I’ve been saying for a long time that MadConnect’s #1 competitor isn’t another platform, ETL, Reverse ETL, or connector library: it’s actually companies building connectors themselves.
The truth is that building a connector is straightforward, and while I don’t know anyone who likes building them, almost any engineer can.
Given that we have made a business out of doing something anyone can do, you have to ask why. So, let us peel this onion back a few layers.
We agree building a single connector is easy. We disagree that building an enterprise-grade connector to completeness, one with full functionality, that’s scalable, self-healing, manages large-volume transfers, optimizes for retries, auto-chunks a file to meet rate limits, and sends multiple concurrent chunked files 10-100X faster, among many other nuances, is simple, or even worth it, if connectivity isn’t your business or core product.
We’re not even talking about developing comprehensive observability to monitor the connector for failures, keeping it up to date, and not dropping innovative product functionality off your roadmap in the process.
Remember, I’m only talking about one connector. But let’s be clear: the average platform in our industry requires building four core connectors. Take a simple use case like activation: you’ll need an audience activation, reporting, conversion tracking, and trafficking connector. But what business are you in? What about measurement, planning, buying, enrichment, or optimization? What if you want to hook up an MCP and have an agent manage these connections for you? Now we’re talking about a synchronous connector, one giving you real-time responses and metadata for every action the agent takes.
I’ll tell you right now: this onion we’re peeling back, we’re nowhere near the core. We’re only a few layers in.
You’ve read this far, and you’re saying, “so what, I know it’s not my core business, I know it won’t differentiate me to build these connectors, but I’m doing it anyway.” I think now is the time to do some math. Let’s stick with audience activation:
- 5 activation platforms (Meta, TikTok, Snap, Pinterest, TTD) x 4 connectors each = 20
- How many engineers: 2
- How many QA: 1
- Build time per connector: 4 weeks
- Testing and QA: 2 weeks
- Maintenance: 4-6 updates a year (we’ll use 4/yr), 2-week fix each time
- Total time per connector spent per year: 14 weeks/year/connector x 20 = 280 weeks of effort, or 5.38 FTE engineers for initial buildout
- Annual update work: 2 wks x 4/year x 20 = 160 weeks, or about 3 FTE engineers/year
- How much does your average engineer in US Advertising/AdTech get paid? $160K-$230K fully loaded
- How much does your offshore engineer get paid? Average, skewing higher for AdTech, is about $75K fully loaded
- Total estimated cost of connectivity for managing audience activation across 5 platforms:
- US: Buildout: $160K x 5.38 = $860,800 || Annual maintenance: $160K x 3 = $480,000
- Offshore: $75K x 5.38 = $403,500 || Annual maintenance: $75K x 3 = $225,000
These are estimates, but from our experience, no company builds 20 connectors in a year. And if they had built them at some point, my guess is they’re likely out of date, failing, or don’t work consistently. Why? Because the company typically can’t allocate a full-time team to connectors, so they make roadmap tradeoffs to update them and/or steal time from engineers on other projects to work on them in their “free time.”
Unfortunately, the costs above, whether you believe them or not, aren’t the full cost of these connectors. Let’s talk about readiness, time to market, and cost of doing business. Let me paint a picture, and you tell me if it’s a familiar story at your company.
You’re the head of product. Your CRO or head of sales comes to you and says, “We have a million-dollar opportunity, but we need to launch the campaign next week, and the client requires that we’re connected to their [activation platform of choice] to run it.”
Your response:
- You say no: the company takes a $1,000,000 loss.
- You say yes: what product tradeoff are you making? Which other customer(s) did you upset or lose? How much further behind on innovation are you falling?
Now multiply that by however many times this happens a year. Add in the reputational cost of being the vendor who always says “we can”t connect to that yet.” That’s not a one-time bill; that’s a subscription. If you’re going to pay for a subscription, shouldn’t it be 5x-10x less expensive than doing it yourself?
Connectivity isn’t a nice-to-have. It’s infrastructure, and in the age of AI, it’s critical infrastructure. So, are you in the business of running your own data center, or are you building your solution on top of an enterprise-grade cloud? Same question, different layer of the stack.
Anyone can build a connector. Few can build connectivity that truly scales. Candid insights from MadConnect CEO Bob Walczak
