LinkedIn "wisdom"

by "Stavros"

How the sunk cost of a multi-year vendor contract quietly reshapes technical strategy long after the original decision-makers have left the organization

BY STAVROS · AUGUST 13, 2026
How the sunk cost of a multi-year vendor contract quietly reshapes technical strategy long after the original decision-makers have left the organization infographic

The proof is indisputable.

THE INSIGHT

I inherited an architecture once that made no sense to me. Three overlapping data tools, a message bus nobody loved, and a set of services running in a region we had no customers in. I spent weeks looking for the engineering reasoning behind it. There wasn't any. The reasoning was in a commercial agreement signed four years earlier by someone I had never met, and the whole shape of the platform was an attempt to spend that money before the term ran out.

That experience changed how I read technical decisions. Look at the numbers. Korn Ferry put average CIO tenure at 4.6 years across the top 1,000 US companies. Vestd's 2025 FTSE100 report puts CTO tenure at about 4 years and 6 months. Meanwhile the agreements those people sign, cloud committed spend deals, ERP terms, the strategic platform contracts that actually determine what your systems look like, routinely run three to five years, and sometimes considerably longer. The math means a meaningful share of the most consequential technical commitments in any large organization will be lived with, renewed, or unwound by somebody who was not in the room for the trade-off.

The classic sunk cost story doesn't quite fit here, and the way it fails is the interesting part. Staw's work on escalating commitment found that people throw good money after bad mostly when they were personally responsible for the original choice. A new leader has none of that ego in the decision. What they have instead is worse, because the lock-in is structural rather than psychological. The discount tiers are real. The shortfall penalties are real. Microsoft's own documentation describes a MACC as a commitment to a specified level of spend over a defined period, and if you don't get there you pay the difference anyway. So a perfectly rational architect ends up routing new workloads toward vendor native services, not because they are the best fit, but because the commitment has to be burned down. Your FinOps team optimizes a workload, usage drops, and suddenly efficiency has created a liability.

Then there is the pace. Zylo reports that companies handle around 211 SaaS renewals a year, and renewal calendars are typically managed on a 30, 60, 90 day horizon. That is a procurement rhythm making decisions with a five year technical half life. The CMA found cloud customers switch providers under 1% annually. When 37signals left the cloud, the engineering was the easy part. They had to wait for contract commitments to expire before they saw a clean year of savings.

Every renewal is an architecture decision. Most organizations treat contract governance and architecture governance as separate functions with separate calendars and separate owners, and Gartner has been pointing at that seam for years without much changing.

The uncomfortable question I now ask new teams: show me the contract expiry dates, and I will show you your real roadmap.

THE EVIDENCE

Research brief: when inherited vendor contracts quietly become architecture decisions

Research findings only for a LinkedIn post on the thesis that the sunk cost of multi-year vendor contracts reshapes technical strategy long after the original decision-makers leave, and that contract governance is really architecture governance. No post drafted. Data organized by theme with named sources and years. Vendor-marketing versus independent research is flagged throughout.

TL;DR

  • The core mismatch is real and quantifiable: average CIO tenure is 4.6 years (Korn Ferry, top 1,000 US firms, 2020) and average CTO tenure is 4 years 6 months (Vestd FTSE100 report, 2025), while cloud committed-spend deals, ERP contracts and strategic SaaS routinely run 3 to 5 years, so a large share of multi-year commitments outlive the executive who signed them.
  • Committed-spend mechanics turn contracts into de facto architecture: AWS EDPs, Azure MACC and Google CUDs reward routing spend to vendor-native services to "burn down" the commitment, and discounts of roughly 15 to 57 percent plus shortfall penalties make switching or de-scoping expensive.
  • The governance gap is structural: procurement and finance own contracts and renewals (companies average 211 SaaS renewals a year, per Zylo) while engineering and architecture own technical decisions, and Gartner explicitly frames harmonizing these functions as an unsolved organizational problem.

