On September 14, 2026, Gordon Company, a U.S. Christmas decoration retailer with nearly 50 years of history, filed for bankruptcy protection. The bankruptcy filing took aim at an inventory and order management system: starting in 2017, it partnered with SAP reseller Vision33, paid more than $2 million for the system, and alleged that it “never functioned as intended.” The system could not keep up with peak-season order volumes, forcing the company to suspend sales through some channels, and its main retail partner Target even imposed a one-week shipping delay on its merchandise. The company abandoned the system in 2021, but the operational damage had already been done, and it ultimately collapsed before the 2026 peak season. The claimed amount is $15 million.
If you put it together with the two supply-side messages from the first week of November, you will find that they are saying the same thing—just standing on opposite banks of the river.
I. Two Messages in November, and a Countdown
The first item comes from the German-speaking region. According to the monthly roundup in much. Consulting's Odoo News November 2026, Benedikt Sauter, founder of German cloud ERP vendor Xentral, has reportedly been removed from day-to-day operations, with Domenico Cipolla taking sole control; Sauter confirmed the personnel change but denied that a new funding round had stalled, saying: “We still have enough money.” The latest disclosed figures are for 2024: revenue of €14.5 million (+22%), a net loss of €9.4 million, and cash falling from €39.6 million to €32.1 million; investors include Sequoia and Tiger Global, with cumulative funding of about $95 million, and the last round was in August 2021.
The second item comes from the other end of the mid-market. In early November, IBM and SAP jointly launched a cloud ERP innovation support program for mid-sized enterprises. The backdrop is SAP's announced maintenance deadline: standard maintenance for Business Suite 7, which supports the legacy ECC system, ends at the end of 2027; subsequent extended support requires additional payment and terminates in 2030. The same report also described IBM's own situation: its preliminary second-quarter financial results fell short of expectations, and its stock fell by about a quarter in a single day, the largest drop in its 115-year history; IBM Z mainframe revenue fell 42% year over year, and its full-year revenue growth forecast was revised down to 4% to 5%.
SAP Business Suite 7 standard maintenance cutoff year; extended support ends in 2030 (quoted from an IBM×SAP midmarket solution report)
On one side, a startup has gone five years without new funding; on the other, a century-old firm is using a single maintenance deadline to move mid-market customers. Further down, there is a repeatedly cited observation: the $150 billion ERP market is decomposing from “integrated suites” toward “modular point solutions,” and a group of cloud-native finance and accounting challengers (Rillet, Pennylane, Ramp, Light, DualEntry, Campfire, Everest, Digits) are focusing on point capabilities covered by a module in the large suites, while the orchestration layer above the modules is opening doors for the next batch of startups.
These three things point to the same buyer question: Can the company you signed with survive until the end of the contract, and will it still be there for you when something happens to you?
II. Contract Price Measures Cost; Accidents Measure the Price Paid
What is most worth remembering about the Gordon case is not the amounts, but the ratio between those two numbers: a $2 million system, a $15 million claim, a 50-year-old company. This is not a pricing error; it is that two different units of measurement are involved.
The contract amount measures the cost of the software and implementation—licensing, person-days, configuration, training; each can be quoted, compared, and written into the tender. The loss caused by system failure measures the cost to business operations—lost orders, suspended channels, retailer penalties, revenue that cannot be collected during peak season, and final liquidation costs. There is no proportional relationship between the two, while the liability cap, acceptance criteria, and milestone sign-off in the contract are all written precisely in the former unit of measurement.
2 million USD
Amount paid by Gordon Company for this system (since 2017, based on public reports)
15 million USD
The claim amount in the bankruptcy documents, including consequential losses such as lost orders, channel suspensions, retail fines, etc.
Framing note: Information on the Gordon case comes from public reports' paraphrase of bankruptcy filings; the existing materials do not specify which SAP product (Business One, S/4HANA, or another) was involved, and this article does not draw conclusions about the product; the claim amount is the plaintiff's assertion and does not equal a court judgment, and liability remains to be determined through proceedings. This case is cited to illustrate the structural asymmetry between the "contractual liability cap" and the "cost of business interruption," not to determine that a particular vendor or product has a problem.
This asymmetry will be amplified once AI comes in. When traditional ERP goes wrong, it usually means it "didn't run right"—documents are wrong, reports don't balance, and it can be investigated, corrected, and remedied. After agents take over the process, problems become "ran something that shouldn't have been run"—it submitted a purchase based on its own understanding, sent a commitment, and changed a master data record. The action has already happened; it's just that no one approved it. At this point, the parties to hold accountable instantly multiply: model vendors, agent developers, implementation partners, integrators, and your own business departments. Responsibility gets sliced very finely along the chain, while the counterparty to the contract is often only one.
