What a stack audit is for
The marketing technology hub describes a first stack review in three artefacts: an inventory, a use-case map and a data flow diagram. Its table of questions for keeping the stack lean (what job a tool does, who owns it, what it connects to, what would break if it went) is the audit's test for every tool. This tutorial turns those into a procedure with an end point, because an inventory that never produces a decision is a list, not an audit.
The audit is not the only review a stack gets. It sits beside several narrower ones, and it uses their results rather than repeating them.
| Review | Question it answers | How the stack audit uses it |
|---|---|---|
| Stack audit (this tutorial) | Which tools does the business have, and should each one stay? | The whole picture and the decisions |
| Tag governance quarterly audit | Do the tags on the site match the inventory, consent and the budget? | Tag findings go there |
| Analytics implementation audit | Can the tracking be trusted for decisions? | Tracking findings go there |
| Access review | Can the right people, and only they, sign in to each tool? | Access findings go there |
| Integration register | What connects to what, and how would a failure show? | Connection findings go there |
| Takeover inventory | What accounts exist when a programme changes hands? | A one-off audit of the same kind, at handover |
The audit's output is three things: the inventory, kept current afterwards as the stack record; the use-case map, showing what each tool is for, where tools overlap and where a use case has no tool; and a ranked list of decisions, each with one owner and a date. Between audits, the marketing stack management playbook keeps that stack record current through request triage, onboarding, access and seat reviews, renewals and retirement, so the next audit starts from the record rather than from scratch.
Before you start
Set the scope first. A whole-stack audit covers every tool with a login, a contract or a tag on the site, including free tiers, trials and tools an agency runs on the business's behalf. A narrower audit can take one layer, such as the tools that send messages, but should say so, because a finding in another layer will still turn up.
Then gather what the audit reads. Finance's records of software spend: invoices, company card statements and expense claims, because tools bought on a card rarely appear anywhere else. Access to the tag manager (GTMTag management system.) container and the ability to inspect the live site in a browser. Administrator access to the tools, or someone who has it. The use-case list from the designing a marketing stack guide, or an afternoon to write one. The tutorial on writing the use-case list walks through that afternoon. Any data flow diagram the business has. And the stack record, if there is one, as the starting list to be corrected.
Agree who is involved with a RACI matrix: one person accountable for the audit, finance and the site's owner consulted for spend and tags, each tool's owner responsible for answering about their tool, and whoever owns data protection consulted on the tools that hold personal data.
The steps
Set the scope, the owner and the decisions the audit must produce
Write down what is in scope, who is accountable, the date the decisions are due and who signs them off. Name the four outcomes every tool will end with: keep as it is, fix (a missing owner, an untidy integration, a tag outside the container), consolidate with another tool, or retire. Saying at the start that every tool gets one of these stops the audit ending as a spreadsheet with no conclusions. Agree also what is out of scope for decisions, such as a platform the business is contractually committed to, so the audit records it without reopening it.
Find every tool from invoices, card statements, logins and the live site
Start with the existing list, then look for what it misses. Read invoices, card statements and expense claims for anything that looks like software. Inspect the live site's main templates in a browser's developer tools and compare what loads with the container; the tag governance guide notes that the difference between the two is the first finding of most audits, and a script or pixelAn ad platform's browser code that reports page views and events on a site back to the platform. loading on the site is often the only trace of a tool nobody remembers buying. Look in each tool's integration settings for connected apps. Ask each team which logins they use every week and which tools an agency runs for them. The takeover tutorial uses the same sources for the same reason: the tools nobody mentioned are found by looking.
Record each tool's owner, cost, renewal date and users
For every tool, fill in the inventory row: the owner, the cost and how it is billed, the renewal date and the date by which notice must be given, the number of people who use it and who holds administrator access. The marketing stack hub says an owner can log in with administrative rights, knows what the tool is connected to, holds the contract and renewal date and is the person a change request goes to; record where a named owner meets none of those. A tool with no plausible owner is already a finding.
Map each tool to the use cases it serves
Take the use-case list and, for each use case, mark the tools it depends on and whether it is live, planned or an aspiration. Then read the map the other way: for each tool, which use cases does it serve? The designing guide calls this the test every tool should pass, whether it serves a written use case and which. Where a tool's category is unclear, the MarTech categories guide helps place it, and a tool that spans several layers can be mapped to more than one.
Mark overlaps, gaps and tools with no use case
Three patterns show up on the map. Overlaps: two or more tools serving the same use case, often for different teams; the categories guide explains where categories typically overlap and the tool categories hub lists the usual resolutions. Gaps: a live or planned use case with no tool, or one served by a manual workaround. And tools with no use case at all, which are retirement candidates unless someone can name the job they do. Mark each one; the decisions come later.
Note what each tool connects to and how
For each tool, record what it receives and sends, from and to which systems, and by what method: a native connector, an interface someone built, a scheduled file, or a person exporting and uploading by hand. Manual steps are worth noting in particular, because they are the integrations that stop when someone is away. The data flow and integrations guide describes the patterns, and the tutorial on documenting marketing integrations builds the register that holds the detail. The audit needs only enough to know what would break if a tool went, and where a CDPUnified first-party customer profiles. or the CRM is the hub, which tools depend on it.
