The Second Ledger

If you want to argue that outsourcing destroys value, the honest place to begin is with the largest study that says it does not.

In 2022, Lahiri, Karna, Kalubandi and Edacherian published a meta-analysis in the Journal of Business Research covering 121 samples drawn from 106 primary studies over 28 years. Their headline finding is that the outsourcing–firm performance relationship is positive. Not neutral. Positive.1 A researcher who wanted to defend the practice could stop reading there, and many executives effectively have.

But the same paper contains the finding that matters more, and it is the hinge on which everything else in this article turns. The association is significantly stronger for non-core outsourcing than for core outsourcing. Interestingly, it does not vary meaningfully between manufacturing and services.1 The question is not whether you outsource. It is what you outsource.

This is consistent with the earlier literature rather than a departure from it. Lahiri's 2016 review of 57 empirical studies across 47 journals concluded that the answer to "does outsourcing improve firm performance" had remained genuinely elusive for two decades.2 Gilley and Rasheed found no significant direct effect at all in 2000; performance depended entirely on firm strategy and environmental dynamism.3 The field's consensus, to the extent it has one, is that outsourcing is a contingent practice whose value is determined almost entirely by what is being handed over and what is retained.

Outsourcing the periphery pays. Outsourcing capability does not. The entire argument is about which is which — and about the fact that the line keeps moving.

That is the framing I will defend. Not "offshore is bad." Not "vendors are bad." The claim is narrower and, I think, harder to dismiss: the categories of work most commonly outsourced by large enterprises — customer support and application development and maintenance — have migrated from periphery to core over the last fifteen years, while the sourcing model applied to them has not changed. Firms are still using a peripheral playbook on work that has become central, and the meta-analytic evidence says that is precisely the configuration where returns fall away.

Two things have moved the line. First, for most large firms, the digital surface is now the product — the app, the portal, the support channel, the recovery experience when something breaks. Prahalad and Hamel argued in 1990 that the discipline of a firm is to identify and protect what it must be uniquely good at.37 That set now includes competences that in 2008 were plausibly clerical. Second, the pace of technical change has raised the premium on adaptability, and adaptability, as we will see, is exactly what outsourcing trades away.

The cleanest evidence on customer-facing outsourcing comes from a study that ran for nine years and was, in one important respect, not the study its authors expected to publish.

Whitaker, Krishnan, Fornell and Morgeson analyzed the sourcing decisions of 150 North American firms and business units from 1998 to 2006, using the American Customer Satisfaction Index as the outcome variable and coding tens of thousands of news reports to identify sourcing events. Their finding: front-office outsourcing — customer service — is negatively associated with customer satisfaction. Back-office offshoring of IT, HR, finance and R&D shows no such decline.4

Now the part that ought to embarrass anyone who wanted a simple anti-offshoring story, including me. Front-office onshore outsourcing was negatively associated with satisfaction too — at a similar magnitude. The authors' own reading is that the damage comes from the firm boundary dimension rather than the geographic location dimension.4 Moving customer contact to a domestic vendor in Ohio does roughly what moving it to a vendor in Manila does. The problem is not where the agent sits. It is that the agent does not work for you.

This should change how the debate is conducted. A great deal of public argument about offshore support is really about accent, geography, and something less defensible than either. The research does not support that argument. It supports a structural one.

Why the boundary matters: the mechanism

The Global Call Center Report — a survey of thousands of centers across more than a dozen countries, led from Cornell — supplies the mechanism in unusually concrete terms. Comparing subcontracted centers against in-house centers, across most countries in the study: subcontractors provide less training (about 14 days against 20), pay roughly 18 percent lower wages, assign lower job discretion (48 percent of subcontractor jobs classified as low-discretion, against 35 percent in-house), monitor performance more intensively, and run annual turnover of about 25 percent against 19 percent.6 A parallel US analysis found outsourced centres carrying higher turnover and a less skilled workforce than in-house centres while enjoying only a modest labour cost advantage.8 Industry-association figures for the largest offshore voice markets report annual attrition in the same territory or well above it, though those are self-reported trade numbers rather than peer-reviewed measurement.38

The same job, two employment systems

None of those five differences is a mystery, and none is a slur on the people doing the work. They are the predictable output of a contract. A vendor competing on price cannot fund twenty days of training against a competitor bidding fourteen. A vendor measured on average handle time cannot grant discretion, because discretion produces variance. A vendor whose agents are assigned to a single client account cannot offer the internal mobility that keeps good people for five years. The employment system is downstream of the commercial structure, and service quality is downstream of the employment system.

