Martech Consulting: Cut the Stack and Build the Core
The build-vs-buy math nobody runs for you, what is actually worth consolidating, and the four things you should keep buying forever.
There's a simple test for martech consulting. If someone walks your stack and comes back recommending a new platform, with a new seat licence and a new onboarding call, they sold you something. If they can point at four tools doing one job and tell you which three to switch off, that's the real thing. The savings usually dwarf the fee.
I've been on both sides of that conversation for the better part of a decade. Below is the math nobody runs for you, what's actually worth consolidating, and the part most agencies skip: what you should never build yourself.
On this page
- Why your stack got expensive without anyone deciding
- What martech consulting actually does
- The build-vs-buy math nobody runs
- What to consolidate first
- What you should never build yourself
- Frequently asked questions
Why your stack got expensive without anyone deciding
Nobody sits down and decides to run fourteen tools. It accumulates.
The email platform came first, and it was fine. Then sales wanted a sequencer. Then someone needed a scheduling link, and the free one was right there. A form tool, because the CMS forms were annoying. A separate analytics thing, because nobody trusted the numbers in the first one. Each decision was defensible on the Tuesday it happened. Nine months later you're paying four vendors to store the same email address and none of them agree on what a "lead" is.
The industry is not helping. There are now over 15,000 martech products across 500+ categories, which is roughly 15,000 more than any marketing team can evaluate. And most of what gets bought sits idle. Zylo's SaaS Management Index found 46% of applications go underutilised or entirely unused. Half your software budget is buying shelf-ware.
But the licence fees aren't the expensive part. The expensive part is the RevOps person exporting a CSV at 11pm on a Friday because the form tool doesn't talk to the CRM. That's a salary being spent on copy-paste. That's the real bill, and it never shows up on a line item anyone reviews.
What martech consulting actually does
Strip away the deck and it's four things. Work out what you actually have. Work out what the data is supposed to do. Connect the pieces that matter. Switch off the rest.
Most of the value is in that last part, and it's the part nobody wants to sell you. There's no recurring commission on turning something off.
Here's my position, and I'll defend it on a call. A logo wall of "40+ integrations" is fake expertise. It signals breadth, which is the opposite of what you need. I'd rather work with someone who knows eight platforms cold, every rate limit, every field-mapping quirk, every place the API lies to you, than someone who has dabbled in forty. We run 40-odd connectors in production, but we're genuinely deep in eight. Those are different claims, and the second one is the one that saves you.
The unglamorous truth about this work is that it's mostly mapping. Mapping, and then more mapping. When you wire systems together the data mapping has to be exactly right. One wrong value and your whole attribution chain breaks silently. Not loudly. Silently. You keep reporting numbers for a month before anyone notices they're fiction.
That's why we do this as marketing technology integration work rather than as a strategy deck. The deck doesn't have the field mappings in it.
The build-vs-buy math nobody runs
Here's the arithmetic almost nobody does, and it isn't complicated.
Take a tool you pay for monthly. Multiply by 12. Then add the integration tax: the hours someone spends every month keeping it stitched to everything else, the annual price rise, the migration cost you'll eat in three years when they get acquired and sunset the plan you're on. That's the real number.
Now compare it to building the same capability once.
For a lot of core functions, a lead pipeline, an internal dashboard, a scheduling flow, a client portal, the build is a few weeks and then it's yours. No seat licences. No per-contact pricing that punishes you for growing. No vendor deciding to reprice at renewal because they know switching costs you three months.
Where this gets dramatic is the middle of your stack, the connective tissue. Point-to-point tools that exist purely to move data between two other tools are the first thing that should go. That's the layer where a small amount of custom code replaces an entire subscription tier and does it better, because it knows your schema.
Two things to be honest about, because this is where "build it yourself" advice usually goes quiet.
You are buying maintenance, not just a build. Something you own is something you patch. That's a real, permanent cost, and anyone who tells you otherwise is selling a project rather than a system.
Custom is only cheaper when it's genuinely core. If it isn't something your business does differently from everyone else, buy it. Nobody has ever won a customer because their internal expense tool was bespoke.
The right split, in my experience: buy the commodity, build the thing that is actually your business. Then wire them together properly so the seams don't leak. That's where custom development earns its keep, and where most of the compounding savings live.
What to consolidate first
Not everything at once. In order of return:
Anything that duplicates your source of truth. If two systems both think they own the customer record, you don't have a stack. You have a disagreement. Pick one. The other becomes a consumer.
