Before uploading receipts to the cloud, first distinguish between "failure" and "not confirmed".

Today is Tuesday, November 17, 2026. First, a piece of news from last week, then an earlier one that is more worth reading.

On November 9, Odoo Italy announced that it is integrating fiskaly's electronic receipt service into its own POS at no extra charge. For a shop in Milan, this sounds like moving a machine away: there is no need to buy, maintain, and periodically submit for inspection that tax cash register called registratore telematico; sales data is sent directly to the tax authority (Agenzia delle Entrate) via the cloud.

One week earlier, on November 2, fiskaly reported an incident: while the tax office portal was unavailable, SIGN IT receipt submissions experienced timeouts and errors; and some of the requests that “appeared to fail” may actually have already been delivered. Its instruction to users was not to “retry,” but to “first check whether this document has been registered, then decide whether to resubmit,” the reason being to avoid duplicate declarations.

One product release, one incident notice. Taken together, they precisely show what the most expensive part of the code in ERP looks like: it is not on the feature list, but in "how do I determine whether this thing actually succeeded."

I. This message changes the direction of a sheet of paper

Italian retail reporting consists of two records. One is called documento commerciale, which records each individual sale: the merchant's VAT number, what was sold, the applicable tax rate, and how the customer paid. The other is called corrispettivi, which is a summary of sales amounts that merchants must record electronically and report. The former is transaction by transaction, the latter is a summary, and the two must reconcile at the daily close.

In the past, there were two approaches. One was hardware: tax cash registers stored transaction data locally and generated a file at the daily close to report to the tax office. The other was the tax office's own online program, where a merchant entered each document on a webpage as it was issued. According to fiskaly, its version connected to the tax office's existing portal (SIGN IT lite) is still running; another version is being piloted under the framework of the tax office's March 2025 technical ruling, and the November 6 update places the planned commercial availability in early 2027, subject to the approval outcome.

Odoo's approach takes the form of the third path: cashiers record a sale in Odoo POS as usual, and fiskaly is responsible for reporting what needs to be reported. This is not the first collaboration between Odoo and fiskaly—Germany uses SIGN DE, Austria uses SIGN AT, and Italy is the third market.

MarketServices usedWhere do the differences lieWhat does it mean for enterprises
GermanySIGN DEDifferent technical requirements for signatures and tamper resistanceIn the same ERP, the generation sequence of accounting vouchers needs to be changed.
AustriaSIGN ATThe declaration pipeline and voucher format are differentStores in both Austria and Germany cannot share a single checkout process.
ItalySIGN ITTwo sets of records: individual vouchers and amount summaries; daily closing declarationThe rules for determining voided vouchers and duplicate declarations are different from those of the previous two countries.

Three markets, three sets of requirements, three ways of doing things. What this reminds us is: localization is not translating the menu into Italian, but bringing the local timing into the system—when vouchers must be issued, when consolidation must happen, and whose records prevail when something goes wrong. If the translation is wrong, you change the copy; if the timing is wrong, you change the books.

II. The number 95% determines the teardown timetable

On the same timeline, there is also a harder piece of evidence. On November 4, Lucia Albano, Deputy Minister of Economy and Finance of Italy, provided a distribution when responding to the parliamentary finance committee: about 95% of merchants use a tax cash register (telematic register), while the rest use the tax authority's online program.

95%

According to the parliamentary reply of the Italian Ministry of Economy and Finance dated 2026-11-04, the share of merchants still using tax cash registers (the remaining approximately 5% use the tax authority's online procedure)

In the same response there is another sentence that is easier to overlook: the tax bureau has never authorized third-party tools to automate its web program through an independent interface, and is testing methods to detect and block them. This sentence targets a type of practice—using RPA or scripts to "click" through the web process once, making it look automated while actually bypassing the approval framework.

"The cloud can replace hardware" does not mean "you can dismantle the machines this month".

The three dates must be plotted on the same Gantt chart: the commercial availability date of the certified version (according to fiskaly, early 2027, and pending approval), the decommissioning schedule for existing hardware (contracts, leases, annual inspection cycles), and the migration of registered payment terminals. If the three dates are not aligned and the equipment is dismantled, then during the intervening gap, every receipt issued in the store can only be handled manually.

This is not to talk down new technology. 95% is a fact of the installed base: the operating habits, audit paths, and employee training of the vast majority of peers are still tied to hardware. The pace of compliance switching is not determined by suppliers' release dates, but jointly determined by three things: approvals, the installed base, and payment terminal registration.

What is more noteworthy is what Odoo did not write in this announcement: which Odoo versions are supported, which hosting methods are available, what steps existing stores need to take to switch from tax cash registers; and how the registration relationships of existing payment terminals will be handled after the switch. These are not oversights, but a typical case of 'the first version first says what it can do.' For implementers, these blank items are exactly the places to fill in on quotations and acceptance forms.

3. “Failure” and “Not Confirmed” Are Two Different States

Let's return to the outage on November 2. What it exposed was not system fragility, but a distinction that is extremely common in ERP and extremely easy to get wrong: a request returning a failure and a request whose result is unknown are not the same thing at all.

Deterministic failureUnconfirmed status
The other party explicitly rejects it, or did not receive it at all. It is safe to retry. Retrying does not change the business outcome, only the timing.The connection dropped or timed out, and the other party may have already processed it. Retrying at this point is equivalent to reporting it again. You must query first, then decide.

In a retail scenario, the consequence of this error is a duplicate tax document; in ERP, it is duplicate posting, duplicate invoicing, and duplicate payment. The nature is exactly the same: there is one extra entry on the books while there is only one transaction in the business, and during reconciliation you cannot tell which entry is the extra one—because the two look exactly identical. This is silent duplication, which is far harder to handle than an error: errors get attention, but duplicates get none.

fiskaly's fallback path for portal-based services also confirms this: when retries still fail, first issue a temporary paper receipt, then supplement the electronic invoice afterward. Note this order—it acknowledges that the "status undetermined" situation must have a deliverable temporary item, rather than leaving the cashier standing there repeatedly clicking submit.