Batt's related work makes the positive case from the same data: centers that invest in workforce skills and give agents the discretion to actually solve a customer's problem show lower turnover, better service quality, and higher revenue.7 The lever exists. Standard outsourcing contracts are structured so that nobody can pull it — the client will not pay for it and the vendor cannot fund it.

There is a further cost that sits entirely outside the vendor's SLA. A 2023 study in Industrial Relations traced what happens to the client's own remaining employees when third-party vendors make errors: internal staff absorb the correction work, absorb the resulting customer anger, and show measurable increases in emotional exhaustion and absenteeism.9 The vendor's cost per contact went down. Somebody else's cost went up, in a different budget, under a different name.

A CFO can reasonably ask: so what? Satisfaction is a survey. Show me the money.

This is the part of the literature that is stronger than most operators realize. Customer satisfaction, measured by ACSI, has been repeatedly linked to hard financial outcomes. Anderson, Fornell and Mazvancheryl found a significant positive relationship between satisfaction and contemporaneous market value metrics including Tobin's Q.12 Fornell, Mithas, Morgeson and Krishnan went further in the Journal of Marketing: a portfolio built on high ACSI scores beat the market by considerable margins, at low systematic risk — and, critically, the news of ACSI results did not move share prices at the time of publication, implying the market was not pricing the information in.10 The authors returned to the question later in Marketing Science with updated data and reported that the above-market returns persisted and remained both economically and statistically significant.11

The Michigan researchers who ran the offshoring study did the obvious arithmetic. Prior work associates the average ACSI decline with a drop of roughly 1 to 5 percent in a firm's market capitalization, depending on industry.5 Set that against the savings line on a customer-service outsourcing business case. For a company with a $20 billion market capitalization, the low end of that range is $200 million. The support contract it is defending might save $30 million a year.

I want to be careful here, because this is where advocacy usually outruns evidence. That 1–5 percent figure is an association drawn from a body of work on satisfaction and firm value; it is not a causal estimate of what a specific outsourcing decision will do to a specific company's stock. The honest formulation is this: satisfaction is an economic asset with a measurable link to equity value, the market appears to underprice it, and a decision that predictably reduces it is therefore a decision to convert an underpriced long-lived asset into a booked short-term saving. That is not a cost reduction. It is a sale of an asset, recorded as income.

Satisfaction is not a soft metric. It is an intangible asset the market undervalues — which is exactly why it is so easy to spend.

Now the harder claim, and the one you asked about specifically: are outsourced IT estates slower to modernize? The evidence says yes, with a well-identified mechanism and a clear boundary condition.

Start with the operational data. The 2018 Accelerate: State of DevOps report, based on nearly 1,900 respondents, examined outsourcing directly. Low-performing teams were 3.9 times more likely to use functional outsourcing overall than elite performers, and 3.2 times more likely to outsource application development, IT operations, or testing and QA specifically. DORA's summary was blunt: outsourcing by function is rarely adopted by elite performers and is linked to lower performance.13

DORA's own caveat is important and I will not bury it. The finding concerns wholesale outsourcing by function — handing an entire discipline to a vendor — and the report explicitly notes that the same reasoning applies to organizations with rigid internal functional silos.14 The report also distinguishes this from integrating individual external engineers into existing internal teams, which does not carry the same penalty. That distinction is not a footnote; it is arguably the whole finding. The damage comes from splitting a delivery system along a contract line, not from the passports of the people involved.

The mechanism: efficiency purchased with adaptability

The strategy literature explains why. Weigelt's 2012 study in the Strategic Management Journal examined firms outsourcing during the emergent stage of a technological innovation, using archival data and two surveys of U.S. banks adopting internet banking. The finding is precise: outsourcing yields efficiency gains up to a point, but it hurts adaptability. Efficiency problems and adaptability problems differ in complexity and structure, and their optimal governance therefore differs — the first benefits from outsourcing, the second is better managed inside the firm.15

Weigelt's earlier 2009 paper isolates the specific failure. Outsourcing gives a firm access to a new technology. It does not give the firm the capacity to assimilate that technology into its business processes and build on it — what she calls integrative capability. Access and integration are different goods, and the literature had rarely distinguished them.16 Anyone who has watched an enterprise buy a platform, hand the implementation to a partner, and then discover three years later that nobody internal can extend it will recognize the finding immediately.

