I've been evaluating ZTNA solutions for my team, and we're down to two main contenders: Zscaler Private Access (ZPA) and Citrix Secure Private Access. The technical demos were solid for both, but now I'm trying to get a real handle on the total cost of ownership, which seems... opaque.
From what I can piece together, pricing models are quite different:
* **ZPA** seems to be primarily user-based licensing, often bundled with their broader Zscaler Internet Access platform. I've heard quotes are per user, per year, with potential volume discounts.
* **Citrix SPA** appears to tie into their existing workspace/VDI licensing ecosystem. If you're already a Citrix shop, the incremental cost might be lower, but the standalone entry point is less clear.
My specific context:
* ~150 developers and SREs needing access to internal Kubernetes APIs, build systems, and monitoring dashboards.
* We already use Okta for identity.
* No existing major Citrix or Zscaler footprint.
I'm less interested in list prices and more in real-world operational costs. Has anyone gone through a similar procurement?
* What were the hidden costs? (e.g., special connectors/VDAs, bandwidth egress, support tiers)
* Does the per-user model get prohibitive for contractor or CI/CD system access?
* How does the agent-based approach (both are agent-based, right?) affect resource overhead on endpoints? I'm curious if anyone has done low-level profiling.
```bash
# Not really code, but this is the kind of resource check I'd run
# on a test machine to see agent impact
$ ps aux | grep -E '(zscaler|citrix)' | awk '{print $3, $4, $11}'
```
Would love to hear from teams who have deployed either, especially at a similar scale. The architecture docs are great, but the business side feels like a black box.
System calls per second matter.
I'm Anna, and I lead platform engineering for a fintech with about 200 engineers. We replaced a legacy VPN for our microservices and internal tooling access last year, and I ran the bake-off between ZPA and Citrix SPA.
Here's my breakdown from our procurement and the 14 months we've run ZPA in production.
* **Real Pricing & Model:** ZPA is strictly user-based. For 150 users, expect $70-110 per user, per year. The quote is annual and includes their cloud connector; there were no extra fees for egress or the service edge nodes. Citrix SPA was quoted as a component add-on to their Workspace Premium at ~$16/user/month, but that required the underlying platform. The standalone "Secure Private Access" SKU started at about $12/user/month, but that was before the required "connector" VMs in our cloud, which added infrastructure and ops cost.
* **Integration & Operational Lift:** ZPA's integration with Okta was a 2-hour job. App segments are IP-based, so defining our internal Kubernetes ingress IPs and dashboard URLs was straightforward. The biggest config time sink was setting up granular access policies. Citrix required us to stand up connector VMs (we tested with 4 in AWS) which added about $350/month in compute costs and meant patching and monitoring another fleet. Their policy engine felt more tied to their legacy VDI concepts.
* **Performance & Scale Gotcha:** For API traffic (like your K8s use case), ZPA's TLS termination happens on their cloud, so your apps see traffic originating from their nodes. We had to adjust some rate-limiting. Citrix's connectors tunnel all traffic, so source IP is preserved, which is nicer for auditing. However, we found ZPA's scaling automatic and transparent, while the Citrix connectors needed manual sizing and HA configuration. Our load of ~2k RPS to internal APIs caused CPU spikes on the medium connector VMs.
* **The Hidden Support Cost:** With ZPA, you're paying for a turnkey service, and support is included but reactive. You'll file tickets for issues. With Citrix, a significant portion of the cost is the support ecosystem itself - forums, TAMs, complex KBs. If you're not already immersed in that Citrix world, there's a learning tax for your team to diagnose connector issues or policy quirks.
My pick for your specific use case of 150 developers accessing internal APIs and tools with an existing Okta foundation is **Zscaler Private Access**. The operational simplicity and predictable cost won for us. If you were heavily invested in Citrix Cloud already, or if preserving exact client source IPs for internal apps was a legal requirement, then Citrix SPA could make sense. To make the call absolutely clean, tell us: do you have a security team that demands full traffic logs reside in your cloud, and is there any existing spend or relationship with either vendor?
Clean code is not an option, it's a sanity measure.
Ah, the old "no extra fees for egress or the service edge nodes" line. That's a classic one, and it's true until it isn't. Did your procurement team actually run a projected traffic analysis for the next three years? In our case, Zscaler was happy to quote the per-user license, but the "premium support" required for any meaningful SLA and the costs associated with integrating their API for our custom workflows became a significant, recurring line item they didn't mention until the third call.
Your point about the Citrix connector VMs adding infrastructure cost is valid, but let's not pretend the operational lift for ZPA ends with Okta integration. Maintaining those app segments and policies when your internal IP ranges shift, or when you're dealing with dynamic cloud environments, becomes a quiet but constant tax. Their cloud connector is "included," but you're entirely at the mercy of their architecture decisions and update schedules.
Trust but verify.
Exactly. The "premium support" requirement is the silent killer. They all quote you the base price, but good luck getting a response inside of 48 hours without that gold tier. Suddenly your simple per-user fee has a 30% annual adder.
And you're right about the operational tax. Their included connector isn't a benefit, it's a cage. You can't tune it, you can't scale it independently, you just get to troubleshoot their black box when things are slow. At least with Citrix you own the VMs and can throw more resources at a problem. With Zscaler, you just open a ticket and wait.
Your stack is too complicated.
thanks for laying out the pricing so clearly, this is exactly the kind of info I was looking for since I'm trying to wrap my head around the same decision for a smaller team.
one thing I keep bumping into when I read comments about these tools is that "operational cost" seems to mean very different things to different people. like, for you with 150 SREs, does the time spent tuning app segments in ZPA actually show up as a real dollar cost on your end? or is it more about frustration during on-call? I'm trying to figure out if these hidden costs are real money or just annoyance.
also, the point about premium support being a silent 30% adder sounds scary. did your sales rep ever bring that up before you asked, or did you have to dig for it?
You're spot on about the support tier being a hidden line item. In our case, the 30% adder was real, but it also came with a dedicated slack channel and a named TAM, which did cut down our MTTR for ZPA-specific issues. That's the trade-off they don't advertise: you're paying to make their black box a little less opaque.
> you just open a ticket and wait
This is the real operational tax during an incident. You can't pull Grafana logs from their connector, you can't see queue depths. Your monitoring stops at the edge of your network, and your dashboard just shows "connection failed." That wait time isn't just frustration, it's prolonged outage duration.
Sleep is for the weak