Amanet: how much of the backlog is truly firm, and how much still depends on customer choice?
Amanet finished 2025 with roughly NIS 545 million of backlog, but the material-contract disclosure shows that this is not a backlog where the customer is committed to draw the full amount in advance. The backlog is real, yet a meaningful part of it still rests on actual usage, service hours, and fairly wide customer flexibility.
The main article argued that Amanet’s backlog expanded, but the cash-conversion test remained unresolved. This follow-up steps back one level and isolates the quality of the backlog itself: how much of it is truly firm in contractual terms, and how much rests on service frameworks where the customer still controls pace, volume, and sometimes even whether the work is consumed at all.
That matters because, at Amanet, backlog does not automatically equal almost-certain revenue. The most important sentence in the material-contract note is the small footnote beneath the table: percentage-of-completion is not relevant in these agreements because they are service contracts performed in practice, usually according to customer demand. That already tells the reader that the large headline number, NIS 545 million at the end of 2025, needs to be read differently from the backlog of a fixed project or a hard supply order.
The correct reading is not that the backlog is weak or fictitious. Quite the opposite. The company says there were no material cancellations in 2025, and backlog remained similar in mid-March 2026 in both the consulting segment and the computing segment. But the quality of Amanet’s backlog rests more on usage patterns, long-standing customers, and operating continuity, and less on hard contractual obligations that force the customer to take the full volume.
Four points matter from the start:
- Duration is not the same as firmness. In computing, software, and hardware, Amanet had NIS 394 million of backlog, including NIS 116 million for 2028 and beyond, yet the same section says the customer can usually cancel binding orders on one month’s notice and sometimes up to three months, without compensation.
- The largest contract is also the most flexible. In the hardware-testing contract, which carried roughly NIS 97 million of expected remaining revenue, the customer can increase, reduce, not order at all, and also use other contractors, with no minimum commitment.
- Even the contract that looks firmer still depends on actual consumption. The software-testing agreement includes a monthly retainer and roughly NIS 77 million of expected remaining revenue, but the retainer is based on estimated average hours, includes a true-up mechanism if usage deviates, and gives the customer a 90-day termination right.
- What really supports the number is not an exit penalty, but the client relationship. In computing, customers with more than five years of tenure generated NIS 201.3 million of revenue in 2025, and in consulting the equivalent figure was NIS 88.5 million. That is a sign of stability, but a very different kind of stability from hard contractual firmness.
What Counts As Backlog Here
At the end of 2025 Amanet reported group backlog of about NIS 545 million. The segment split already tells half the story: NIS 394 million in computing, software, and hardware versus NIS 151 million in consulting, project management, and logistics. The time structure is also different. Computing carries the longer-dated visibility layer, including NIS 116 million for 2028 and beyond. Consulting is concentrated almost entirely in 2026 and 2027.
But the second half of the story matters more. In both segments the company defines backlog as binding orders for projects and/or services that have not yet been recognized as revenue, and in both segments it adds that the customer can cancel with one month’s notice up to several months’ notice, without compensation. In computing the company explicitly says this is standard in the sector and may sometimes mean up to three months. So even the longer-duration part of the backlog is not a contract where the customer must consume the full volume in advance.
| Segment | Backlog at 31.12.2025 | Main timing split | What the company says about cancellation |
|---|---|---|---|
| Consulting, project management, and logistics | 151 | NIS 95m in 2026 and NIS 56m in 2027 | The customer may cancel on notice ranging from one month to several months, without compensation |
| Computing, software, and hardware | 394 | NIS 198m in 2026, NIS 80m in 2027, and NIS 116m from 2028 onward | The customer may usually cancel on one month’s notice and sometimes up to three months, without compensation |
The easy mistake here is to assume that longer backlog automatically means harder backlog. At Amanet that is not the case. Duration gives visibility. It does not remove the customer’s room to adjust.
The Four Contracts That Reveal Backlog Quality
The material-contract note is where the backlog headline starts to break into layers. The four disclosed contracts together include about NIS 206.9 million of expected remaining revenue, almost 38% of group backlog at year-end. That is already large enough to show how Amanet’s backlog really behaves.