The boundary condition is the useful part for a practitioner. In both Weigelt studies, the firm's absorptive capacity — its ability to recognize, assimilate and apply external knowledge, in Cohen and Levinthal's original formulation — moderates the trade-off.1519 Firms with strong internal technical depth can outsource and still adapt. Firms that have outsourced their way out of internal depth cannot. And absorptive capacity is itself a function of retained internal expertise, which means the trade-off compounds: each round of outsourcing reduces the capacity that would have made the next round safe.

This is what Bettis, Bradley and Hamel called hollowing out in 1992, before most of the current offshore industry existed.17 Handley and Benton put numbers to the managerial consequence in the Journal of Operations Management: capability loss degrades the client's ability to manage the outsourcing relationship itself, which degrades outcomes further.18 You lose the ability to build, then you lose the ability to judge whether what you are being sold is any good, then you lose the ability to negotiate. That is not a slippery-slope argument. It is a documented sequence with a name.

So the answer to "are they slower to modernize" is: yes, and the reason is not vendor incompetence. It is that modernization is an adaptability problem, adaptability is the thing outsourcing trades away, and the internal capacity that would have protected you is the same capacity you spent.

Here is the accounting problem in its purest published form, and it comes from hospitals rather than IT — which is why it is so clarifying.

Toffolutti, Reeves, McKee and Stuckler linked MRSA incidence per 100,000 bed-days to cleanliness surveys across 126 English acute hospital Trusts from 2010 to 2014. Outsourcing cleaning services was associated with greater MRSA incidence, fewer cleaning staff per bed, worse patient perceptions of cleanliness, and worse staff perceptions of handwashing facility availability. It was also associated with lower economic costs — and the authors add the decisive clause: without accounting for the additional costs of treating the infections.20

Read that twice. Both books are open on the same page. The savings are real and they are measured. The harm is real and it is measured. They land in different ledgers, on different timescales, under different owners — and only one of them was ever presented to the committee that made the decision. Litwin, Avgar and Becker found a parallel result in U.S. hospitals: outsourced environmental services workers, more likely to be under-rewarded, undertrained and detached from the care team, associated with higher incidence of healthcare-associated infections.21

I am not claiming a help desk is a hospital. I am claiming the accounting structure is identical, and that IT and support functions have their own version of the untreated infection.

Three enterprise cases: what was booked, and what was not

Booked — appeared in the business case	Not booked — appeared years later
NHS acute Trusts, cleaning Outsourcing associated with lower measured economic cost.Recorded as savings	Higher MRSA incidence, fewer staff per bed, worse cleanliness ratings; treatment costs excluded from the comparison.Not in the model20
Marks & Spencer, IT help desk A decade-long outsourced service desk contract with a global provider.Annual run-rate saving	Attackers impersonated an employee and persuaded a third-party service desk agent to reset credentials; roughly £300m of operating profit.−£300,000,0002223
TSB, core banking migration Platform intended to cut operating costs by about £160m a year.£160m/yr target	Failed migration: £330m total cost, £48.65m in regulatory fines for failures including management of outsourcing risk, chief executive resigned.−£378,650,0002829
Boeing, 787 airframe Outsourced roughly 65% of the airframe; improved return on net assets.RONA improvement	Approximately three years late, billions over budget; Boeing spent about $1bn buying back a supplier's plant to regain control.−$1,000,000,000+32

The M&S case, in detail

In April 2025 Marks & Spencer suffered the most operationally destructive cyberattack in recent UK retail history. Chairman Archie Norman described it as a sophisticated impersonation attack involving a third party: attackers impersonated an employee and convinced a third-party service desk agent to reset credentials, which gave them access to internal systems.22 M&S warned investors the incident would take roughly £300 million off group operating profit.23 In the half-year to 27 September, statutory pre-tax profit fell from £391.9 million to £3.4 million; the company booked £101.6 million of direct incident costs, partly offset by about £100 million of cyber insurance.24 In July 2025, following a competitive procurement, M&S ended its decade-long IT help desk contract with the incumbent provider. The company said the move was routine and unrelated.25 TSB reached a comparable conclusion after its own failure, announcing that an in-house team would take direct ownership of the banking platform and of the contractual relationships with technology suppliers.30

