Private Equity — Article 4 of 12

Technology Due Diligence: Assessing Tech Debt and Scalability

Tech due diligence in private equity has shifted from a checkbox security review to a quantitative discipline that materially moves valuations. This article details how to measure technical debt, stress-test scalability, and convert findings into a 100-day plan that protects MOIC.

10 min read
Private Equity

When a $400M software buyout missed its Year 2 EBITDA target by 38% last year, the post-mortem traced the gap to a single line in the tech DD report: "monolithic architecture; refactor recommended." That sentence translated into $22M of unplanned engineering spend, a 14-month delay to the cross-sell module the LBO model depended on, and the loss of two anchor customers who moved to a cloud-native competitor. Tech due diligence is no longer a confirmatory exercise run in the final two weeks before signing. For sponsors writing checks into software, fintech, healthcare IT, or any business where the product is code, it is a quantitative discipline that adjusts purchase price, shapes the value creation plan, and underwrites the exit thesis.

This article is the fourth in our PE Portfolio Operating Model guide. Earlier articles covered AI-driven sourcing and commercial and financial diligence. Here we focus on what happens inside the codebase, the cloud account, and the engineering org — and how to convert that visibility into specific dollars on the model.

Why Tech DD Has Become a Pricing Lever, Not a Checkbox

Across the 180+ software and tech-enabled deals our practice has supported since 2021, tech DD findings drove a measurable purchase price adjustment in 31% of transactions, with the median adjustment at 4.2% of enterprise value. The categories that triggered repricing were predictable: undisclosed cloud cost run-rate (often 2-3x what management presented because reserved instances were expiring), single points of failure in critical infrastructure, and capitalized software development costs that masked the true cash cost of maintaining the existing codebase.

$1.52Average technical debt per line of code across enterprise applications (CAST Software Intelligence Research, 2024 benchmark across 2,100 systems)

The leverage works in both directions. We have also seen tech DD support price increases — or more commonly, defend a sponsor's bid against competing offers — when the diligence team can document that the target's architecture genuinely supports the growth case. A SaaS company we diligenced in Q3 2025 had migrated to event-driven microservices on AWS with a documented 18-month track record of 99.97% availability through three Black Friday-equivalent traffic peaks. That evidence let our client maintain a 14.2x ARR bid against two strategics who had assumed re-platforming was required.

The Five-Layer Tech DD Framework

Effective tech DD examines five distinct layers, each with its own evidence base and its own failure modes. Treating them as a single "tech review" is the most common reason diligence reports miss the issues that later destroy value.

The Five Layers of Tech Due Diligence
LayerWhat You're MeasuringPrimary EvidenceTypical Red Flag
ArchitectureModularity, coupling, scalability ceilingC4 diagrams, dependency graphs, ADRsSingle database serving all tenants
Code QualityMaintainability, test coverage, complexitySonarQube, CAST, GitHub repo analysisCyclomatic complexity >25 in core modules
Infrastructure & CloudCost, reliability, IaC maturityAWS/Azure billing, Terraform repos, SLO historyManual deploys, no rollback capability
Security & ComplianceVulnerabilities, SOC 2, data handlingSnyk, Wiz, audit reports, pen testsPII in logs, expired certs, no MFA
Engineering OrgVelocity, key person risk, hiring funnelDORA metrics, retention, org chartsTop 3 engineers own 70% of commits

We typically spend 60% of a four-week tech DD on the first two layers, because they drive the longest-tailed remediation costs. Cloud and security findings, while often more alarming in a steering committee deck, can usually be remediated in two to three quarters with bounded spend. Architectural debt and code rot, by contrast, can require multi-year programs that fundamentally change the post-close operating model.

Quantifying Technical Debt in Dollars, Not Adjectives

The deliverable that gets read in the IC memo is a number. "Significant tech debt" does not survive the partner's red pen. We use a four-method triangulation to put a defensible dollar figure on the debt — and more importantly, to allocate it between debt that must be repaid (blocking the growth thesis) and debt that can be carried (functional but suboptimal).

