The CTMO Quarterly Volume 1 · Issue 3 · Q3 2026

The
Compounding
Stack

Most companies have a technology stack. Very few have one that compounds. The difference is not which tools you bought. It is whether the system was designed to get better by itself.

John Kirker, CTMO July 1, 2026 ctmo.com
Read
In This Issue
Section 01 Editor's Letter

Most Stacks Depreciate. A Few Compound.

JK
John Kirker
CTMO · ctmo.com

Most companies at your stage have a technology stack. Some have a good one. Very few have one that compounds.

The difference is not about which tools you bought. It is about what happens to the value those tools produce over time. In a typical stack, each tool does its job: your CRM stores contacts, your marketing automation sends emails, your analytics platform generates reports. Each tool depreciates. It gets older. It requires maintenance. It occasionally gets replaced by something newer. The value it produces is roughly constant, minus the cost of keeping it current.

A compounding stack behaves differently. Each component makes the other components more valuable over time. The data your CRM collects makes your marketing automation smarter. The patterns your marketing automation surfaces make your analytics platform more precise. The precision in your analytics makes every dollar of future spend more efficient. The system improves without anyone adding a new tool or increasing the budget.

"What separates a depreciating stack from a compounding one is not budget or talent. It is architectural intent: whether the system was designed to get better by itself."
John Kirker · The CTMO Quarterly · Q3 2026

I have built compounding stacks twice in my career. Once for a $17 million home improvement company that grew to close to $100 million over 14 years. Once for a $2 billion pest control company where a single system, PinID, generated more than $500 million in trackable revenue over a decade. In both cases, the stack was not a collection of tools. It was an architecture that treated every customer interaction as an input that made the next interaction better.

By the end of this issue, you will have a framework for evaluating whether your technology infrastructure compounds or depreciates. You will also know the three architectural conditions that separate the two, and what it takes to shift from one to the other.

The question you should be sitting with as you read: If your technology stack is a depreciating asset that requires increasing effort to maintain its current performance, what does that mean about every dollar you plan to spend on it next year?

Section 02 Featured Analysis

The Compounding Stack: What It Looks Like, and How to Know If Yours Does

The Belief Worth Challenging

The belief running through most $10M to $100M companies right now sounds like this: "We need to invest in our technology stack to stay competitive."

It is repeated in board decks, strategic plans, and budget requests. It is the consensus view. And it is hiding a question that almost no one asks: what kind of asset are you building when you invest? There are two kinds of technology infrastructure. Most companies are building the first kind and calling it the second.

The reframe: A depreciating stack does what it was configured to do on the day it was deployed, and requires rising spend to hold that line. A compounding stack was architected to treat every interaction as an input that improves the system's future performance. One is economically a fleet of trucks. The other is an asset that appreciates.
Two Kinds of Stack: The Divergence Shows Up at Year Three
Kind One
Depreciating Stack
Data flowOne direction, into dashboards
Value over timeFlat, minus the cost of upkeep
Cost of ownershipRises 15–25% a year after year two
HandoffsEach function re-learns the customer
AI, deployed hereFaster reports on the same flat performance
Diverges
Kind Two
Compounding Stack
Data flowClosed loop, back into production
Value over timeMarginal value rises faster than cost
Cost of ownershipMay rise, but output rises faster
HandoffsInformation accumulates across functions
AI, deployed hereCompounds on an already-compounding system

A depreciating stack looks identical to a compounding one in year one. The divergence becomes visible at year three or four, and by then the compounding stack has a head start that budget cannot close.

The Three Architectural Properties

A compounding stack has three properties. If any one is missing, the stack depreciates. If all three are present and connected, it compounds. There is no partial credit.