M&S statutory pre-tax profit, half year to late September

I want to be scrupulous about attribution here, because the temptation to overclaim is strong. M&S has not said the outsourcing arrangement caused the breach, and a third-party desk is not automatically worse than an internal one — internal desks have been socially engineered too. What can be said, and what the U.S. and allied cybersecurity agencies have said in a joint advisory, is that the threat group involved specifically targets large companies and their contracted IT help desks as a category.26 That is a government agency identifying an outsourced function as a named attack surface. In the MGM Resorts intrusion using comparable tradecraft, full tenant compromise was reportedly achieved in a single call of under ten minutes; MGM reported roughly $100 million in losses and later settled a class action for $45 million.27

The structural reason is not hard to see. A help desk exists to help people who are locked out. Verification rigor is friction, and friction is measured in average handle time, and average handle time is in the contract. An internal desk that knows the person on the phone has a defense that no SLA can specify. That knowledge is a form of firm-specific human capital, and it is exactly the form that outsourcing dissolves.

The Boeing precedent, and the metric that caused it

None of this is new. In 2001, before the 787 program was formally launched, a Boeing senior technical fellow named L. J. Hart-Smith circulated a paper titled Out-Sourced Profits — The Cornerstone of Successful Subcontracting. Drawing on the Douglas Aircraft DC-10 experience, he argued that outsourcing exports the profits along with the work; that the prime manufacturer absorbs the overruns while subcontractors capture the margin and the decades-long spares business; and that the dollar value of outsourced work is a very poor surrogate for internal cost savings.31

Every one of those predictions held. The 787 ran approximately three years late with billions in overruns, and in 2009 Boeing spent about $1 billion to acquire the underperforming fuselage plant of a supplier that had contributed to the delays. Boeing's own commercial airplanes chief later acknowledged that the company spent far more recovering than it would have spent keeping key technologies closer to home, and that the pendulum had swung too far.32

Hart-Smith identified the culprit as a metric: return on net assets. If your performance measure rewards the removal of assets from the balance sheet, and capability sits on the balance sheet as people and plant, then the measure will reward the removal of capability and will keep doing so until the capability is gone.31 Substitute "IT headcount" for "plant" and you have most enterprise sourcing strategy since 2005.

If capability is what is being spent, the question becomes what it is worth. Two bodies of evidence answer that, and the second one is unusually rigorous because it uses stock returns rather than profits as the dependent variable.

Crook, Todd, Combs, Woehr and Ketchen meta-analyzed 66 studies of the human capital–firm performance relationship for the Journal of Applied Psychology. Human capital relates strongly to performance — and, critically, the relationship is strongest when the human capital in question is not readily tradable in labor markets, and when performance is measured operationally rather than through profit measures subject to appropriation.33

Sit with that moderator, because it is the hinge of this entire article. The human capital that generates competitive advantage is the kind you cannot buy on the open market: firm-specific, tacit, accumulated by working on your systems with your customers under your constraints. Outsourcing is, definitionally, the act of converting a firm-specific capability into a tradable one. It takes the exact form of human capital the evidence says produces advantage and reclassifies it into the form the evidence says does not. The vendor may be excellent. The commodity is still a commodity, and by construction your competitor can buy the same one tomorrow.

You cannot purchase a competitive advantage from a supplier who is also willing to sell it to your competitor. That is not a moral objection. It is a definition.

The equity evidence comes from Alex Edmans, in the Journal of Financial Economics. A value-weighted portfolio of the "100 Best Companies to Work For in America" earned an annual four-factor alpha of 3.5 percent from 1984 to 2009, and 2.1 percent above industry benchmarks, robust to controls for firm characteristics, alternative weighting, and removal of outliers. The Best Companies also showed significantly more positive earnings surprises and announcement returns.34

The methodological point is what makes this compelling rather than merely encouraging. If satisfaction were simply the result of high profits, those profits would already be in the price at the start of the return window, and the firms would not outperform going forward. They did outperform, and the positive earnings surprises indicate the market was repeatedly underestimating them. Edmans's conclusion: the stock market does not fully value intangibles, even when independently verified by a highly public survey of large firms.34

The necessary caveat, from Edmans's own later work with Li, Pu and Zhang in Management Science: the relationship depends on labor market flexibility, and is weaker in countries where employment is more rigid.35 That is a real limit on generalization. It also implies the effect should be strongest in the U.S. market where most of these sourcing decisions are made.