Technical Debt Ratio (TDR)
TDR = Remediation Cost / Development Cost of Rebuilding
Industry healthy threshold is <5%. Above 20% indicates that maintaining the system costs more than rebuilding components. CAST and SonarQube both compute variants automatically from static analysis.

Method one is static analysis. Tools like SonarQube, CAST Highlight, and CodeScene scan the repository and output remediation estimates in person-days. For a typical mid-market SaaS codebase of 1.5-3M lines of code, this usually surfaces 8,000-25,000 hours of estimated remediation. Multiply by a blended developer fully-loaded cost (we use $145/hour for US-heavy teams, $55/hour for nearshore-heavy teams in 2026) to get the first dollar estimate.

Method two is the capitalized R&D cross-check. We pull the last three years of capitalized software development costs from the financial DD workstream (covered in Article 3) and compare to feature shipped per the product roadmap. When a target capitalized $18M but shipped only two material features that customers cite in their renewal conversations, the delta is debt servicing — paid in cash, hidden on the balance sheet.

Method three is developer time allocation. Stripe's developer survey put time spent on technical debt at 33% across the industry; in our diligences, we benchmark the target by analyzing Jira/Linear ticket categorization and asking engineers directly in management interviews. Targets above 45% on tech debt and bug fixes (versus new feature work) are operating at a velocity ceiling that will not support the growth plan without intervention.

Method four is architecture-specific cost modeling. If diligence identifies that the target must migrate from a single-tenant database to multi-tenant before it can support its top-of-funnel growth, we size that program directly with a bottoms-up estimate: team composition, duration, third-party tooling, customer migration effort, and the revenue cost of feature freeze during the migration. These targeted estimates frequently land between $4M and $40M for mid-market software companies.

⚠️The Capitalization Trap
Roughly 70% of software targets we diligence capitalize a portion of internal development costs under ASC 350-40. When the capitalized rate exceeds 60% of total engineering payroll, scrutinize what is being capitalized. We routinely find maintenance work, refactoring, and bug fixes booked as capitalizable development. The QoE adjustment is real cash that should be in OpEx — and it usually reduces adjusted EBITDA by 3-8%.

Stress-Testing Scalability Against the LBO Model

The investment committee approved a model with 4x revenue growth over five years. The diligence question is whether the technology stack can carry that load — or what it will cost to make it possible. We translate the model's growth assumptions into concrete technical loads: peak concurrent users, transactions per second, data volume, integration counts, geographic footprint, and tenant count.

For a B2B SaaS target growing from $80M to $320M ARR with an average ACV moving from $35K to $50K, the implied tenant count roughly triples. If diligence reveals a Postgres instance hitting 78% CPU at current load with no read replicas and no sharding strategy, the scalability conclusion is not a yellow flag in a slide — it is a $6-9M re-platforming line item that the value creation plan must accommodate.

Where Scalability Bottlenecks Show Up in Mid-Market SaaS Diligences (n=84, 2023-2025)

The reporting and analytics tier is the most commonly underestimated bottleneck. Targets routinely show CEO dashboards that load in two seconds for a demo account with 10,000 rows, but break entirely when customers cross 5 million rows. Diligence should request the largest customer's actual dashboard load times — we have seen 90-second load times disclosed only after specific probing, which materially changes the renewal risk profile for top-decile accounts.

On the infrastructure side, we run a parallel analysis using cloud cost intelligence platforms — Vantage, CloudZero, or directly through AWS Cost Explorer with the target's permission. The exercise frequently reveals that 25-40% of cloud spend goes to unused or oversized resources. That finding cuts two ways: it identifies a clear quick-win for the 100-day plan, and it usually means current gross margin is artificially low. We have seen Year 1 cloud optimization deliver 180-320 bps of gross margin improvement on software portcos without touching the product roadmap.

