In 2011 a marketing technologist in the United States drew a chart of every marketing software product he could find. There were about a hundred and fifty. This May he published the same chart with fifteen thousand five hundred and five, and for the first time in fifteen years it barely grew, by less than one per cent.
The people buying those tools, according to Gartner, are spending a smaller share of their budgets on them than at any point in five years, and using somewhere between a third and a half of what they buy. Fifteen thousand tools. A startup needs perhaps five. This episode is about how to choose them, and what to do with them once chosen.
Recorded in Sydney, thirty-three degrees fifty-two minutes south. This is The Founder's Stack. I'm Daniel Drakopoulos.
Last episode we talked about go-to-market blueprinting, the document that sets out audience, positioning, channels and sequence before the first campaign. This episode is the stage that follows it in my firm's framework: marketing stack configuration. Selecting and configuring the platforms the marketing function runs on.
I put configuration in the name deliberately. The industry talks about buying. The value is in configuring. And this year, with the martech landscape flat, budgets shifting toward AI and paid media, and the major vendors moving to consumption pricing, the gap between what a company buys and what it uses has become a line in the financial model.
Scott Brinker and Frans Riemersma's 2026 Marketing Technology Landscape lists fifteen thousand five hundred and five products. Up a hundred and twenty-one on last year. Nought point seven nine per cent. Brinker's own reading was that the market has, finally, hit peak martech.
Underneath that, the churn is fierce: fourteen hundred and eighty-eight products added, thirteen hundred and sixty-seven removed. Growth has moved to specific places, namely content management, e-commerce, analytics, integration, governance, and what they call AEO and GEO, which is optimising to be cited by AI answer engines.
Their companion study makes a stronger claim. Stacks are flattening. SaaS is devolving into infrastructure. And AI agents are taking over the decisioning layer that used to sit inside each tool. So the number of tools is not the story. The story is what sits above them, and who owns it.
Now the buyers. Gartner's 2026 CMO Spend Survey covered four hundred and one marketing leaders, mostly at large companies, so read it with that caveat. It puts marketing budgets at seven point eight per cent of revenue. Martech at nineteen point four per cent of that budget, down from twenty-six point six in 2021, a five-year low. AI initiatives at fifteen point three per cent. Paid media up to thirty-one point four. Labour at twenty-four and a half.
Then utilisation. Gartner's marketing technology survey had marketers using fifty-eight per cent of their stack's capability in 2020, forty-two in 2022 and thirty-three in 2023. The 2025 survey reported a rebound to forty-nine. The reasons given were ecosystem complexity, customer-data problems and inflexible governance.
Those are enterprise figures. In my experience the startup version is worse rather than better, because nobody in a twelve-person company has time to configure anything.
The real cost is not the subscription. It is that every unconfigured tool holds a piece of customer data nobody reconciles. You end up with three definitions of a lead, and the licence is the cheap part.
There is a fair objection to that, which is that an unused tool is sometimes a cheap option on a future capability. An option is fine if someone knows it is an option. The problem is the tool nobody remembers buying.
Here is the part I think founders have not priced in. The way these platforms charge is changing.
HubSpot, from the fourteenth of April this year, moved its Customer Agent and Prospecting Agent to outcome-based credits. Fifty credits per resolved conversation, about fifty US cents at the standard credit price, down from a dollar per conversation. A hundred credits per recommended lead, about a dollar. Credits at ten dollars per thousand monthly, nine annually. You pay when the task completes.
Salesforce sells Agentforce three ways. Two dollars per conversation. Flex credits at five hundred dollars per hundred thousand, roughly twenty credits an action. Or per-user licences from a hundred and twenty-five a month. Salesforce has reported eighteen and a half thousand Agentforce deals, about nine and a half thousand paid, and five hundred and forty million dollars of Agentforce ARR by its third quarter. The three models exist because the original two-dollar rate priced out smaller buyers.
And from Gartner again: fifty-six per cent of CMOs have increased the share of martech budget on consumption-based pricing. Nine per cent decreased it.
That changes what a configuration error costs. Under seat pricing, a badly configured tool wasted a subscription. Under outcome pricing, a badly configured agent spends money on every wrong resolution. Configuration has become a cost control.