01
Closed-Loop Data Flow
The outcome of every interaction feeds back into the system that generated it, in real time, and changes the next interaction. The loop is operational, not a report.
Test: Did the last 90 days of campaign data automatically retarget the next campaign, with no human touching the settings? If not, the system learned nothing from the money you spent.
02
Shared State Across Functions
One continuously updated representation of the customer that every function reads from and writes to. Information accumulates instead of being re-established at every handoff.
Test: When a prospect fills out a form, how many times is their intent re-established before they buy? Five or more is a leak. Compounding stacks have one or two.
03
Increasing Returns on Data
Historical data gets more valuable over time because the system mines it for patterns that were invisible at collection. This is where AI accelerates a stack that already has the architecture.
Test: Is data from 12+ months ago an active input into a system that adjusts today, not just context in a slide? If not, your data depreciates alongside your stack.
Where AI actually fits: AI does not create a compounding stack. It accelerates one that already has the right architecture. Deployed on a depreciating stack, AI produces faster reports about the same flat performance. That is a more expensive way to watch value depreciate.

Three Stacks, Three Outcomes

The following are composite examples drawn from engagements, market observation, and documented outcomes. Names and specifics are masked. The patterns are real.

Composite A The Tool Collector

A $42M B2B SaaS company spent roughly $1.2 million a year on twelve tools: CRM, marketing automation, ABM, intent data, sales engagement, a new AI content generator. Each did exactly what it was purchased to do.

Three years in, spend had risen to $1.8 million. Lead volume was flat. Conversion had not moved. The marketing team spent about 30% of its time exporting, cleaning, and re-importing data between tools. They called it "operating the stack." It was maintaining the depreciation. No tool made any other tool smarter, so nothing compounded.

What it cost them: $1.8M a year for the marketing performance they had at $1.2M, plus 30% of a team's capacity consumed by data plumbing.
Composite B The Integration Project

A $28M professional services firm recognized the problem and commissioned a $400,000, six-month integration to unify CRM, marketing automation, and analytics. It worked. Conversion rose about 15%. Manual data work dropped.

But the integration was static. It moved the same data between the same three systems in the same patterns, day after day. When the firm bought an AI lead-scoring platform eight months later, it required a new integration: $180,000, four months. The marginal cost of adding capability never fell. That is the definition of a depreciating system.

What it cost them: $580,000 over 18 months for a 15% lift that did not itself compound. The integration was a one-time act, not a continuous one.
Composite C The Compounding Architecture

The home services engagement referenced in last quarter's issue began from a worse position than either case above: $17 million in revenue, $160,000 a month in advertising, zero tracking, no closed loop of any kind. The difference was architectural intent from day one.

Every system was designed with all three properties in mind. Call tracking fed conversion data back into ad targeting. The website created a shared customer profile that the call center, sales, and marketing all read from and wrote to. Campaign performance went into the system that allocated the next dollar of spend, not into a dashboard. Ad spend dropped from $160,000 to about $60,000 a month while lead volume rose 160%, and the improvement compounded every month as the loop processed more data.

The outcome: Over 14 years, $17M grew to close to $100M. The infrastructure was not a cost center that supported the business. It was a marketing asset that compounded.

What Building One Actually Requires

The difference between a depreciating stack and a compounding one is not determined by what you buy. It is determined by how you connect what you buy, and whether the connections are designed to learn. Three things most purchasing processes never evaluate:

1. Architecture before tools. Before a tool is evaluated, ask how it will contribute to the loop. If the answer is "it stands alone," it is a depreciating purchase, however useful.

2. The loop as the unit of investment. Budget for loops, not tools. "Closed-loop conversion optimization, expected to cut cost-per-acquisition 8–12% a year as the loop matures" produces a business case that appreciates. "Marketing automation, $120K/year" produces a cost center.

3. A function that owns the architecture. Compounding loops cross functional boundaries. If no one owns the loop, the stack depreciates by default. This is the Translation Layer function from Q2, pointed at infrastructure.