Silent duplication is the most expensive class of bug: it raises no alerts, blocks nothing, and never appears in exception reports.

Two concrete lines of defense: First, give each declaration an idempotency key (document number + terminal/device number + business date, or the unique identifier returned by the tax authority), so that resubmission does not create a second record; second, retries must have an upper limit and backoff, and when the number of attempts is exhausted, they must enter a manual queue—the script must not be allowed to quietly loop in the background until dawn.

There is one more detail worth copying into the design: during an outage, what should the cashier's screen display? Odoo's release materials do not explain what the cashier sees after a submission fails, whether the POS will retry automatically, or how staff should verify before closing that everything that needed to be reported was reported. These three questions are not product defects; rather, they are questions that every project that connects a local action to an external authoritative system must answer for itself—because the person who answers them is the person who has to close the store that night.

IV. Acceptance Checklist for CIOs: State Machine Before Features

If you want to implement this type of tax interface on your own ERP (whether using native Odoo, third-party modules, or self-built middleware), it is recommended to reverse the acceptance order: verify status first, then verify functionality.

Acceptance itemSpecific phrasingTypical consequences of failure
State machineHow many statuses are there on a document? Who can change them? Does a failure directly trigger a retry?"API failure" is treated as "business failure"; each retry adds one more record
Idempotency keyIf the same transaction is resubmitted three times, how many records are there in the system?Duplicate declarations and duplicate invoicing, with tax amounts and book amounts unreconciled over the long term
Daily closing reconciliationAre the four numbers—local transaction count, confirmed transaction count, rejected transaction count, and discrepancy count—equal?Underreport one section and no one will know; the fine arrives a few months late.
Manual fallbackWhen external systems are unavailable, do store staff have any temporary credentials on hand that can be delivered?Store shutdown, or issuing invoices first and backfilling entries later, resulting in duplicates
Switch windowAre the certification commercial launch date, hardware retirement date, and payment terminal migration date on the same chart?During the gap period, we relied on manual operations, and the weeks with the highest error rates happened to fall right when we were busiest.

Four statuses are worth writing directly into the document model: pending submission, submitted awaiting confirmation, confirmed, and rejected (terminal state). Automatic transitions between the third and fourth states are not allowed; there must be a query action; rejected is a terminal state and must not be overwritten by retries. The mapping in Odoo is roughly like this: after a POS order is persisted, accounting entries are generated; tax reporting goes through an asynchronous queue as an additional status of the entry; if the queue fails, it goes to an activity or to-do instead of being thrown back to the cashier; each request and response keeps the original text so that later it can answer "whether it was actually reported that day".

Three actions for the CIO:

  • Set up “Unconfirmed” as a separate status, and do not allow it to fall under “Failed.” This is the starting point for all governance. With one status too few, all subsequent reconciliation is guesswork.
  • Make daily closing a four-number reconciliation, not a comparison of “whether there are errors.”Local transaction count, confirmed, rejected, and discrepancies—only when these four numbers are equal is closing allowed; if they are not equal, issue a discrepancy list and assign a designated person to handle it that day.
  • During the transition period, schedule according to “three-party alignment,” not according to supplier release dates.Certification version commercialization, hardware retirement, and payment terminal registration migration run on three parallel tracks; if any one is not in place, run in parallel first—don’t hop on one leg.

The value of a compliance interface does not lie in “whether it can report automatically,” but in whether these three questions—“Was it reported? How many times was it reported? Who confirmed it?”—can be answered at any time. Automation that cannot answer them merely swaps paper risk for electronic risk.

Odoo has filled in the Italy stop, showing that the localization of open-source ERP has become granular enough to reach the level of “a country’s tax sequence,” which is good news for companies operating across borders. But the same thing has also raised the implementation threshold: in the past, one only had to match tax accounts; now one has to catch the failure semantics of external authoritative systems. If this step is done solidly, all subsequent e-invoicing, customs, and banking interfaces will be reuse of the same set of skills.

Source: ERP News, “Odoo Adds Cloud Fiscal Receipt Reporting to Its POS in Italy” (reporting on Odoo’s Italy announcement of 2026-11-09, including fiskaly SIGN IT, no additional charges, and the three markets of Germany SIGN DE and Austria SIGN AT); fiskaly’s 2026-11-02 incident statement and 2026-11-06 version update (commercial availability timeline); Italian Deputy Minister of Economy and Finance Lucia Albano’s 2026-11-04 reply to the parliamentary finance committee on retail reporting (about 95% of merchants use fiscal cash registers; unauthorized third-party automated web programs and testing of detection and blocking); the software approval framework established by the Italian Revenue Agency’s March 2025 technical ruling.

Definitional note: Odoo and fiskaly are both vendor announcements and do not disclose the number of customers, number of stores launched, or failure rate; early 2027 is fiskaly's planned commercial availability time, explicitly "subject to approval," and does not constitute a commitment. The 95% and 5% are oral percentages given by government officials in parliamentary replies; the original text does not provide the statistical methodology or sample. The state machine, idempotency keys, and four-number reconciliation covered in this article are summaries of implementation experience, not official specifications of Odoo or fiskaly.

关于我们

​我们致力于帮助中小企业实现数字化转型,我们的团队由一群充满激情和创新思维的专业人士组成,他们具备丰富的行业经验和技术专长。

扫一扫获取顾问以及手册

Sign in to leave a comment
Your company — AI has to get to know it all over again every time