Key findings by theme

1. Contract lengths: multi-year is common for strategic systems, but the "average" is shorter than people assume
  • Vertice reports the average SaaS contract length is 1.5 years, "indicating that many businesses still lean toward shorter-term flexibility," and notes business-critical platforms such as CRMs carry longer commitments than average. [1][1]
  • Multi-year deals are a minority of contracts but concentrate in the highest-value, most architecturally central systems. One benchmark cited multi-year agreements at roughly 11 percent of agreements, with churn of 8.5 percent versus 16 percent for monthly plans.
  • Cloud committed-spend deals: the AWS Enterprise Discount Program (EDP) requires roughly $1M+/year spend and runs "typically one to three years" (up to five in some sources). Azure MACC is "typically one to three years." Google CUDs are 1-year or 3-year. [2]
  • ERP and large government deals run far longer: the UK Home Office Oracle E-Business Suite/Fujitsu deal was a seven-year contract (2009 to 2016). [3]
  • Data platforms use multi-year capacity commitments: Snowflake capacity contracts have a $25,000/year minimum with discounts rising by contract length (2 to 3 year terms preferred), and Databricks pre-purchase DBU commitments are "the standard commercial model for enterprise customers." [4][5]
  • Trend: more organizations are reportedly opting for multi-year SaaS deals to manage SaaS inflation, though vendors are "dialling back multi-year discount incentives." Per Vertice's SaaS Inflation Index, as of January 2025 SaaS pricing was up 11.4 percent year over year, "a stark difference from the 2.7% average market inflation rate of G7 countries."
  • Source-quality flag: much category-level contract-length data comes from SaaS-management vendors (Vendr, Vertice, Zylo, Torii) who sell negotiation services, so treat their benchmarks as directional rather than independent.
2. Executive tenure versus contract length: the mismatch is the whole story
  • CIO tenure: Korn Ferry (analysis of top 1,000 US companies by revenue, published Jan 2020) found average CIO tenure of 4.6 years, versus CEO tenure of 6.9 years, which was "down from an average of 8.0 years when Korn Ferry conducted the same analysis in 2016." An earlier Korn Ferry survey (late 2016) put CIO tenure at 4.3 years, second-lowest of five C-suite titles. [6]
  • CIO tenure by sector (Korn Ferry 2020): "CIOs in the energy sector have the longest average tenure at 5.3 years. The shortest-tenured CIOs are in the healthcare sector at an average of 3.9 years."
  • CTO tenure: Vestd's C-Suite Churn Report 2025 (FTSE100) found CTO average tenure of 4 years 6 months (53.5 months); a separate Pave analysis of 14,000 executives is cited for other C-suite roles. [7]
  • Broader trend: overall C-suite tenure has been decreasing while age increased (Korn Ferry). One recruiter estimate put average executive tenure at about 4.8 years in 2026 versus about 7.5 years a decade earlier (JRG Partners, treat as a search-firm estimate, not peer-reviewed). [8]
  • The direct comparison: a 4 to 5 year executive tenure against 3 to 5 year committed-spend and ERP contracts means a meaningful share of strategic commitments are inherited, renewed, or exited by someone other than the original signer. Many decisions get made "for" a successor.
3. Sunk cost and escalation of commitment: strong academic grounding
  • Foundational work: Barry M. Staw, "Knee-deep in the Big Muddy: A study of escalating commitment to a chosen course of action," Organizational Behavior and Human Performance, 16(1), 1976, pp. 27 to 44. In the experiment, business-school students allocated funds between divisions of a hypothetical company; those personally responsible for the initial choice invested disproportionately more in the underperforming division after receiving negative feedback. Staw's summary: people "committed the greatest amount of resources to a previously chosen course of action when they were personally responsible for negative consequences." [9]
  • Replication and its limits: the Schoorman/Holahan review notes "the effect of choice found by Staw (1976) has been replicated in at least eight other published experiments" (Bazerman, Beekun & Schoorman 1982; Bazerman, Guiliano & Appelman 1984; Caldwell & O'Reilly 1982; and others). Flag: replication is contested. Martens & Orzen (2021, European Economic Review) and a 2021 Economics & Business Review study (N=365) both failed to reproduce the escalation effect, so present the mechanism as well-established but not unchallenged.
  • IT-specific escalation: Mark Keil et al., including "A Cross-Cultural Study on Escalation of Commitment Behavior in Software Projects," MIS Quarterly 24(2), 2000, pp. 299 to 325. Keil, Mann and Rai (2000) found that between 30 and 40 percent of IS/software projects exhibit some degree of escalation.
  • Keil (1995) established that the level of sunk cost can influence decision-makers' willingness to continue an IS project. [10]
  • Application to contracts: the literature is about projects, not vendor contracts specifically. Extending it to contract renewals is an inference, not a directly-studied finding, and should be flagged as such. The mechanism (personal responsibility for the original decision drives escalation) actually implies a nuance worth stating: a NEW leader who did not sign the original deal may face LESS psychological escalation pressure but MORE structural and contractual lock-in.