A single system failure can bring down a company with annual revenue in the tens of millions of dollars, but the liability cap in the contract is usually calculated only based on the price of the software and implementation. This gap does not disappear; it only shifts—to creditors, employees, and shareholders.
So the missing row in the selection scorecard is counterparty risk: what this company will live on over the next three to five years, whether its revenue structure can withstand a renewal negotiation, and whether you can take your data and configurations with you when you exit.
3. The seller's cash flow is also the buyer's risk
In the past, few people filled in this column carefully, because ERP seemed to be a very stable industry. But the figures disclosed this year are not all stable.
| Main body | Key figures | Implications for buyers |
|---|---|---|
| Xentral (Germany, cloud ERP) | 2024 revenue €14.5 million (+22%), net loss €9.4 million, cash €39.6 million → €32.1 million; cumulative funding approximately $95 million, last round 2021-08; founder removed from day-to-day operations | With no new funding coming in for five years, the credibility of the product roadmap should be discounted. |
| IBM (mid-market ERP solution provider side) | Q2 preliminary financial results below expectations, share price fell about 1/4 in a single day; IBM Z revenue -42% year over year; full-year revenue growth forecast revised down to 4%–5% | Big companies are also under pressure, but their ability to withstand risk is not on the same level as startups |
| Yonyou (First Half of 2026, Summary of Third-Party Research Reports) | Revenue of 3.727 billion yuan (+4.08%), net profit attributable to the parent of -928 million yuan, operating cash flow of -1.001 billion yuan, monetary funds of 2.170 billion yuan (-45.16% from the beginning of the year), and short-term borrowings of 4.202 billion yuan | Cash flow pressure during the transition period will pass through to delivery resources and response speed |
| Kingdee (First Half 2026, Summary of Third-Party Research Reports) | Revenue RMB 3.625 billion (+13.6%), net profit attributable to shareholders RMB 54 million (loss of RMB 98 million in the same period last year), cloud accounted for 86%, ARR RMB 4.413 billion (+18.3%), AI-native RMB 296 million (+189%) | Turnaround through expense discipline and subscription stock, with a different path from the former |
| A certain small and medium-sized ERP service provider (2025 annual report, cited from a research report) | Revenue from software products and technical services RMB 101 million (+6.41%), gross margin 17.75%→11.08%, accounts receivable RMB 127 million accounting for 48.21% of total assets, operating cash flow -RMB 19.9064 million | Pure project-based + high receivables is the most fragile type of counterparty |
Data hierarchy note:Xentral's financial data comes from much. Consulting's November 2026 overview, which compiled German media reports, and is not an audit report; the data for Yonyou, Kingdee, and small and medium-sized service providers comes from summary articles based on third-party research reports and industry databases, without checking the original company announcements item by item, and may represent semi-annual unaudited figures; IBM's figures come from media reports in early November on its financial statements and share price. These figures are used to assess trends and magnitude and cannot replace the original financial statements used in supplier due diligence.
Viewed together, a useful stratification emerges: what is truly dangerous is not “small and beautiful,” but the category that is purely project-based, with customer concentration, high receivables, and no subscription base. Their products may be easy to use and their teams may know the domain, but a single delayed payment from a major customer or a missed financing round can cause delivery resources to collapse before the product does. And what ERP fears most is precisely “no one maintaining it in the third year after launch.”
A counterintuitive judgment: When choosing a vendor, don't just look at how many agents it talks about; look at how much of its revenue is predictable. Subscriptions and ARR are the buffer; project-based revenue is the heartbeat—once the heartbeat stops, the people are gone immediately. Models like Odoo, with public pricing and per-user monthly billing, often have stronger anti-cyclical capability, at the cost of accepting the trade-off between standardization and customization.
IV. Four Practical Recommendations for CIOs and Procurement Leaders
First, add a “Counterparty Health” column to the scorecard and give it real weight. The four items of documentation required before signing the contract are: audited financial statements for the past two years or an equivalent credit report; the timing and amount of the most recent financing round (if there has been no financing for more than three years, clarify how long the cash can last); customer concentration; R&D expense ratio and changes in R&D personnel. These four items do not involve confidential information, and reputable vendors can provide them.
Second, treat “exit” as a feature to be accepted. Three things are required: data can be exported (master data, documents, attachments, configurations, in usable formats rather than only PDF); custom code and configuration documentation can be delivered (including who maintains them and how the next version will remain compatible); transition-period support (no less than 6 months of technical assistance after termination, at the agreed rate). Open-source solutions have a structural advantage here—the source code is in your hands, which is also why many enterprises this year have written “whether it can be self-hosted” into their tenders. But it must be made clear: access to source code does not mean you can maintain it.
