The Marketing Tool Graveyard (And How to Stop Adding to It)

Open the company card statement of almost any agency we talk to and the same pattern is there: four figures a month going to marketing tools nobody opens. Six or seven platforms on the bill, three people who actually log in, and not one person who can tell you which tool drives revenue and which is dead weight. That is the pattern behind the truth card on our homepage, the £4k a month on tools nobody opens, and it is not an accounting problem. It is a process problem wearing an invoice. This page takes that leak apart properly: why tools end up in the graveyard, how to audit the stack you already pay for, and the process-first alternative that stops you adding to it. It goes deeper than the tool-sprawl section of our pillar on AI for marketing agencies, which is where the three leaks first get named.

Why tools end up in the graveyard

A tool dies in the same way every time, and it is almost never because the tool was bad. It dies because it was bought before the process was mapped. Someone hit a problem, a client asked an awkward question, a competitor mentioned a platform on a panel, and the answer was a subscription. The tool arrived first and the process it was meant to serve was assumed, never drawn out. So nobody quite knew which step it replaced, who owned it, or how you would know it was working. It got a trial, the trial became a renewal because cancelling felt riskier than paying, and within two quarters it was another login that one person remembered and everyone else forgot.

This is tool-first adoption, and it is the default in our market because it feels like progress. Buying a tool is fast, visible, and easy to defend in a meeting. You can point at the new platform and say you are doing something about the problem. Mapping the process is slow, quiet, and uncomfortable, because it usually shows that the work the tool was meant to fix was never designed in the first place. So the agency keeps reaching for the card. Every new client problem gets met with a new subscription, every shiny launch gets a trial, and the stack grows because subtracting feels like admitting a mistake and adding feels like momentum.

Process-first adoption runs the other way, and it is the discipline almost nobody applies. You map the work before you buy anything. You draw the steps a person actually takes, you find where the hours and the errors leak, and only then do you ask whether a tool belongs there at all. A lot of the time the honest answer is that the step should be removed, not tooled. When you map first, a new platform has to earn its place against a clear picture of the work, which means most of the tools that would have ended up in the graveyard never get bought. The reason the graveyard fills up is not weak willpower. It is that the buying decision keeps getting made before anyone has seen the process it is supposed to improve. That is the one idea worth holding onto here: tools die because they were bought before the process was mapped.

There is a quieter cost on top of the wasted spend, and it is the one owners feel without naming. Every extra tool is another place the data lives, another export, another thing to keep in sync by hand. The more platforms you add to save time, the more manual stitching you create, and a person ends up being the integration layer between tools that do not talk to each other. Tool sprawl does not just cost the licence fee. It feeds the reporting leak and the prospecting leak by scattering your numbers across systems that were each bought in isolation. We unpack that connection in the glossary entry on tool sprawl.

How to audit your stack

The audit is not glamorous and it does not need a tool of its own, which is the first joke worth getting ahead of. You need a spreadsheet, the company card statement, and an honest afternoon. The aim is to replace the comfortable assumption that everything you pay for is earning its keep with a list of what is actually true. There are three questions, and you run every line item on the bill through all three.

The first is usage versus licence count. Pull the real login data, not the seat count you are billed for. A lot of agencies pay for seats nobody opens, and the gap between licences bought and humans logging in is usually the single most revealing number in the whole exercise. If you are paying for ten seats and three people have logged in this quarter, you are not buying capability. You are buying a default that renewed itself. Write the last-login date next to every tool. The ones that have not been touched in a month are already in the graveyard, you just have not filed the paperwork.

The second is revenue attribution per tool. For each platform, answer one blunt question: if this disappeared on Monday, what client work would actually stop or get worse? Not what might theoretically suffer, what would visibly break. Some tools earn their place instantly here, because a delivery line depends on them and you would feel it the same day. Others cannot be tied to any client outcome at all, which tells you they were bought to solve a feeling rather than a process. You are not looking for perfect attribution. You are looking for the tools that nobody can connect to revenue, because those are the ones the graveyard is built from.

The third produces the output that makes the audit worth doing: the kill list. Any tool with near-zero logins and no clear line to client work goes on it. Cancel those first, because they are pure leak with no defence. Then the harder middle tier, the tools one person uses for one task that another platform you already pay for could cover. Consolidate those, even though it means a short migration, because two tools doing one job is the sprawl pattern in miniature. What survives the kill list is the small set of platforms that real people use to do work clients pay for. That is your actual stack. Everything else was the graveyard, and you have just stopped feeding it.

The process-first alternative

Killing what nobody uses buys back the spend, but it does not fix the habit that filled the graveyard in the first place. For that you need to change the order of operations permanently: map first, then decide build, buy, or kill. Before any new tool gets onto the card, the step it is meant to serve has to exist on a process map, with an owner and a measure attached. If the step cannot be drawn, the tool cannot be bought, because there is nothing for it to improve yet.

Once the process is on the wall, the decision is rarely as simple as buy. Sometimes the right answer is build, a thin layer that pulls existing data together so a person stops being the glue, and sometimes that layer uses AI and sometimes it is just plumbing. Often the right answer is kill, removing the step entirely because mapping it revealed it added nothing. Buy is only correct when a mapped, owned, measured step genuinely needs a capability you do not have and cannot reasonably build. That is a much higher bar than the card has been clearing, and it is meant to be. This is the same discipline we apply to every agency leak: it is how we treat the reporting grind in where to start with AI in your marketing team, and it is the through-line of the whole pillar. It also lines up with the uncomfortable industry numbers: MIT NANDA found that 95% of organisations investing in generative AI report no measurable return [MIT NANDA, 2025], and McKinsey found that while 88% of businesses now run AI in at least one function, only 39% can point to a measurable effect on profit [McKinsey, 2025]. The gap is not the technology. It is buying before mapping, at scale.

Where to start

You do not start by buying a tool to manage your tools, which would be the graveyard eating its own tail. You start by mapping. The value stream mapping tool walks you through drawing one process end to end, so you can see the steps before you decide which of them deserve a platform, a build, or a quiet cancellation. Map one painful process first, run its tools through the three audit questions, and you will usually find the first few lines of your kill list inside an hour. The spend you free up is real, it is recurring, and it is the cheapest first win an agency can give itself.

→ Map a process with the value stream mapping tool

References

• MIT NANDA. "The GenAI Divide: State of AI in Business 2025." MIT Media Lab Project NANDA, 2025. Source of the 95% no-measurable-return figure.
• McKinsey & Company (QuantumBlack). "The State of AI." 2025. Source of the 88% adoption versus 39% measurable profit-impact gap.