| Contract | Expected remaining revenue | How revenue is actually generated | Where flexibility still sits with the customer |
|---|---|---|---|
| Dialog, treatment coordination | NIS 19.288m | Actual hours worked at an hourly rate, through June 30, 2026 or until a tender winner is chosen | The Ministry of Health may terminate on 45 days' notice, and the engagement itself only lasts until a new tender winner or the period end |
| Smart ticketing | NIS 13.6m | Supply of equipment and maintenance through August 31, 2026 | The customer may terminate on notice, and the company itself says it cannot estimate the revenue scope if the contract is extended again |
| Software testing | NIS 77m | A global monthly retainer based on average expected hours, with a true-up if usage deviates | The customer may terminate at any time on 90 days' notice, there are additional termination cases, and the retainer may be revisited at the end of each calendar year |
| Hardware testing | NIS 97m | Service calls billed according to actual hours and tender-set rates | The customer may increase, reduce, not order at all, shift work to others, and is not committed to any minimum volume |
This is the core point. There is not a single contract in the table that looks like a hard customer commitment to take the full stated amount. Even the expected remaining revenue figures are estimates of future revenue based on hours, retainers, maintenance, or service calls, not amounts the customer is obliged to consume.
That is especially visible in the two technology contracts. Together they account for about NIS 174 million of expected remaining revenue, roughly 84% of the amount disclosed in the material-contract note. Yet that is also where customer flexibility is widest. In software testing the client holds a straightforward 90-day termination right, while in hardware testing there is not even a minimum order floor.
Where Firmness Ends, And Where Stability Still Exists
After reading the notes, it helps to separate three different layers that can blur together on a first pass.
The first layer is visibility. Here Amanet looks reasonably solid. The timing is spread out, the client base is long-standing, and the company reports no material cancellations in practice during 2025. In March 2026 it still saw backlog broadly similar to year-end.
The second layer is contractual firmness. Here the picture is weaker. In both consulting and computing the company does not present broad minimum-purchase mechanisms, meaningful exit penalties, or structures where the customer must use the entire budget. In the material contracts, revenue is earned only if the service is actually provided.
The third layer is customer stickiness. That is a different kind of strength. In computing the company says customers with more than five years of tenure generated 60% of group revenue, and in consulting the equivalent figure was 26% of group revenue. That is not legal firmness, but it is still evidence that the business is built on recurring relationships rather than on one-off orders.
That is also why the bottom line here is not flatly bearish. Amanet’s backlog is not weak, because customers have not historically cancelled in a material way and because the client base is seasoned. But the quality of the backlog is measured here less by whether a contract was signed and more by whether the customer keeps ordering, renewing, and consuming.
The Dialog agreement shows this well. On one hand, it is an existing and ongoing service with a roughly NIS 19 million cap for the first half of 2026. On the other hand, it runs only until a new tender winner is selected or until the end of June 2026, and the ministry may terminate on 45 days' notice. It is a short bridge with visibility, not a hard-protected contract.
The software-testing contract sits in the middle. There is a monthly retainer, there is term through August 2029, and there is even an indicated annual value of roughly NIS 21 million at the current level. But the economic base is still average service hours, with a true-up if usage deviates and with an early termination option. So it is better read as recurring activity with adjustment rights, not as sealed backlog.
The hardware-testing contract is where the big number needs the greatest caution. It sounds like a step change, and it is important. But because the customer may order through service calls, change scope, and even not order at all, the number says more about framework potential than about hard execution.
What This Means For 2026
That is why Amanet’s 2026 test is not only whether backlog stays high, but whether these flexible frameworks translate into real revenue at a pace that justifies the headline. On a timing basis, the company attributes NIS 293 million of group backlog to 2026. That is a respectable figure, but it does not change the fact that a meaningful part of it still depends on how much service the customer actually chooses to consume.
Three signs matter most. The first is the pace of service calls in hardware testing, because there is no minimum floor there. The second is the stability of the software-testing retainer through the year, especially given the annual retainer-review mechanism. The third is continuity in the shorter public-sector contracts, where tenders, extensions, and supplier replacement can all matter.
If those three signals stay strong, one can argue that the contracts may be legally flexible but operationally sticky, which is a different kind of quality. If one of them cracks, the NIS 545 million headline will start to look less like a revenue floor and more like a potential ceiling.
Bottom Line
Amanet’s backlog deserves to be counted, but it should not be treated as if all of it were equally firm. The contract note shows a clear picture: most of the material commitments are service-based, usage-dependent, and leave the customer with meaningful room to adjust. The strongest part of the backlog is therefore not a contractual penalty mechanism, but a long-standing client base and the ability to provide recurring services.
That is exactly why the debate around Amanet is not whether it has backlog. It does. The debate is what kind of backlog it is. As of the end of 2025, the answer is backlog with good visibility but only moderate contractual firmness. That is why, even in this continuation, the real test remains the same one: not the size of the framework, but the rate at which it converts into revenue, profit, and eventually cash.
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.