There is a moment in every growing B2B company when martech stops being a tooling conversation and becomes a capital allocation conversation.
It rarely announces itself. It arrives quietly, in the form of a finance business partner asking why the marketing technology line has grown faster than pipeline, or a security questionnaire from a prospect that nobody can answer without three internal emails, or a data protection officer asking which of your platforms currently holds behavioural data on EU citizens and under what lawful basis.
That is the moment the stack becomes a portfolio. And portfolios are managed differently from tools. Tools get bought on features. Portfolios get managed on allocation, concentration risk, return, and exit discipline.
Most B2B organisations are still buying tools while operating a portfolio. That gap is where the value leaks.
Part 1: Why Stacks Drift Out of Alignment
Stack drift is not a failure of judgement. It is the predictable output of three entirely rational behaviours compounding over time.
Buying is additive, never subtractive. Every quarter brings a new gap and a new tool to close it. Almost no quarter brings a formal retirement. The stack only ever grows because nobody is incentivised to shrink it. A marketer who lands a new capability gets credit. A marketer who removes four redundant platforms gets a quieter kind of credit, if any.
Ownership fragments as teams specialise. Early on, one or two people know the whole stack. At scale, demand gen owns the ad platforms, content owns the CMS and SEO suite, sales ops owns the CRM and enrichment layer, and someone in the middle owns the automation platform that connects them. Nobody owns the whole. Overlap becomes invisible because no single person can see it.
Renewals run on autopilot. Auto-renew clauses, multi-year lock-ins, and staggered contract dates mean most tools are re-bought without ever being re-evaluated. The default is continuation, and defaults are powerful.
Add growth to those three forces and the arithmetic turns uncomfortable. Headcount grows, so seat counts grow. Volume grows, so usage-based pricing tiers escalate. Geography expands, so compliance surface area expands with it. The stack does not become misaligned because anyone made a bad call. It becomes misaligned because a hundred reasonable calls were never reconciled against each other.
The signals worth reading
Drift is easy to detect once you know what to look for. The signals below tend to appear in this order.
| Signal you observe | What it actually means |
|---|---|
| Two teams present the same metric with different numbers | Your source of truth is contested, not defined |
| A tool renews and nobody can name its internal owner | The capability is orphaned, and orphaned tools never improve |
| Onboarding a new marketer takes more than a week of tool training | Stack complexity has become a hiring tax |
| Reporting requires manual export and reconciliation | Your integration layer is failing, and analysts are the workaround |
| Security questionnaires trigger an internal scramble | Governance is reactive, which means it is already a risk |
| The same data is enriched by two different vendors | You are paying twice for one asset, and trusting neither |
None of these is fatal on its own. Three or more together mean the portfolio needs active management, not another point solution.
Part 2: The Three Costs That Never Appear in the Budget Line
The subscription fee is the visible cost, and it is usually the smallest of the three.
The adoption gap. Licences are bought for the team you plan to have, and used by the team you actually have. Seats sit dormant. Modules bought as part of a bundle are never switched on. Advanced features that justified the premium tier are used by two people, occasionally. Every dormant seat is a direct transfer from your budget to a vendor's renewal forecast, and it is entirely invisible unless you go looking for it.
Integration debt. Every tool added to a stack of n tools creates potential connection points with all of them. Complexity does not grow linearly, it compounds. That debt is paid in engineering hours, in middleware subscriptions, in brittle automations that break silently, and in the analyst time spent reconciling systems that were supposed to talk to each other. Integration debt is the single most underestimated cost in martech, because it never arrives as an invoice.
Compliance exposure. Every platform holding personal data is a node in your regulatory footprint. Every sub-processor those platforms use extends that footprint further, usually without your explicit knowledge. A stack of forty tools is not forty vendor relationships. It is forty vendor relationships plus their sub-processor chains, each one a place where consent lineage can break, where data residency assumptions can quietly fail, and where an AI feature enabled by default may be doing something with your customer data that your privacy notice does not describe.
The first cost is inefficiency. The second is friction. The third is liability. Only the third can end up in front of a regulator or a board.
Part 3: Governance Is No Longer a Back-Office Concern
Three shifts have moved martech governance from an operational detail to a leadership agenda item.
Regulation has become plural and specific. GDPR set the template, and a widening set of national frameworks including India's Digital Personal Data Protection Act, evolving US state privacy laws, and sector-specific rules have layered on top. The practical consequence is that "we are compliant" is no longer a single answer. It is a different answer per jurisdiction, per data category, and per processing purpose.