4. Vendor lock-in, switching costs, and committed-spend shaping architecture
  • Lock-in sentiment: Parallels' 2026 State of Cloud Computing Survey (540 IT pros, US/UK/Germany, fielded Nov 2025) found 94 percent of organizations concerned about vendor lock-in, with 57 percent citing fears over future support and 46 percent citing uncertain product roadmaps. [11]
  • Switching is rare: the UK CMA found customers rarely switch cloud providers, less than 1 percent annually, and that multi-cloud adoption remains limited among SMEs.
  • Egress fees as a switching barrier: research cited by the CMA put egress fees at an average of about 6 percent of organizations' cloud storage costs. AWS standard egress was $0.09/GB for the first 10TB/month (2024). [12]
  • Committed-spend shapes architecture (the crux of the thesis): AWS EDP purchases through AWS Marketplace count toward the commitment (capped at 25 percent), creating "a powerful incentive for buyers to consolidate their software procurement through the marketplace." Mission Cloud states committing to an EDP "may influence a company's cloud architecture decisions by encouraging broader adoption of AWS services to maximize the value of their spend commitment" and that overcommitments "can trap you in a sub-optimal architecture just to keep your usage growing sufficiently." [13]
  • Azure MACC: Microsoft's own docs define it as "a contractual agreement in which your organization commits to a specified level of Azure spending over a defined period," and note customers "can contribute to their MACC by purchasing MACC eligible third-party solutions in Microsoft Marketplace." A shortfall triggers a charge for the remaining balance. Independent negotiation-advisory firms (LicenseQ, Redress Compliance) warn customers "end up paying shortfall charges or buying technology they didn't need just to 'burn down' the commitment," and warn of "architecture drift" where FinOps optimization wins reduce burn and leave a shortfall. [14][15]
  • Google CUDs: 1-year and 3-year terms; resource-based Compute Engine discounts of about 37 percent (1yr) to 55 percent (3yr), up to about 57 percent; flexible/spend-based CUDs of 28 percent (1yr) and 46 percent (3yr). CUDs are non-cancellable and apply to Google-native services, reinforcing native architecture.
  • FinOps framing: the FinOps Foundation calls commitment-based discounts "the largest percentage discount you can achieve in cloud" and warns of the governance failure of "Tech teams autonomously making commitments without considering WACC/NPV" and "Finance folks buying without proper understanding of planned infrastructure changes." This is the governance gap stated in FinOps's own words. [16]
  • Regulatory attention: the UK CMA ran a multi-year cloud market investigation; in July 2025 its report estimated AWS and Azure each at 30 to 40 percent of UK cloud spend, Google 5 to 10 percent. On 31 March 2026 the CMA accepted voluntary commitments from AWS and Microsoft on egress and interoperability rather than imposing Strategic Market Status, and opened a new investigation into Microsoft's software licensing.
  • EU Data Act (Regulation 2023/2854): applied from 12 September 2025; from 12 January 2027 cloud providers serving EU customers are prohibited from charging switching/egress fees. In the interim switching charges must be cost-covering. Google (11 Jan 2024), AWS (5 Mar 2024) and Microsoft (13 Mar 2024) all launched free-egress-for-switching programs in anticipation. Microsoft extended its free switching window from 60 to 180 days.
5. Renewal dynamics: renewals are financial events, not architecture reviews
  • Volume: Zylo reports companies average 211 SaaS renewals per year, "and more in the enterprise," which forces procurement to triage rather than deeply evaluate each one. [17]
  • Ownership: Zylo positions renewal management squarely with Procurement and SAM teams, framing it as a financial and optimization exercise (license true-ups, cost allocation), with IT and business units consulted. This is direct evidence that renewals are owned as procurement events. [18]
  • Auto-renewal prevalence: SaaS-management vendors describe auto-renewal as the default trap; a Vendr-referenced practice is inserting a "no auto-renewal; opt-in renewal only" redline, and companies that adopt it "consistently cut their effective SaaS spend by mid-single-digit to low-double-digit percentages." One vendor claims 60 percent of SaaS contracts renew at inflated counts (marketing claim, flag). [19]
  • Retention as the vendor mirror image: KeyBanc Capital Markets' 2024 Private SaaS Survey found gross retention around 90 percent and net retention around 101 percent; these high renewal/retention numbers reflect how sticky and rarely-switched these systems are. [20]
  • Lead time: renewal calendars are recommended to look 30/60/90 days ahead, which is short relative to the multi-year architectural consequences being locked in. [17]
6. The governance gap: contract governance and architecture governance are separate functions
