9 min read · 1,924 words
A martech stack audit fails when it scores tool features instead of the organisational decisions those tools were bought to settle. Gartner has tracked utilisation of the average martech stack falling for three straight survey cycles, and the fix is not better procurement. It is asking what each tool is actually for before you ask what it does.
I inherited a stack I had not bought once, in a role change at the end of last year, moving into a brand and communication mandate at a large industrial company. The login list that landed in my inbox in the first week ran well past a dozen platforms, several with renewal dates already close, owners no longer in the building, and no shared memory of why half had been signed. Roughly a third looked, on a first pass, like they were doing the same job under a different logo. That is not a procurement failure. It is what happens when tools get bought to solve a people problem or a budget-cycle problem, and nobody goes back to check whether the technical problem got solved along the way, a pattern I have also seen play out in marketing functions inside product-led, engineering-heavy companies.
Why does martech utilization keep falling?
The number to start with is Gartner’s. Marketers used 58% of their martech stack’s capability in 2020. That fell to 42% in 2022, and to 33% in 2023, which Gartner’s own survey release called a second consecutive year of decline. That same survey found martech absorbing 25.4% of the marketing budget in 2023, meaning the average team spent a quarter of its budget on tools it used one-third of. Nobody designs a function to run at those numbers on purpose. It accumulates.
The accumulation has not reversed since. Martech’s share of the marketing budget slid further to 23.8% in 2024, its lowest point in a decade, according to MarTech’s read of the Gartner CMO Spend Survey data. That looks like restraint until you notice what it is really measuring: CMOs losing budget authority to IT and finance, not marketers suddenly using their tools better, a shift I have written about in the context of what a B2B company should actually spend on brand. Utilisation is a separate line, and nothing in the newer surveys suggests it turned a corner.
How big has the stack actually gotten?
Scott Brinker’s landscape, the closest thing this industry has to a census, counted 15,384 martech products in the 2025 edition, a 9% rise on the 14,106 counted the year before, per MarTech’s coverage of the count. Growth has slowed from the boom years, but each new tool still adds a fresh decision point on top of a stack already outgrowing its use.
The waste this produces has a number too. Zylo’s 2026 SaaS Management Index puts 53% of enterprise SaaS licenses as either unused or used too little to justify the line item, and martech is one of the categories driving it, because campaign tools and single-channel point solutions get bought under deadline pressure and never get un-bought. India’s stack is also growing before it gets audited: Grand View Research sizes the domestic martech market at roughly $30.1 billion for 2025, growing at a 25.5% compound rate through 2033, per its India martech market outlook.
The audit that fails, and the one that works
Most martech audits ask the wrong question. They ask what a tool does, list its features against a shortlist of competitors, and score it on capability. That audit almost never produces a cut list, because every tool in a mature stack does something. The question that surfaces redundancy is different: what organisational decision does this tool exist to settle, and is that decision still being made anywhere near this tool.
Here is the claim a fair number of peers will push back on. The utilisation problem is not primarily a buying-discipline problem. It is what happens when martech gets purchased to resolve a political question rather than a technical one: a regional team that wanted its own analytics because it did not trust the central dashboard, a function that bought a listening tool after losing an argument about budget visibility, a campaign owner who needed a platform with their name on the license to justify headcount at the next review. None of those tools were bought to close a capability gap. They were bought to close a trust gap, and no feature audit will find that in the license terms.
The four-question test for every tool in the stack
Score every tool against four questions rather than a features sheet. A tool that fails two or more is a live candidate for the cut list, regardless of how capable it is on paper.
| Question | What a fail looks like |
|---|---|
| 1. What specific decision does this tool exist to inform or execute | No one can name the decision without describing a feature instead |
| 2. Who used it in the last 90 days, by name | Login logs show one or two users, or the seat count exceeds active users by more than half |
| 3. Does another tool in the stack already answer question 1 | Two or more platforms produce the same report or execute the same workflow |
| 4. What breaks in the business, not the marketing function, if this is switched off for 60 days | The honest answer is an argument, not a process |
Question four is the one that separates a real audit from a features exercise. A tool can fail every capability comparison and still be indispensable because a regional sales head reads its dashboard every Monday. A tool can win every capability comparison and still be dead weight because the one person who used it left the company eleven months ago.
The cut list: four categories, not one
Once the scoring is done, the tools that fail sort into categories that need different treatment, not one blanket removal.