The Terminix PinID system is the clearest example. Direct mail campaigns carried unique PIN codes that triggered personalized web experiences and real-time lead routing. Every interaction improved the next campaign's targeting, the next web experience, and the next lead's routing. It ran more than a decade and generated an estimated $500 million in trackable revenue, not because someone kept optimizing it, but because the architecture optimized itself. The compounding was structural, not operational. That system was built with technology that is primitive by today's standards. What has not changed is the requirement: someone has to design for compounding from the beginning. The tools cannot do that part.

The Compounding Question

Most CEOs who read this will recognize their stack in one of the three composites, and most will recognize it in the first two. That recognition is a starting point, not a problem. The decisions that produce a compounding stack can be made at any point. But every year you delay, the competitor with the compounding stack pulls further ahead, and the gap is not linear.

"For every dollar we spent on marketing technology last year, how much more effective is our system today than it was 12 months ago, without any additional investment?"
The question to take into your next leadership meeting · Q3 2026

If the answer is "about the same," your stack is depreciating. If the answer is "we do not know," that is worse, because it means you are spending without measuring what kind of asset you are building. The companies that will dominate their markets over the next five years are not the ones spending the most on technology. They are the ones whose technology spending compounds.

Section 03 The Panel Debate

Architecture or Execution?

Six voices. One compounding question. Real disagreement about what decides whether a stack appreciates or depreciates.