  • Gartner explicitly frames the separation as a problem to solve: "IT Vendor Management and Procurement Must Become Collaborators, Not Competitors," arguing that "IT VM, procurement, and sourcing must collaborate to define responsibilities and decision authority." [21]
  • Gartner's operating-model research for sourcing, procurement and vendor management (SPVM) 2024/2025 treats these as functions that must build "stakeholder relationships" with the rest of IT, implying they are structurally separate from enterprise architecture.
  • FinOps Foundation's commitment-management guidance (above) is the clearest statement that the two sides (finance/procurement making commitments; engineering changing infrastructure) operate without a shared decision process.
  • Concept of "renewal calendars driving roadmaps": the raw materials are here (Zylo renewal calendars, FinOps commitment tracking) but there is no single named analyst report using the exact phrase "contract-driven architecture." That specific framing is thinly sourced and should be presented as the author's synthesis, not a cited term.
7. Case studies: inherited contracts constraining successors
  • UK NHS National Programme for IT (NPfIT): green-lit 2002, main supplier contracts let in 2003 under the Labour government. CSC's Local Service Provider contract was originally valued at £3.1 billion. The Coalition government elected in 2010 inherited it and announced in September 2011 it would "urgently dismantle" NPfIT. Total programme cost is cited at roughly £10 billion (NAO figures ranged £9.6bn to £12.7bn; a DoH FOI put end-of-life cost at £10.1bn). The Public Accounts Committee found: "Despite the contractor's weak performance, the Department of Health is itself in a weak position in its attempts to renegotiate the contracts." The Department argued cancelling could be "more expensive than allowing it to complete." This is the strongest example of a successor administration locked into a predecessor's deal. [22]
  • Queensland Health payroll / IBM: contract signed December 2007 under the Bligh (Labor) government; original budget about A$6.19 million, IBM fees A$25.7 million, final cost about A$1.2 billion over eight years, with over 78,000 employees hit by pay problems. The successor Newman (LNP) government received the 2013 Commission of Inquiry (Commissioner Chesterman), which found "IBM should never have been appointed as the prime contractor" and that the prior government's decision to settle with IBM rather than litigate foreclosed the successor's ability to recover funds. Newman called it "arguably the worst failure of public administration in Australia's history." (IBM's self-defence that its fees were "less than 2 percent" of the total is a litigation-posture statement, flag.) [23]
  • UK Home Office Oracle/Fujitsu: locked into a £330m contract until 2016, signed 2009 for Oracle E-Business Suite across 29,518 users. The Coalition government could not drop it despite moving ERP to a Steria shared-services centre, so it effectively paid twice, a clean small-scale example of inherited term lock-in. [3]
  • Cloud repatriation (contract mechanics): 37signals (Basecamp/HEY), led by CTO David Heinemeier Hansson, exited AWS/Google Cloud starting 2022 after a $3.2M annual cloud bill; spent about $700,000 on Dell servers and cut the bill by about $2M/year, projecting over $10M savings across five years. Notably, contract mechanics gated the exit: "it took until the end of the year for various contract commitments to expire" before 2024 was "the first clean year of savings," and AWS waived about $250,000 in egress fees to let them delete their account. This shows contracts, not just technology, govern the timeline of architectural change. [24]
8. Counterpoints worth acknowledging
  • Real discounts: multi-year SaaS deals are commonly cited at 15 to 25 percent savings with price protection against annual increases; annual prepay discounts have a median around 18 percent (Vendr 2024 dataset via Torii); committed cloud spend yields far larger discounts (Google CUDs up to about 55 to 57 percent; Snowflake capacity 15 to 40 percent). [25]
  • But term is a weak lever versus volume: a large analysis of about 15,000 contracts (Mostly Metrics) found multi-year term commitments yield only about a 2 to 3 percent additional discount, and concludes "volume discounting has a massive benefit over term discounting." This undercuts the argument that long terms are worth the lost flexibility. [26][26]
  • Lock-in fears may be partly overstated: hyperscalers have removed egress fees for full account exits since 2024, the EU Data Act removes switching fees from 2027, and full cloud repatriation remains rare. Per IDC (October 2024, Server and Storage Workloads Survey, via Digital Chiefs), only 8 to 9 percent of companies are planning full repatriation; the widely-cited 83 to 86 percent figures (Barclays CIO Survey Q4 2024) refer to moving at least one workload, not full exits. Committed-spend deals like EDPs also offer flexibility to shift workloads between services without losing the discount tier. [27]
  • Price protection has real value amid double-digit SaaS inflation, and predictable budgets and vendor relationships are legitimate reasons to commit.