- Orphaned tools: the owner left, the licence auto-renewed, nobody has logged in this quarter. Cut immediately.
- Duplicate capability: two tools answering question one identically. Keep the one with higher adoption and lower switching cost, sunset the other over one renewal cycle.
- Feature-shelf tools: the organisation bought the platform for one module and pays for the whole suite. Renegotiate down, or replace with a point solution.
- Political tools: fails on capability but exists to answer a trust or visibility problem for a specific stakeholder. The category most audits mishandle, cutting it and reopening the original fight, or leaving it and losing credibility for the whole exercise.
What if a tool fails the test but cannot be cut?
This is where consolidation efforts actually die, not in the scoring. A tool that exists to give a regional head or a founder’s office its own view of performance cannot simply be switched off, because the trust problem that created it is still there. Cutting it without solving that problem just moves the demand to the next procurement cycle under a different vendor name.
The workable move is to keep the underlying data feed and retire the redundant application sitting on top of it. Give the stakeholder a real seat in the central tool with the specific view they were trying to build for themselves, and put a 12-month sunset clause on the standalone platform tied to that access being delivered. That turns a political fight into a project with a deadline, which is a fight most CFOs will let you win.
Why AI bolt-ons make consolidation more urgent, not less
The 2026 layer on top of this is AI features added to tools already in the stack, often duplicating a capability another platform already has. Content generation is now a checkbox feature inside the CRM, the email platform, the social scheduler and the analytics suite at once, and almost nobody is asking whether the fifth AI writing assistant solves a different problem than the first four, a version of the confusion I saw take hold once AI features first arrived inside brand management tools.
The evidence that this bolt-on approach is not paying off is direct. Gartner surveyed 413 martech leaders between June and August 2025 and found 45% saying vendor-offered AI agents already in their stack failed to meet the business performance promised. The 2026 CMO Spend Survey found CMOs now allocating 15.3% of budget to AI, while only 30% call their organisation ready to scale it, and 70% say their processes are not mature enough to make the investment work. That is the utilisation gap repeating itself, on newer technology, inside the same unaudited stack.
Adding an agent layer on top of an unaudited stack does not fix the problem, it adds a fifth unanswered question to a stack that already had four. The case for consolidating before adopting more AI tooling is not caution for its own sake. An agent asked to orchestrate five overlapping systems will faithfully automate the overlap along with everything else.
What this means for your next audit
Run the four-question test against the whole stack once a year, not tool by tool as renewals come up, because the duplication only shows up when you can see all the answers to question three side by side. Budget defence gets easier when the audit produces a document, not a memory, so keep the scoring table and update it at each renewal. And treat the political category honestly rather than pretending every cut is a technical decision, because the ones you mishandle come back with a new vendor name attached within a year.
A stack audit that only counts features will always come back clean, because every tool does something. Score the decisions instead, and the redundant third of the stack finds itself.
Frequently asked questions
What is a martech stack audit? A structured review of every tool in a marketing technology stack, scored not on features but on the specific organisational decision each tool exists to inform, who actually uses it, whether another tool already covers the same decision, and what genuinely breaks if it is switched off.
How much of a typical martech stack goes unused? Gartner’s 2023 survey found marketers using just 33% of their stack’s capability, down from 42% in 2022 and 58% in 2020. Zylo’s 2026 SaaS Management Index separately found 53% of enterprise SaaS licenses sitting unused or underused across categories including martech.
Why does martech utilization keep falling instead of improving? Because tools keep getting bought to resolve organisational trust or visibility problems rather than technical capability gaps, and no feature-based audit ever catches that at renewal time. The landscape also keeps growing, past 15,384 tracked tools in 2025, adding new candidates faster than old ones get retired.
Should we cut a martech tool that a senior stakeholder insists on keeping? Not by force. Give that stakeholder the specific view or access they built the tool for inside your central platform instead, then attach a firm sunset date to the standalone tool tied to that access being delivered. Cutting it outright usually just moves the same demand to a new vendor next year.
Do AI features inside existing martech tools reduce the need for a stack audit? No, they increase it. Gartner found 45% of martech leaders saying vendor AI agents already in their stack underperformed what was promised, and most CMOs allocating budget to AI in 2026 say their organisation is not yet mature enough to use it well. Layering agents onto an unaudited, duplicated stack automates the duplication rather than removing it.
The next time a vendor renewal lands on your desk, ask what decision that tool exists to settle before you ask what it does. What is still sitting in your stack because nobody has asked that question yet?