EV
Dr. Eleanor Vance
Systems Architect
"The compounding stack is not aspirational. It is table stakes for any company that intends to be competitive at scale in five years."
MO
Marcus Okafor
Revenue Operator
"Eleanor is right about the architecture. She is wrong about what kills most companies' stacks. It is execution decay, one shortcut at a time."
NC
Nadia Chen
AI Strategist
"ML on a depreciating stack does not compound. It produces very precise wrong answers, delivered quickly."
WS
William Stover
Devil's Advocate
"I question whether the compounding stack is actually attainable for most companies at this stage. The examples are biography, not a playbook."
LP
Linda Park
Historian
"Every compounding infrastructure in business history started as an inaccessible advantage that became standard practice. Choose your cost."
JK
John Kirker
Practitioner
"The bottleneck is not the technical skill to design a compounding stack. It is the organizational authority to protect one."
EV
Dr. Eleanor Vance
The Systems Architect
The compounding stack is not aspirational. It is table stakes for any company that intends to be competitive at scale in five years. State should be shared, loops should be closed, and every component should make the others more valuable over time. That is how well-designed systems have worked since before the term "tech stack" existed. What is revolutionary is that most companies still do not do it, and the reason is not budget or talent. It is the purchasing model. Companies buy tools the way consumers buy appliances: one at a time, with no regard for how the new purchase interacts with everything already in the kitchen. Twelve tools that do not talk to each other is not a stack. It is a collection, and collections do not compound. They accumulate costs. One caution on the featured analysis: the claim that these decisions can be made at any point is technically true and practically misleading. Retrofitting closed loops onto a stack designed with open ones is expensive and rarely compounds as well as designing for it from the start. The best time was when you built the stack. The second best time is today, but do not pretend the second option costs the same as the first.
MO
Marcus Okafor
The Execution Rebuttal
Eleanor is right about the architecture. She is wrong about what kills most companies' stacks. In every $50M-plus company I have run revenue operations for, the architecture was designed with good intentions. Closed loops were in the requirements. Shared state was in the original design. Then the quarter happened. Marketing needed a campaign in two weeks, the integration that would have closed the loop was four months deep in the backlog, so the campaign launched without it, the data went into a spreadsheet, and the optimization happened in someone's head. Six months and three campaigns later, the team had built a manual workflow around the gap and nobody fought for the integration anymore. This is how compounding stacks degrade into depreciating ones: not through bad architecture, but through execution decay, one shortcut at a time. My prescription is to fix the operating model. Every quarter, audit the loops. Are the closed ones still closed? Are new manual workarounds replacing what the system should do automatically? That audit is worth more than any redesign, because it catches the decay before it becomes structural.
NC
Nadia Chen
The AI Argument
AI does not just accelerate a compounding stack. It changes what compounding means. Before machine learning was accessible at this stage, compounding meant rules-based optimization: human-authored logic the system executed, bounded by the quality of the rules. With ML models on top of a closed-loop architecture, the compounding becomes unbounded. The system discovers patterns no human would have authored, surfacing segments and pathways that emerge from volume, not intuition. That is the "increasing returns" property, and it is the most important of the three. But here is the trap nobody in the AI vendor space is honest about: ML models deployed on depreciating stacks do not compound. They produce the same flat performance faster. Point a model at broken, one-directional data and it will find the patterns in your broken data and optimize toward them. You will get very precise wrong answers, delivered quickly. The precondition for AI to compound is not AI capability. It is data architecture. Get the loops, the shared state, and the increasing returns right first. Skip the foundation and the AI spend just depreciates alongside everything else.
WS
William Stover
The Devil's Advocate
I want to do something uncomfortable: question whether the compounding stack is actually attainable for most companies at the $10M to $100M stage. The featured analysis cites two examples. One took 14 years. The other ran for over a decade. Both had a single practitioner with rare expertise operating at the architectural level across marketing and technology for years. That is not a playbook. It is a biographical fact about one person's career. What percentage of companies at this stage even have access to someone who can diagnose whether their stack compounds, much less redesign it? My estimate is under five percent, and the ones who do probably already have a compounding stack. For the other 95%, the realistic options are an expensive consultant who may lack genuine architectural expertise, a CTO who optimizes for stability, or a CMO who optimizes for campaigns. I am not disputing the theory. I am disputing the implied accessibility. Before you commission an overhaul, answer this: do you have someone who can credibly evaluate whether your current architecture compounds? Not someone with an opinion. Someone who has built one. If not, that is your first investment. Everything else is premature.
LP
Linda Park
The Historical Pattern
Stover raises a fair accessibility objection, but history reframes it. Every compounding infrastructure in business history started as an inaccessible advantage that became standard practice. Double-entry bookkeeping, invented in the 15th century, was a compounding architecture for financial management, and for almost a century it was available only to the most sophisticated trading firms. Then it became standard, not because it got simpler, but because the firms that adopted it first outperformed so visibly that everyone else had to follow. The same pattern played out with relational databases in the 1970s and 80s, ERP in the 90s, and cloud in the 2010s. Each was initially accessible only to companies with rare expertise, and each became standard within 10 to 15 years. We are in year one or two of the compounding-stack transition. The expertise gap Stover describes is real today and it will close, not because the problem gets easier but because the cost of ignoring it becomes unbearable. The cost of early adoption is expertise. The cost of late adoption is catching up to competitors whose systems have been compounding for years while yours depreciated. Choose your cost.
JK
John Kirker
The Practitioner Close
Stover's objection is one I hear from CEOs constantly, and he is half right. He is right that the expertise is rare, that most CTOs will not build a compounding architecture by default because their optimization function is stability, and that most CMOs will not specify one because their horizon is the campaign. Where he is wrong is the implication that it requires heroic individual capability. It requires intent. The WinDor engagement did not succeed because I am uniquely talented. It succeeded because someone decided, at the beginning, that the infrastructure would be designed to improve its own performance over time, and then held that standard through every purchasing decision and every proposed shortcut. Marcus is right that execution decay is the real killer. What protected the architecture was not my expertise alone. It was that someone had the authority and the accountability to say no when a shortcut would break the loop. That is what most companies are missing: not the technical skill to design a compounding stack, but the organizational authority to protect one. Stover says find someone who has built one. I say find someone willing to protect one. The building is the easier part. The protecting is where most stacks die.