Collect the findings on tags, access and data handling
Gather the findings that belong to narrower reviews, without solving them inside the audit. Tags: tags of lapsed tools, tags outside the container and tags with no owner go to tag governance, and the tutorial on moving hard-coded tags covers bringing code into the container. Access: former staff or agencies with logins, shared accounts and tools with one administrator go to the access review. Data handling: which tools hold personal data and whether a privacy review was recorded when each arrived, as the marketing tool onboarding SOP requires; decisions on these belong to whoever owns data protection. Numbers that disagree between tools go to the data quality guide. Licence findings, such as more paid seats than users, go to the licence and seat review.
Score and rank the findings
List every finding and every candidate decision in one place, then sort them with the impact-effort matrix: the expected effect of acting against the effort it takes. Quick wins, such as removing a lapsed tag or an unused subscription, go first. Major projects, such as consolidating two messaging tools, need a plan. Fill-ins wait for a quiet week, and low-impact, high-effort items are dropped and recorded as such. The matrix is a first sort, not a final ranking; dependencies and contract dates can reorder it, and the next step takes those into account.
Decide keep, fix, consolidate or retire, with owners and dates
For each tool, record the decision, the reason, one accountable owner and a date. Time decisions that touch a contract to its notice date rather than its renewal date, because the notice date is the one that can be missed. Consolidation follows the stack design rather than the renewal calendar, as the tools hub puts it. A tool to be retired leaves in the order the tutorial on retiring a marketing tool gives. Gaps go to the design roadmap. Then update the stack record, attach the use-case map and the decision list, and set the date of the next audit. The KPIA metric tied directly to an objective. a kept tool is meant to move goes in its row, so that the next audit can judge it.
The audit inventory
The inventory is one row per tool. It becomes the stack record once the audit ends, so it is worth keeping simple enough to update. From then on the stack recordThe live register of every marketing tool: owner, deputy, cost, notice date, users, use cases and status, kept current by onboarding, reviews and retirement. is changed by each routine that brings in, reviews or retires a tool, not only by the next audit.
| Field | What to record | Where it comes from |
|---|---|---|
| Tool and category | What it is, in plain words, and its category | The tool; the categories guide |
| Owner and deputy | A named person who can log in as administrator, and a second | The tool's user list; the team |
| Use cases served | The live or planned use cases it supports | The use-case map |
| Users | Who uses it, and how many | The tool's user list |
| Cost and billing basis | What it costs and what it bills on | Invoices, card statements, expense claims |
| Renewal and notice dates | When the contract renews and when notice is due | The contract or terms accepted |
| Connections | What it receives and sends, from and to where, and how | Integration settings; the integration register |
| Tag on the site | Whether it loads a script, where and how it is deployed | The live site and the container |
| Personal data | Whether it holds personal data, and whether a privacy review is on file | The onboarding records; data protection |
| Findings | Overlaps, gaps, tag, access and data findings | Steps five to seven |
| Decision | Keep, fix, consolidate or retire, with owner and date | Step nine |
A veterinary group audits its stack (illustrative)
Haldrupex Vet Group audits its marketing stack
Haldrupex Vet Group, an invented group of 7 veterinary practices, audits its marketing stack ahead of the budget round. Its own list shows 14 tools. Card statements and expense claims add 5 more, mostly bought by individual practice managers, and a scan of the site's booking and practice pages finds 2 further scripts: a booking widget tag left from a trial that lapsed 10 months earlier and a chat tool nobody on the central team knew about. The inventory ends at 21 tools.
Mapping the tools to the group's 12 written use cases shows two appointment reminder tools doing the same job, one bought centrally and one by 3 of the practices on a card, and 4 tools that serve no use case at all. The connection notes show that the second reminder tool receives appointment data through a weekly file someone uploads by hand. Access findings go to the access review: 2 tools have only one administrator, and one is a former practice manager.
Ranked on the impact-effort matrix, the quick wins are removing the lapsed booking tag and cancelling 3 of the 4 unused tools; the fourth turns out to host the forms for the group's puppy classes and is kept with a named owner. The major project is consolidating reminders. The group keeps the central reminder tool, moves the 3 practices onto it, gives notice on the other tool 6 weeks ahead of its contract's notice deadline and retires it in the order the retirement tutorial sets. Every one of the 16 tools that remain has a named owner, a deputy and a renewal date in the stack record, and the next audit is in the calendar.
Common mistakes
- Starting from the list you already have. Card statements, expense claims and the live site always find more; the audit's value is in what the list missed.
- Auditing without decisions. An inventory with no keep, fix, consolidate or retire against each tool is a list. Name the decisions before starting.
- Solving every finding inside the audit. Tags, access, integrations and tracking each have their own review; send findings there and keep the audit on the tools.
- Ranking by cost alone. A cheap tool with no owner and personal data inside it can matter more than an expensive one that works. Rank by impact and effort.
- Timing to the renewal date. The notice date comes first; a decision made after it is a renewal by default.
- Decisions with no single owner. Shared actions wait for each other; give each one accountable person.
- Treating the audit as a one-off. The inventory becomes the stack record; without a next date, it goes stale as fast as the last one did.