Anything a human is manually moving. Every CSV export is a subscription you're paying for twice, once in licence and once in salary. When we wire lead capture straight through, leads land in the CRM within seconds and nobody touches a spreadsheet. That isn't a productivity gain. It's a job you get to stop doing.
Anything that only exists to report. Most standalone reporting tools are compensating for bad plumbing upstream. Fix the plumbing and the reporting tool stops being necessary. We had a large South African ecommerce brand mid-migration from UA to GA4 that was, candidly, a train wreck. Weeks of combing through data and fixing broken calls got them a 70% improvement in the data quality coming into GA4, with event properties finally landing. No new tool. Just the existing one, finally correct. If you're facing that same migration, we wrote up the how-to.
Anything blind to your outcomes. The most expensive tool in most stacks is the ad platform that never learns a deal closed, so it keeps spending to retarget people who are already customers. Closing that loop cut CAC by 42% on one engagement, not by spending less, but by spending it at people who hadn't bought yet.
What you should never build yourself
I'd be a bad consultant if I let you walk away thinking everything is worth building. Some things you should buy forever, and I mean forever.
Payment processing. Card data is a regulated liability. Stripe and its peers employ compliance teams larger than most agencies. Use them.
Email deliverability. You can build a sender. You cannot build a decade of IP reputation, and your mail will land in spam while you learn that.
Authentication. Rolling your own password storage is how you end up in a breach notification. Use a real identity provider.
Anything with an annual filing obligation. Payroll tax, 1099s, sales tax nexus. The penalties for getting these subtly wrong compound quietly, and the software is cheap by comparison.
Anything where a bug is a headline. If the failure mode is a regulator or a journalist, that's a bought problem.
Notice the pattern. Buy the things where the cost of being wrong is catastrophic and someone else has already absorbed it. Build the things where the cost of being wrong is a Tuesday afternoon.
The same test applies to AI, which is where this question is loudest right now. A general chatbot is a commodity, so buy it. But a model running on your own hardware, on your own client data, that never sends a privileged document to someone else's API? That's core, and that's why private AI agents are worth owning outright if your data is the kind that can't leave the building.
Frequently asked questions
What is martech consulting? It's the work of auditing, selecting, connecting and operating your marketing technology: CRM, automation, analytics, tracking and the data pipelines between them. Good martech consulting is judged on whether the stack got simpler and the numbers got trustworthy, not on how many tools were added.
How much does martech consulting cost? Ours starts around $18k for an initial wiring engagement, typically 4 to 8 weeks, with ongoing work from $6k a month. Compare that against what you're spending on redundant licences before you decide it's expensive. For a lot of teams the consolidation pays for the engagement inside a year.
Is it cheaper to build your own marketing tools? For core functions specific to how your business runs, usually yes. You pay once instead of monthly, and you're not exposed to renewal repricing. For commodity functions like payments, email sending or authentication, no. Buying those is cheaper and considerably safer.
How do I know if my stack has too many tools? Two quick tests. Can you name, without looking, which system is the source of truth for a customer record? And is anyone on your team exporting a spreadsheet on a regular schedule? A "no" to the first or a "yes" to the second means you have plumbing problems, not tooling problems.
How long does consolidating a martech stack take? An initial integration is typically 4 to 8 weeks depending on how many systems are in play and how bad the field mapping is. The mapping is the long pole. It's slow, it's unglamorous, and rushing it is how you get attribution that lies to you for a quarter.
Will I lose data if I switch off a tool? Not if it's sequenced properly. You export and reconcile before anything gets decommissioned, and you run the new path in parallel until the numbers agree. Anyone who wants to cut over on a single date without a parallel run should be politely ignored.
Do I need a martech consultant or a developer? If you know exactly what to build, hire a developer. If you're not sure which of your fourteen tools should survive, that's a consulting question first, and the answer determines what actually needs building.
Fewer tools, better wiring
We started Social Catnip in 2016 and we've been doing this from Glenrock, Wyoming ever since. A small senior bench, average engineer tenure of 11 years, veteran-owned. That tenure is the whole point: this work rewards people who have already made the mistakes and remember where the bodies are.
If your stack has grown past the point where anyone can explain it in one breath, the first conversation isn't a demo. It's a walk through what you're paying for and what you'd stop paying for. Let's map your stack and find the three tools you can switch off.