Consent lineage now has to survive the whole journey. It is not enough to capture consent at the form. You have to be able to demonstrate that the consent captured in your CMS travelled intact through your automation platform, your CRM, your enrichment vendor, your ad platforms, and your analytics layer, and that it was honoured at every stop including withdrawal. Most stacks cannot demonstrate this end to end. They can demonstrate it in pieces.
AI features arrived faster than AI policy. A significant share of martech platforms have shipped generative and predictive features into existing contracts, sometimes enabled by default. The questions of what data those features access, where inference happens, whether outputs are retained, and whether your data contributes to model improvement are now standard in enterprise procurement. If you cannot answer them for every platform in your stack, you have an exposure you have not priced.
Governance, framed correctly, is not a brake on growth. It is what allows you to sell into regulated industries, close enterprise deals without a six-week security review, and expand into new geographies without re-architecting your data flows. Treat it as an enabler and it behaves like one.
Part 4: The ACES Framework
Assessment fails when it produces a score and stops. A useful framework produces a decision.
The ACES framework evaluates every platform in your stack across four axes, then passes each one through a governance gate. The output is not a ranking. It is an instruction.
A: Adoption. Is it actually used, by the people it was bought for?
C: Contribution. Does it demonstrably move a business outcome?
E: Economics. Is what you pay proportionate to what you get?
S: Staying power. Will it still fit the business twenty-four months from now?
Scoring the four axes
Score each axis from 1 to 5. Anchor your scoring in observable evidence, not in how the tool feels.
| Axis | What you actually measure | Score 1 | Score 3 | Score 5 | Weight |
|---|---|---|---|---|---|
| Adoption | Active seats as a share of licensed seats over 90 days; depth of feature use; frequency of login by intended role | Under a quarter of seats active; one champion keeps it alive | Roughly half of seats active; core features only | Most seats active weekly; advanced features in routine use | 30% |
| Contribution | Traceable influence on pipeline, conversion rate, cycle time, retention, or hours saved | No line of sight to any outcome | Plausible contribution, not measurable | Clear, attributable, and defensible in a QBR | 30% |
| Economics | Total cost of ownership per unit of output, including admin time, integration effort, and internal support load | Cost rising faster than usage or output | Cost proportionate but not benchmarked | Below market for comparable capability; favourable terms | 20% |
| Staying power | Fit with the two-year GTM plan; vendor viability; switching cost; overlap with platforms you already own | Redundant, or vendor future uncertain | Fits today, unclear tomorrow | Strategic, embedded, and hard to replicate elsewhere | 20% |
Weighting matters. Adoption and contribution carry the majority because a tool that is used and works can usually be renegotiated. A tool that is cheap but unused is simply cheap waste.
The governance gate
The gate is binary. It sits underneath the score and it overrides it.
| Gate check | Pass condition |
|---|---|
| Data inventory | You can state precisely what personal data the platform holds and why |
| Lawful basis | A documented basis exists for every processing purpose |
| Consent lineage | Consent state and withdrawal propagate to and from this platform reliably |
| Residency | Storage and processing locations are known and contractually fixed |
| Sub-processors | The current sub-processor list is known, reviewed, and change-notified |
| AI processing | You know what AI features are active, what they access, and whether your data trains anything |
| Access control | Role-based access is enforced and reviewed at least twice yearly |
| Exit | Data export and deletion on termination are contractually guaranteed and tested |
A platform that fails any gate check cannot be classified as healthy, regardless of how well it scores. High performance on a non-compliant foundation is not performance. It is accumulated risk with good reporting.
Part 5: From Score to Decision
Combine the weighted score with the gate result, and every platform in your stack resolves into one of five actions.
| Weighted score | Gate result | Decision | What you do next |
|---|---|---|---|
| 4.0 to 5.0 | Pass | Retain and invest | Expand usage, negotiate multi-year for better terms, make it a system of record |
| 3.0 to 3.9 | Pass | Optimise | Drive adoption before the next renewal; if adoption does not move in two quarters, reclassify |
| 2.0 to 2.9 | Pass | Consolidate | Fold the capability into a platform you already own; do not renew standalone |
| Below 2.0 | Pass | Retire | Exit at renewal; redirect budget to a retain-and-invest platform |
| Any score | Fail | Remediate or retire | Fix within one renewal cycle with a named owner and a date, or exit |
Two disciplines make this work in practice.