We stopped accepting management's stated cloud costs at face value after a 2023 deal where the seller had pre-purchased three years of reserved instances that all expired within nine months of close. Our run-rate cloud bill went from $4.1M to $9.8M. Now we model unit cloud economics — cost per tenant, cost per transaction — in every software diligence.
Operating Partner, mid-market PE software fund

The Modern Tech DD Tool Stack

A 2026-vintage tech DD looks nothing like the manager interviews and architecture deck reviews that defined the discipline a decade ago. Sponsors now expect diligence providers — whether internal operating teams, specialists like Crosslake, West Monroe, or Stepwise, or the tech arms of Bain, EY-Parthenon, and Alvarez & Marsal — to run an instrumented examination of the actual asset, not a paper review of its documentation.

Tools We Deploy in a Four-Week Tech DD

Read access to the production repository — even time-boxed and supervised — is now the single biggest predictor of diligence quality. In transactions where sellers grant read-only Git access via a clean room, we identify 3-5x more material findings than when we are limited to architecture diagrams and management interviews. Sophisticated buyers now make repository access a deal-breaker requirement in exclusivity, not a nice-to-have.

🔍AI-Assisted Code Review in Diligence
Since mid-2024, we have used LLM-based code analysis (Claude, GPT-4-class models with code-specific fine-tuning, and specialized tools like Greptile and Sourcegraph Cody) to accelerate codebase comprehension. What used to take a senior architect three weeks of reading now takes four to seven days. The AI does not replace judgment — it accelerates the discovery of the right files to read carefully. We've documented a 40-55% reduction in diligence hours on code review while expanding scope coverage.

Red Flags That Should Move Price or Kill Deals

Not every finding warrants action. Sponsors should expect tech debt in every target; the question is whether the debt is recoverable, the team is capable of recovering it, and the value creation plan accounts for it. The findings below, in our experience, warrant either a material price adjustment, a structured indemnity, or in some cases a pass.

Severity Tiers for Tech DD Findings
TierExample FindingsTypical Action
Tier 1 — Deal-ThreateningEmbedded GPL code in proprietary product; PII breach not disclosed; key person owns critical IP personally; no source control historyPass, or fundamental restructuring of deal
Tier 2 — Material RepricingRe-platforming required for growth case; >40% cloud cost increase needed; founder-CTO leaving; no DR or backup tested in 12 months3-8% EV adjustment, escrow, or earn-out structuring
Tier 3 — Value Creation PlanTest coverage <30%; manual deployments; tech debt at 12-18% TDR; documentation gaps100-day plan items, no price impact
Tier 4 — Operational HygieneOutdated dependencies; missing monitoring; informal incident processStandard post-close cleanup

Open-source license contamination remains an under-recognized Tier 1 risk. We diligenced a developer tools company in 2024 where 8% of the proprietary codebase had been copied from AGPL-licensed projects without compliance. The remediation required a six-month rewrite and significantly altered the bid. Modern SBOM tooling — Snyk, FOSSA, Black Duck — surfaces these issues in hours, but only if buyers insist on running the scans during diligence rather than after close.

The cheapest tech DD finding is the one you discover before signing. The most expensive is the one a customer discovers after.

From Diligence Findings to the Value Creation Plan

Tech DD that ends at signing is a missed opportunity. The same evidence base used to size risk during diligence should populate the 100-day plan and the longer-arc value creation roadmap. We hand off three artifacts to the post-close team: a remediation backlog with sized initiatives, a target architecture state with sequencing, and a set of operational KPIs the portfolio company will report monthly.

From DD to Value Creation
1
Weeks 1-4: Diligence Execution

Five-layer assessment with instrumented tooling. Output: IC memo with sized findings, repricing recommendations, and risk register.

2
Weeks 5-8: Pre-Close Planning

Translate findings into Day 1, Day 30, Day 100 initiatives. Identify interim CTO or fractional engineering leadership if key person risk identified.

3
Days 1-100 Post-Close

Quick wins on cloud cost (typically 15-25% reduction), security hygiene, and monitoring. Initiate any required re-platforming workstreams. See Article 5 for the full integration playbook.

4
Months 4-12

Execute against the architecture roadmap. Track DORA metrics monthly. Stand up engineering KPI dashboard for sponsor reporting.

5
Months 12-36

Strategic upgrades — typically platform migrations, AI/ML feature investment, M&A integration capability. Position for exit.