Third, change the liability clause from a "price multiple" to "scenario-based tiers." Stop accepting the vague "no more than 100% of the total contract amount." Tier by scenario: data loss and leakage, business interruption exceeding the agreed duration, erroneous actions leading to external commitments or payments, and regulatory penalties—each with its own cap and remediation obligations; at the same time, clarify the re-acceptance responsibility after "capability changes"—problems arising after a change cannot be blocked by using the acceptance conclusions from before the change.
Fourth, price the "can't switch away" risk in advance. Run a stress test: suppose the supplier stops service tomorrow—how long would it take you to recover, how much would it cost, and who could take over? If the answer exceeds six months or more than half the original contract amount, the true risk of this procurement is far higher than the quoted price. There are three mitigation measures: use open standards for key interfaces, keep readable copies of core data in your own database, and separate the implementation partner from the product vendor.
The question most easily skipped: If this vendor is acquired next year, will the price protection, service levels, and data ownership in the contract still hold? There have been quite a few M&A deals in the ERP and industrial software space this year; acquirers usually retain customers and restructure terms. Asking before signing, "What rights do we have upon a change of control?", is far cheaper than negotiating afterward.
Back to Gordon's case. What it reminds us of is not the correct but empty platitude that “ERP has risks,” but a sharper judgment: what you are buying is not a software package, but a counterparty that must continue to deliver and upgrade over the next five to ten years, and must still be there when something happens to you.
Over the past two decades, that line could get by on “choosing a big vendor.” Not anymore—the market is deconstructing, suites are being broken into point solutions, point solutions are being strung together by an orchestration layer, and vendor size and stability are no longer inherently proportional; maintenance end-of-life dates for old platforms like ECC are also forcing everyone to make a decision. In years like this, factoring counterparty risk into selection is not conservatism—it is doing the full accounting.
Sources: Shang Zhuangyuan, “How Can One ERP Kill a 50-Year-Old Company?” (2026, including publicly reported information on Gordon Company’s 2026-09-14 bankruptcy filing, the Vision33 system, US$2 million investment, and US$15 million claim); much. Consulting, “Odoo News November 2026” (November 2026, including a monthly roundup of Xentral personnel and 2024 financial data, Odoo’s Munich office, and adjustments to industry-specific events); BigGo Finance, “IBM Joins Forces with SAP to Launch Cloud ERP Innovation Support for Mid-Market Enterprises” (early November 2026, including the end of standard maintenance for Business Suite 7 at the end of 2027, termination of extended support in 2030, IBM Q2 financial results and stock price data, and IBM IBV’s 68% survey); Toutiao, “ERP Ushers in a Super Upgrade! What Huge Changes Are Happening to Enterprise Core Systems in 2026?” (including the US$150 billion market and the challenger list of Rillet, Pennylane, Ramp, Light, DualEntry, Campfire, Everest, Digits, etc., as well as the judgment on the orchestration layer); Three Cobblers Report and Toubao Science and Innovation Network industry database (summary of 2026 H1 financial data for Yonyou and Kingdee; 2025 annual report metrics for small and medium-sized ERP service providers); third-party reviews such as epicflow’s “Best SAP Alternatives” and Parameter LLC’s “Odoo vs SAP (2026)” (Odoo’s public price list and two subscription tiers, with pages marked as verified from 2026-11-05 to 11-07).
Data basis note: The product model in the Gordon case is not explicitly identified in public materials; the claim is the plaintiff's assertion and has not been adjudicated, and this article only cites its structural implications. The Xentral data come from a German media report relayed through an industry overview and are not audited figures. Data on Yonyou, Kingdee, and small and medium-sized service providers come from third-party research reports and database compilations and were not checked one by one against the original company announcements; they may be on a semi-annual unaudited basis. IBM data come from media reports on financial statements. The $150 billion market size and challenger list come from media relay of industry views, with no original research institution or statistical boundaries identified. Odoo pricing is a third-party review page citing odoo.com public pricing (Standard and Custom tiers, per user per month, annual payment); other reviews give different currency bases, and at procurement the official website's current pricing shall prevail. On 2026-11-02 and 11-03, this account covered respectively the Q3 earnings divergence among international vendors and AI commercialization in Yonyou's Q3 report; this article examines counterparty risk and contractual liability structure from a buyer's perspective and does not repeat their performance breakdown.
Zhongdi Information (Zodioo) · Enterprise Digitalization Insights · 2026-11-10