First, every platform gets a named owner, a single person accountable for its score. Shared ownership produces shared inaction.
Second, the assessment runs on a fixed cadence, ideally quarterly for the top ten platforms by spend and annually for the long tail, timed to sit sixty to ninety days ahead of each renewal date. Assessment after a renewal is an autopsy. Assessment before a renewal is leverage.
Part 6: A Worked Example
Consider a mid-sized B2B organisation with a fourteen-platform stack. Eight of those platforms score as follows.
| Platform category | A | C | E | S | Weighted | Gate | Decision |
|---|---|---|---|---|---|---|---|
| Marketing automation | 4 | 5 | 3 | 5 | 4.3 | Pass | Retain and invest |
| CRM | 5 | 5 | 3 | 5 | 4.6 | Pass | Retain and invest |
| Intent data provider | 2 | 3 | 2 | 3 | 2.5 | Pass | Consolidate |
| Second enrichment vendor | 2 | 2 | 2 | 1 | 1.8 | Pass | Retire |
| Content experience platform | 2 | 2 | 2 | 2 | 2.0 | Pass | Consolidate |
| Conversational or chat tool | 4 | 4 | 4 | 3 | 3.8 | Fail | Remediate or retire |
| Analytics and attribution | 3 | 4 | 3 | 4 | 3.5 | Pass | Optimise |
| Sales engagement | 5 | 4 | 3 | 4 | 4.1 | Pass | Retain and invest |
Read the pattern rather than the individual rows. Two platforms are carrying the portfolio and deserve more investment, not less. Three are candidates for consolidation or exit, and between them they are likely absorbing meaningful budget for marginal contribution. One is performing well but sitting on a governance failure, which makes it the single most urgent item on the list despite its healthy score.
That is the value of running the gate alongside the score. Without it, the chat tool looks like a success story.
Part 7: A Ninety-Day Rationalisation Sequence
Assessment without a sequence becomes a spreadsheet nobody opens twice.
| Phase | Window | What gets done | Output |
|---|---|---|---|
| Inventory | Days 1 to 15 | List every platform, cost, renewal date, owner, and data category. Include tools bought on individual cards. | A single stack register |
| Instrument | Days 16 to 35 | Pull real usage data. Licensed versus active seats, feature depth, login frequency by role. | Adoption evidence |
| Score | Days 36 to 55 | Run ACES with the owners in the room, not on their behalf. Disagreement here is signal. | Weighted scores |
| Gate | Days 36 to 55 | Run the governance checklist with legal or your DPO. | Pass or fail per platform |
| Decide | Days 56 to 70 | Map to the five decisions. Sequence by renewal date, largest spend first. | A dated action plan |
| Execute | Days 71 to 90 | Serve notice where required, open renegotiations, launch adoption plans, remediate gate failures. | Realised savings and reduced risk |
Two practical notes. Serve notice early on anything you intend to exit, because notice periods are the most common reason a retirement decision slips a full year. And treat the savings as reallocation, not reduction, because the fastest way to lose organisational support for rationalisation is for the recovered budget to disappear into a general pool.
What Good Looks Like Twelve Months On
Organisations that manage martech as a portfolio tend to arrive at a similar place, and it is a recognisable one.
The stack is smaller and the capability is larger. Fewer platforms, deeper adoption, and a clear system of record for each core function. Reporting is faster because reconciliation has largely disappeared. New marketers are productive in days rather than weeks. Security questionnaires are answered from a document rather than a scramble. Renewals are negotiated from a position of evidence, because you know exactly what usage and contribution you are bringing to the table. And the budget conversation shifts from defending the line item to allocating within it.
None of that comes from buying better tools. It comes from managing what you own with the same rigour you would apply to any other portfolio of assets.
Five Questions for Your Next Stack Review
- Which platform in our stack has the highest cost per active user, and can we defend it?
- If we had to demonstrate consent lineage from form fill to ad platform tomorrow, could we?
- Which two capabilities are we currently paying for twice?
- Which platform would break something important if we switched it off on Monday, and which would nobody notice?
- When was the last time we retired something, and what happened to the budget?
If any of those questions takes more than a week to answer, the answer itself is the finding.
The stack you have is a set of decisions you made in the past. The stack you need is a decision you make deliberately, on a cadence, with evidence. The difference between the two, compounded over a few growth years, is one of the larger controllable variables in go-to-market efficiency.