both. But the economic center of gravity has moved toward a much narrower top layer of large accounts.
| 2025 customer layer | Disclosed exposure | Economic reading |
|---|---|---|
| Two main customers | 15% and 12% of revenue, together $10.835 million | More than a quarter of the top line rests on two accounts |
| Six customers above $1 million | About 49% of data collection revenue | The high-growth layer became very narrow |
| 53 customers in the $100 thousand to $1 million range | About 34% of NetNut revenue | There is a middle layer, but it no longer carries most of the weight |
| 176 customers in the $10 thousand to $100 thousand range | About 15% of revenue | The long tail exists, but its economic weight is still limited |
What matters most is the combination of the two earlier signals. High concentration alone is not automatically a problem if it comes with strong NRR, because then the argument can be that the product is deepening within the installed base while still keeping broad demand healthy. At Alarum, the two metrics moved in the opposite direction. Concentration rose while the metric designed to test the health of the older base fell. That makes the picture less about higher-quality scale and more about growth becoming dependent on a few deep pockets.
It is also important not to overstate what the filing discloses. The report does not explicitly say that the two largest customers are the AI customers, and it does not name them. So the identity cannot be assumed. But the more important point is still clear: in the same year that management described strong growth from strategic AI customers, concentration moved from a setup with no customer above 10% to one where two customers reached 27%. That does not prove identity. It does sharpen the quality-of-revenue risk.
The Geography Table Also Signals Rotation, Not Broad-Based Expansion
The geography table does not tell us who the customers are. It tells us where they are located. Even so, it points to a sharp mix shift. China rose by $6.646 million versus 2024, the U.K. Virgin Islands rose by $3.368 million, and the U.S. rose by $2.797 million, while the United Arab Emirates fell by $5.699 million and the Middle East and Africa fell by $1.414 million.
The right reading here is not that the company found one winning geography. That would go too far. The right reading is that 2025 growth was not distributed smoothly across a broad, stable customer base. It came through a sharp swap between demand pockets. When that happens alongside weaker NRR and higher concentration, it becomes harder to argue that the business simply got wider. It looks more accurate to say that it became more dependent on a few strong pockets of demand that appeared quickly.
Just three lines in that table, China, Hong Kong and the U.K. Virgin Islands, add up to roughly $20.3 million in 2025 revenue. Again, that is not a substitute for customer-level disclosure, but it does reinforce the feeling that the new revenue structure is becoming more concentrated along the location axis as well. The more growth looks like a sharp rotation across demand pockets, the harder it is to treat it as a newly stabilized base.
Why This Matters More Than the AI Headline
The temptation in reading 2025 is to stop at a half-positive conclusion: Alarum finally found a new demand wave, especially around AI-linked customers and higher-layer data collection products. That is true, but it is only half the picture. The other half is that this growth wave did not repair the economics of the older base. It masked them.
That gap changes how 2026 should be read. If Alarum had shown both stable or improving NRR and only a modest rise in concentration, the argument could have been that the company opened a new engine without losing the older structure. But when NRR falls to 0.83 and two customers reach 27% of revenue, the message is different: growth exists, but it is still not proving that the company is building a broader recurring-demand base.
That is also what the market can miss if it looks only at the top line. Another strong revenue quarter would not necessarily improve the read if it arrives again without an NRR recovery and without softer concentration. In that case, each additional dollar of revenue would mostly say that the company is still getting more work from large accounts, not that it is building a more stable platform.
| What needs to be watched next | Why it is critical |
|---|---|
| NRR needs to stop falling, and ideally move back toward 1 | That is the direct test of whether the older cohort is stabilizing |
| The weight of the two largest customers needs to fall, or at least stop rising | Without that, top-line volatility will remain elevated |
| The newer engines need to grow without further erosion in IPPN | Only then can they be treated as additive rather than merely replacing the old base |
| Revenue by customer location should look less jumpy | Too much rotation across pockets of demand keeps the stability question open |
Conclusion
This follow-up does not contradict the main article. It sharpens it. Alarum did show in 2025 that it can attract meaningful new demand, including demand linked to AI and to higher-layer data collection products. But at the same time it also showed that this demand arrived while two assets that should have improved alongside growth actually weakened: retention inside the older base and the diversification of revenue sources.
Current thesis in one line: Alarum accelerated the top line through large customers and newer products, but it did so while the older revenue base contracted and concentration rose.
The fair counter-thesis is that 2025 was a cohort-transition year. Under that reading, the AI customers are still early in their lifecycle, so NRR looks temporarily weak and concentration looks temporarily high. If those accounts deepen in 2026 and the company manages to build a broader customer layer around them, 2025 may look in hindsight like a reset year rather than an erosion year. That is a real possibility. But as of year-end 2025, the evidence still supports the harsher interpretation more than the softer one: growth came back, but broad revenue quality did not.
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.