So the selection decision is now a financial model decision. Which tool, at which price model, against which volume assumption from the blueprint. That is why, in my framework, this stage sits directly after blueprinting and before anything is automated.
Let me set out how my firm does platform selection, because it is the opposite of how most stacks get built. The order is: the blueprint decides the requirements, the requirements decide the platform, and configuration is where the value is.
In practice, the blueprint from stage one says who the company is for, how it is positioned, which two or three channels, and in what sequence. From that you can write down requirements rather than features. We need to capture leads from these two channels into one record. We need one definition of a qualified lead that marketing and engineering both accept. We need lifecycle messaging on these three triggers. We need reporting that reconciles with finance.
Only then do you look at platforms. And when you look, you look for the smallest set that meets the requirements and can be configured by the people you actually have. Not the most capable. The most configurable by this team. Then comes the implementation roadmap: what gets configured first, what waits, and what is explicitly not being done this quarter.
The pattern I see in platform decisions is consistent. They come down to data model fit and to who will maintain the thing, rather than to features. The runner-up usually had more capability and nobody to run it. Nobody to run it is the seed-stage reality, and the right platform is the one your team will actually configure.
What belongs on the roadmap rather than in the cart is anything that needs clean data the company does not yet have. Attribution modelling. Predictive scoring. Multi-touch anything. Those are quarter-three items for a company that has not yet agreed what a lead is.
Let me be concrete about the short list, as an opinion rather than a rule, because people ask. A system of record for customers, a CRM, configured with one lifecycle definition. A messaging and automation layer that reads from it. Analytics that capture what the product and the website actually do, owned jointly with engineering. A place the data reconciles, which for a small company may be the CRM itself and later a warehouse. And a consent layer, which we will give its own episode.
That is four or five things. Everything else is an option on the future and should be bought as one, when the requirement arrives rather than before.
Two notes on that list. First, configured with one lifecycle definition is doing most of the work in the sentence. The CRM is not the asset. The agreed definition of a lead, a qualified lead, an opportunity and a customer is the asset, and the CRM is where it is written down. If marketing and engineering do not share that definition, no platform will reconcile the numbers for them.
Second, owned jointly with engineering, on analytics, is not a courtesy. The events that matter most for acquisition, meaning sign-up, activation and first value, are fired by the product rather than by the website. If engineering does not own the instrumentation, marketing will be reporting on a funnel that stops at the form.
The comparison everyone asks about. The 2026 reviews line up fairly consistently. Salesforce with Agentforce for enterprises that need deep customisation and cross-cloud orchestration. HubSpot for small and growth-stage teams that need fast deployment without a large implementation budget or a dedicated administrator.
That rule breaks in three places. It breaks when a startup sells into enterprise and needs to match its customers' systems. It breaks when a technical team would rather run a warehouse plus lighter tools than either suite. And it breaks on pricing once agent usage scales.
The third option is the one I would flag. A warehouse as the system of truth, with lighter tools reading from it. It suits technical founding teams, and it is where a lot of the growth in the martech landscape is pointing, namely integration, analytics and governance. We will spend a whole episode on that pipeline.
One more thing from Brinker's 2026 predictions. The shift he describes is from AI for marketers to AI for customers, with buyers using agentic browsers and assistants to do the comparing.
What changes in configuration when the buyer has an agent too is mostly about data rather than tools. Structured, machine-readable product and pricing information. Answer-engine visibility. Fewer gated assets. Consistent data across every surface an agent might read. Which is an argument for the same discipline, one data model consistently exposed, rather than for more tools.
The question I ask everyone on this show is this. What is the one thing in your stack you would rip out tomorrow, and what would you put in its place?
That is The Founder's Stack for this episode. The summary, the reading list with every source, including Brinker's Marketing Technology Landscape, Gartner's survey and the HubSpot and Salesforce pricing, and the full script are at danieldrakopoulos dot com under Media. The archive is at drakopodcast dot net. The monthly newsletter covers emerging platforms in more detail.
Next time: one data set for the board. Data pipeline integration, and what it does for investor reporting.
Alignment between creative marketing and engineering removes operational friction and accelerates product-market fit. I'm Daniel Drakopoulos.