Where the panel landed: Eleanor argued the problem is purchasing architecture; Marcus, execution discipline; Nadia, that AI raises the ceiling but requires the foundation first; Stover, that accessibility is the real barrier; Linda placed it in historical context and called the transition inevitable; John argued the bottleneck is organizational willingness to protect the loop. They agreed on one thing: most companies at $10M to $100M are building depreciating stacks and do not know it.

Section 04 The Framework

The Compounding Stack Diagnostic: A Test You Can Run Today

The diagnostic evaluates whether your technology infrastructure gets better over time without proportional increases in investment, or whether it requires rising effort and budget to hold its current performance. Your stack compounds only if all three properties are present. Missing any one converts compounding into depreciation.

The Three-Property Test

Property 1 · Closed-Loop Data Flow
Pick any campaign from the last 90 days. Did its performance data automatically change how the next campaign was targeted, messaged, or distributed, without a human adjusting the settings?
YesClosed loop. The system learned from the spend.
NoOpen loop. A human read a report and adjusted the next campaign. The system did not learn.
Property 2 · Shared State Across Functions
When a prospect fills out a form today, how many times will their intent, qualification, and context be re-established by different teams before they become a customer?
1 to 2 timesShared state. Information accumulates across functions.
3 or moreFragmented state. Each function starts fresh. Every handoff loses data.
Property 3 · Increasing Returns on Historical Data
Is data from 12 or more months ago an active input into systems that adjust their behavior based on what it reveals, not just context in a presentation?
YesHistorical data is appreciating. Patterns surface over time.
NoHistorical data is depreciating. It lives in reports nobody opens.

Scoring

3/3
Your stack compounds. Protect the loops. Audit quarterly for execution decay.
2/3
Compounding potential, one structural leak. Identify and close the missing property.
1/3
Depreciating with one bright spot. Budget for architecture, not tools.
0/3
A cost center. Every dollar spent maintains current performance or worse.

The Four Questions for Your Next Tool Purchase

1. What data does this tool produce? Not its reports. The raw data it generates or collects.

2. Where does that data flow after production? Into another system that acts on it, or into a dashboard?

3. What other system's performance improves as a result? Name it specifically. If you cannot, it is a standalone purchase.

4. Does this tool close a loop, or open a new one? Closing a loop compounds. Opening one is a maintenance commitment.

The 90-Day Architecture Audit

Run this once a quarter. It takes less than an hour.

1. List every integration between marketing and technology systems.
2. For each: is data flowing in both directions, or only one?
3. For each one-way flow: what would it take to close the loop?
4. For each closed loop: is it still functioning, or has a manual workaround replaced it?
5. Count the manual workarounds introduced since the last audit. That count is your decay rate. If it is rising, your compounding architecture is degrading into a depreciating one, regardless of what the original design intended.

Apply it in ten minutes: run the Three-Property Test, write down the score, then pull up your next 12 months of technology budget and ask how much of it closes a loop versus maintains a tool that operates independently. The ratio tells you whether you are investing in a compounding asset or funding a depreciating one.

Section 05 The Provocation

The Case Against Designing for Compounding

WS
William Stover
Steel-Manned Counterargument · Written in Good Faith

I accept the framework. A compounding stack outperforms a depreciating one over time. Closed loops beat open ones. Shared state beats fragmented state. Historical data that teaches the system beats data that sits in a report. My objection is not to the destination. It is to the journey.

The featured analysis implies the correct response is an architectural intervention: redesign how your tools connect, close the loops, build for compounding. That sounds right. For most companies at this stage, it is the wrong first move. The ones I have watched attempt large-scale stack redesigns share a pattern. Year one: ambitious plan, big budget, executive attention. Year two: 60% complete, the original architect has left, and the business has changed enough that some of the integration designs no longer match operations. Year three: a partially compounding stack that requires constant attention from a senior resource who could be doing something else.

The total cost of the architecture project exceeds the value of the compounding it produced over a three-year window. The investment was not wrong. It was premature.