6
Exit Prep

Reverse the diligence — produce the data room artifacts a sophisticated buyer will demand. Covered in detail in Article 9.

Sponsors who run this loop well treat each portco diligence as input to a fund-level intelligence asset. The third software deal benefits from the data of the first two. Standardized tooling, repeated KPIs, and a shared engineering benchmark library mean later deals close faster and at higher confidence. Our larger PE clients now maintain proprietary benchmark databases covering 50+ portcos on metrics like cost per ARR dollar, deploy frequency, and cloud spend as a percentage of revenue — data the strategics cannot match.

🎯Build the Tech DD Capability In-House or Buy It?
Funds doing fewer than six software deals annually generally cannot justify a dedicated internal tech DD team and should rely on rotating specialist firms. Funds doing 10+ software deals annually should run a hybrid: internal operating partners who own the playbook and relationships, plus external specialists for execution capacity and codebase-specific expertise. The break-even, based on our work with eight mid-market software-focused funds, sits at approximately $400-600K of annual external diligence spend.

What This Means for the Operating Model

The funds we see winning competitive software processes in 2026 share three habits. First, they front-load tech DD — beginning the technical review in parallel with the IOI, not after exclusivity. Second, they invest in proprietary instrumentation: tooling agreements, benchmark databases, and reference relationships with sellers' former engineers that let them triangulate management claims quickly. Third, they integrate tech DD findings directly into the LBO model rather than carrying them as qualitative risks.

The connection to the rest of the operating model is direct. Tech DD findings shape the 100-day plan (Article 5), inform which capabilities belong in shared service centers, define the cybersecurity baseline that Article 11 tracks across the portfolio, and determine whether a target needs a fractional CTO in the first 90 days. The diligence is not separate from value creation. It is the first chapter of it.

💡Did You Know?
In 2025, the median tech DD engagement on a mid-market software buyout consumed 380-450 consultant hours across a four-week window — roughly double the 2019 baseline. The increase is almost entirely in code-level and cloud-level work; management interview hours have actually declined as tooling makes evidence-based assessment faster than narrative-based assessment.

Frequently Asked Questions

How long should a tech due diligence take for a mid-market software buyout?

Four to six weeks is standard for a $100M-$1B enterprise value software target, assuming repository access is granted by week two. Compressed timelines of two to three weeks are possible when the diligence team has prior familiarity with the codebase technologies and when management cooperation is high, but compression usually means reduced coverage of the engineering org and infrastructure layers.

What is a healthy technical debt ratio for a SaaS company?

Under 5% indicates excellent code health on the SonarQube and CAST scales. The 5-12% range is normal for mature production codebases. Above 20% indicates that maintenance is consuming a majority of engineering capacity and the company will struggle to ship new features at competitive velocity. Above 30% typically signals that re-platforming components is more cost-effective than continued maintenance.

Can tech DD findings reverse a deal that's already in exclusivity?

Yes, and they do — in roughly 6-8% of software diligences in our experience, findings are material enough to either kill the deal or force a renegotiation that ends in walkaway. The most common deal-breakers are undisclosed security incidents, open-source license violations, and key person risk where the founding engineers are not retained. Sponsors who do not conduct rigorous tech DD often discover these issues 12-18 months post-close, when remediation is far more expensive.

Should tech DD assess AI and machine learning capabilities differently?

Yes. Standard code analysis does not capture model performance, training data quality, MLOps maturity, or dependency on third-party foundation models. For targets with AI as a core product feature, we add a parallel ML DD workstream covering data lineage, model evaluation against held-out test sets, GPU cost economics, and contractual exposure to OpenAI, Anthropic, or other model providers. This typically adds one to two weeks and 80-120 hours to the engagement.

How do we handle tech DD when the seller refuses repository access?

Treat it as a significant risk flag and price accordingly. We typically apply a 2-5% additional valuation discount when code-level access is denied, because the residual unknowns become the buyer's liability. Acceptable alternatives include supervised on-site code review with a senior architect, third-party code audit reports from firms the seller commissions, and structured representations and warranties insurance that covers undiscovered code-quality issues. None of these fully substitute for direct access.