Caveats and data-quality notes

  • Much of the contract-length, discount and renewal data comes from SaaS-procurement vendors (Vendr, Vertice, Zylo, Torii and similar firms) whose business is negotiation; their benchmarks are directional and self-interested, not independent research.
  • Executive-tenure figures come from search and recruiting firms (Korn Ferry, Spencer Stuart, Vestd, Pave) using different samples (US top 1,000, FTSE100, compensation databases), so numbers cluster around 4 to 5 years but are not strictly comparable.
  • The escalation-of-commitment literature is about projects, not contract renewals; applying it to renewals is a reasoned extension, not a directly measured result, and recent studies (2021) have failed to replicate the core effect, so it is well-established but contested. The "new leader" twist cuts both ways: less psychological escalation, more structural lock-in.
  • The specific phrase "contract governance is architecture governance" is the author's thesis, not an established analyst term; the supporting evidence (Gartner on procurement/VM separation, FinOps on commitment governance) is real but assembled here rather than pre-packaged.
  • The strongest "buy technology you did not need to burn down the commitment" claims come from Microsoft-negotiation advisory firms and ISV marketing, not from FinOps Foundation or the hyperscalers; the burn-down mechanic itself is confirmed by Microsoft's and Google's own documentation, which are authoritative on how the programs work but are vendor self-descriptions on whether they are good for the buyer.
  • Several sources carry 2026 datelines with forward-looking or promotional framing; product mechanics from Microsoft and Google docs are authoritative for how programs work but should not be read as neutral on buyer value.
  1. Vertice — https://www.vertice.one/blog/reasons-to-consider-a-multi-year-saas-contract
  2. nOps + 2 — https://www.nops.io/blog/ultimate-guide-aws-edp/
  3. theregister — https://www.theregister.com/2014/12/08/government_locked_into_330m_oracle_contract/
  4. Select — https://select.dev/posts/snowflake-pricing
  5. Softwarepricingguide — https://softwarepricingguide.com/snowflake-vs-databricks-pricing-2025-which-data-platform-costs-more-and-what-you-actually-get/
  6. RetailWire — https://retailwire.com/discussion/why-is-the-turnover-rate-so-high-for-technology-leaders/
  7. Vestd — https://www.vestd.com/blog/c-suite-churn-report-2025
  8. JRG Partners LLC — https://www.jrgpartners.com/executive-turnover-statistics-2026-c-suite-departure-rates-industry/
  9. ScienceDirect — https://www.sciencedirect.com/science/article/abs/pii/0030507376900052
  10. ScienceDirect — https://www.sciencedirect.com/science/article/abs/pii/S0959802299000041
  11. Parallels — https://www.parallels.com/newsroom/news/press-releases/20260217-cloud-survey/
  12. DigitalDefynd — https://digitaldefynd.com/IQ/cto-navigating-cloud-vendor-lock-in/
  13. Automatum + 3 — https://www.automatum.io/blog-posts/aws-edp-enterprise-discount-program-guide
  14. Microsoft Learn — https://learn.microsoft.com/en-us/partner-center/marketplace-offers/azure-consumption-commitment-enrollment
  15. LicenseQ — https://licenseq.com/understanding-microsoft-macc-and-azure-commitment-traps/
  16. Medium — https://christophershayan.medium.com/cloud-journey-part-10-finops-fe545344d365
  17. Zylo — https://zylo.com/blog/guide-saas-renewal
  18. Zylo — https://zylo.com/blog/software-renewal
  19. SeatCompress — https://www.seatcompress.com/blog/saas-auto-renewal-traps
  20. Zenskar — https://www.zenskar.com/blog/renewal-rate--vs-retention-rate
  21. Gartner — https://www.gartner.com/en/documents/3390418
  22. UK Parliament — https://committees.parliament.uk/committee/127/public-accounts-committee/news/181704/npfit-report/
  23. Henrico Dolfing — https://www.henricodolfing.ch/en/case-study-9-the-payroll-system-that-cost-queensland-health-au1-25-billion/
  24. cbinsights + 3 — https://www.cbinsights.com/company/aws
  25. SoftwareSeni — https://www.softwareseni.com/how-to-negotiate-better-saas-contracts-and-lock-in-lower-prices/
  26. Mostly metrics — https://www.mostlymetrics.com/p/your-guide-to-negotiating-multi-year-deals
  27. Vendorbenchmark — https://vendorbenchmark.com/benchmarks/cloud-infrastructure-pricing-guide

EDITORIAL BRIEF

Commissioned from our research desk. Subject to final editorial discretion.

How the sunk cost of a multi-year vendor contract quietly reshapes technical strategy long after the original decision-makers have left the organization. Explore how inherited commitments constrain new leaders who weren't in the room for the original trade-off, and how contract renewal cycles become de facto architecture decisions. Research data on average enterprise SaaS contract lengths versus average tenure of the CTOs who sign them. The takeaway is that contract governance is architecture governance, and most organizations treat them as completely separate functions.