Put the two findings side by side and you have a coherent financial thesis. Firm-specific human capital produces measurable operational advantage. The market systematically underprices it. Therefore a decision that trades firm-specific capability for booked cost savings will look good to the market in the short run — because the market cannot see what was sold — and will show up as underperformance later, when the operational consequences finally reach the income statement in a form nobody can trace back to a sourcing decision made six years earlier.

Anyone who has spent a career in this industry has watched confident theses collapse. Here is the strongest version of the opposing case, stated as its advocates would state it.

Which of these survives? The macro finding and the meta-analysis survive completely, and they constrain the claim: outsourcing per se is not the problem, and a blanket in-house policy is not supported by the evidence. The offshore-versus-onshore correction survives completely, and I think it is the single most useful thing in this article. The selection-on-outcome objection to my case studies is correct as stated — those cases illustrate a mechanism; they do not measure a frequency.

What survives on my side is narrower and, I would argue, sufficient: the core/non-core moderator is robust and replicated; the front-office satisfaction penalty is measured over nine years across 150 firms; the efficiency–adaptability trade-off has a named mechanism, a moderator, and independent operational corroboration; and the human capital evidence tells you precisely which kind of capability is worth keeping. That is enough to change a decision rule. It is not enough to justify a slogan.

If the evidence supports a decision rule rather than a slogan, here is the rule I would put in front of a board.

Four questions before any sourcing decision

Does this work generate knowledge we need in order to make the next decision? If the team doing it is the team that will tell you whether the vendor's next proposal is sound, you cannot outsource it without also outsourcing your ability to evaluate. This is the absorptive-capacity test, and it is the one most often skipped.1518

Does it touch a customer at the moment something has gone wrong? Recovery is where satisfaction is won and lost, and it is the moment that most requires discretion — the thing outsourcing contracts are structurally unable to fund.46

Is the capability firm-specific or market-tradable? If a competitor can buy the identical service from the identical provider next quarter, it is not a source of advantage and should be sourced on price. If it is genuinely yours — your data model, your regulatory position, your customers' particular failure modes — the evidence says keep it and invest in it.33

Are we outsourcing a function, or augmenting a team? The operational data distinguishes these sharply. Integrating external engineers into internal teams does not carry the penalty that handing over an entire function does.14 Most of the value of external capacity is available under the first model. Most of the damage comes from the second.

Model your own second ledger

The following does no research and proves nothing. It performs arithmetic on assumptions you supply, so that the second ledger has a number in it rather than a shrug. The retention and satisfaction inputs are yours to argue about; the point is that they belong in the model at all.

What I would actually do

Keep the escalation path in-house. Whatever else you contract out, the last line — the people who handle the case that has already failed twice, and the engineers who own the systems those cases run through — stays on your payroll, because that is where the firm-specific knowledge is manufactured and where recovery either happens or does not.

Contract for capability transfer and audit it. If a vendor arrangement is not measurably increasing your people's ability to operate the thing without the vendor, it is depleting absorptive capacity on a schedule, and you should know the schedule.

Report both ledgers to the board on the same page. Booked savings next to attrition, first-contact resolution, deployment frequency, change failure rate, and satisfaction — with the sourcing decision named. Most organizations can produce all of these numbers today. They simply never appear in the same document as the savings.

And stop treating headcount as a cost and only as a cost. The finance function has a depreciation schedule for a forklift and none for a twelve-year applications engineer who knows why the settlement job runs at 2 a.m. The forklift is on the balance sheet. She is not. That asymmetry is not a fact about economics; it is a fact about accounting conventions written for a manufacturing economy — and it is the single largest reason capable organizations keep making a decision the research says will cost them.

The caliber of your talent is not an expense line that a disciplined operator minimizes. It is an underpriced, slowly-compounding, hard-to-rebuild asset that the market does not see and your competitors cannot buy. Spend it if you must. But know that you are spending it, and write it in the ledger where somebody will read it.