"Take the Three-Property Test. Identify the one loop that, if closed, would produce the highest marginal return. Close that loop. Run it for 90 days. Prove the organization can protect it. Then build the stack."
William Stover · The Provocation · Q3 2026

Compounding through discipline is slower than compounding through architecture. It is less elegant. It requires patience most CEOs do not have. But it has one advantage the top-down redesign does not: it tests your organization's ability to sustain a compounding loop before you commit to building twelve of them. The featured analysis is right that compounding stacks win. I am arguing that building one loop at a time, proving you can sustain it, and expanding from demonstrated success is safer than assuming execution discipline your organization has never demonstrated at scale.

Section 06 The Evidence Locker

Every Number Has a Basis

Every claim in this issue has a labeled basis. Composites are identified as composites. Confidence levels are assigned to any figure that involves practitioner analysis or extrapolation from limited data.

2008–22
Primary · John Kirker
Home services engagement: $17M to ~$100M over 14 years. All three compounding properties architectural from day one.
Starting conditions: $17M/year revenue, $160K/month advertising, zero tracking, no closed loops. Architecture deployed: closed-loop attribution tying every ad dollar to every closed deal; shared state across marketing, sales, and call-center operations; historical data fed back into targeting. Compounding evidence: ad spend cut to ~$60K/month, lead volume up 160%, conversion and average deal size up, improvements accumulating as the loop processed more data. Outcome: company acquired by a major national manufacturer.
5/5 · Primary source, direct engagement documentation
2000–12+
Primary · John Kirker
Terminix PinID system: an estimated $500M+ in trackable revenue over a decade. Compounding built into the infrastructure, not dependent on a person.
Direct mail carrying unique PIN codes triggered personalized web experiences and real-time lead routing. Each interaction improved the next campaign's targeting (closed loop), was visible to marketing and call-center operations (shared state), and accumulated value as the system learned which segments and routes converted best (increasing returns). The system kept generating revenue after active consulting involvement ended, which is the signature of architectural rather than operational compounding. Deployed across 200+ enterprise customers; Terminix ran for more than a decade.
4/5 · Primary source; $500M+ is a practitioner estimate based on campaign volume, duration, and revenue per transaction; not independently audited
Pattern
Pattern · IT Asset Literature
Absent deliberate architectural intervention, technology systems depreciate. Peak value lands in the first 18 months; TCO rises 15–25% a year after year two while marginal value stays flat.
A frequently cited benchmark from IT asset management literature and multiple analyst reports. The pattern holds across company sizes but hits hardest at the $10M to $100M stage, where budgets are constrained and each tool is a larger share of total infrastructure investment. Specific percentages vary by source and context.
3/5 · Widely cited pattern; specific figures vary by source
2025–26
Observed · Editorial
The gap between compounding and depreciating stacks appears to be widening faster in 2025–26 than in any prior two-year period, with AI adoption as the accelerant.
Companies with closed-loop architectures that deploy ML models on accumulated data see improvements that compound on top of already-compounding systems. Companies with open-loop architectures deploying the same models on fragmented data see marginal or no improvement. The practical implication: the advantage window for building a compounding stack is narrowing, because ML makes the gap exponential rather than linear. This is an editorial inference, not independently measured data.
2/5 · Directional observation; no systematic measurement at this company-size segment
Section 07 · The Trailing Question

If you stopped investing in your technology stack entirely for 12 months, maintaining what exists but adding nothing new, would your marketing performance improve, stay flat, or decline?

If the answer is "improve," your stack compounds: it is still learning from its own data, still closing loops, still surfacing patterns. If the answer is "stay flat" or "decline," your stack depreciates and requires continuous investment just to avoid regression. That answer tells you what kind of asset you have been building, and whether next year's technology budget is an investment or an expense.

Forward this to the person on your team who approves technology purchases. That conversation is the point.