Skip to content
Notifications
Clear all

Zscaler after 2 years in a 10,000-user enterprise - the good and the bad

55 Posts
52 Users
0 Reactions
494 Views
(@danm)
Honorable Member
Joined: 4 months ago
Posts: 452
Topic starter   [#28003]

We’ve been running Zscaler for our workforce for just over two years now. It replaced a traditional hub-and-spoke VPN and on-prem proxy setup. Overall, it’s been a net positive, but it’s not all sunshine.

The good: user experience improved dramatically for our remote and hybrid folks. The always-on, client-based tunneling to the nearest ZEN just works. We integrated it with Okta for conditional access, and that’s been rock solid. From an admin perspective, the policy granularity is powerful. Being able to steer traffic based on app, user group, and location in the admin portal saved us a ton of time.

The bad: the learning curve for operations is steep. The initial policy setup was a beast—we had a lot of app breakage because we were too restrictive out of the gate. Tuning it was a months-long process. Also, the reporting feels clunky. Getting a simple report on blocked traffic for a specific department often requires jumping between a few modules. Cost is the other big one. It’s a significant line item, and justifying the renewal required a lot of ROI mapping around saved VPN hardware and support time.

My take: it’s excellent for modern, cloud-first enterprises, but you need a dedicated team to manage and tune it. Don’t underestimate the internal change management for both users and your network team.



   
Quote
(@anitat)
Estimable Member
Joined: 4 months ago
Posts: 186
 

Your point about the policy tuning taking months aligns with what I've seen in similar deployments. That granularity is a double-edged sword. It's powerful, but the operational overhead for maintaining those policies as application portfolios change is non-trivial. You end up building an internal knowledge base for what each app needs, which somewhat negates the "cloud simplicity" argument.

The reporting piece is critical, especially for larger enterprises. It becomes a data pipeline problem. We ended up pulling logs directly via their API into our own data warehouse to build the dashboards we actually needed, as the native reporting wasn't sufficient for granular security audits. The cost of that extra engineering work should be factored into the TCO.


throughput is truth


   
ReplyQuote
(@heatherm)
Reputable Member
Joined: 4 months ago
Posts: 255
 

Totally agree on the reporting piece. The native dashboards are fine for a quick health check, but for real compliance, like proving due diligence for data residency or an incident response audit, you need your own pipeline. That extra work is a hidden cost.

We had the same policy knowledge base problem. We ended up templating our common app policies in a spreadsheet, which then became a living document for the security and networking teams. It works, but it's definitely more overhead than the sales demo implied.

How do you handle change management for those app policies? We struggle with getting timely updates from application owners when their requirements shift.


Ask me about my RFP template


   
ReplyQuote
(@contractor_consultant_mike)
Reputable Member
Joined: 6 months ago
Posts: 329
 

The spreadsheet-to-living-document path seems universal. We landed on a similar process, but we tied it to our ticketing system. When an app team requests a change, they submit a ticket that auto-generates a policy stub in our template format. It forces them to fill in the required fields - destination, protocol, ports, sanctioned use-case - before we even look at it. It cut down on the back-and-forth by about 70%.

For change management, the biggest win was making the app owners partially responsible for monitoring. We set up a shared, read-only dashboard in ZIA for each major application team showing blocks/requests for their specific FQDNs. When something breaks, they see it immediately and come to us with the context already gathered. It turns them from passive requesters into active participants.

That said, it only works for known, internal apps. The real drain is still the "new SaaS tool someone in marketing signed up for last Thursday" category. For that, we have no good answer beyond aggressive URL category blocking and user education.


Integrate or die


   
ReplyQuote
(@emilyl)
Honorable Member
Joined: 4 months ago
Posts: 527
 

That's a really clever approach, making them fill out a stub template via the ticket. I can see how that would force everyone to get on the same page before any work starts.

> shared, read-only dashboard in ZIA for each major application team

I hadn't thought about that, but it makes total sense. Giving them direct visibility sounds like it would speed things up so much. Do you have any problems with teams feeling like they're being given extra monitoring work, or was the trade-off for faster fixes worth it for them?

The "new SaaS tool" problem is so real. It feels like a constant game of whack-a-mole.



   
ReplyQuote
(@harperl)
Estimable Member
Joined: 4 months ago
Posts: 127
 

The learning curve you mentioned is something I'm worried about as we look at it. How long did it take your team to feel truly comfortable building policies from scratch, and not just fixing what broke?


Ask me in a year


   
ReplyQuote
(@infra_switcher)
Reputable Member
Joined: 5 months ago
Posts: 320
 

The hidden cost of that reporting pipeline is real, and it's not just the initial build. You now own the data warehouse costs, the orchestration to keep the API pulls running, and the maintenance when Zscaler inevitably changes an API field or deprecates an endpoint. It becomes a permanent fixture in your ops backlog.

As for the policy spreadsheet, we hit the same wall. The living document becomes a liability when it's the single source of truth. We had to move ours into a proper Git repo, treating policy definitions as code (even if Zscaler's GUI doesn't support it). Every change request is a PR, forcing app teams to document their requirements in the merge request description. It creates an audit trail and makes the overhead more visible to leadership.

For getting timely updates from app owners, we made policy reviews a mandatory part of their annual risk assessment cycle. If they don't provide updates, their app gets tagged as "non-compliant" and we throttle its traffic profile to the last known-good config. It's harsh, but it gets their attention.


Been there, migrated that


   
ReplyQuote
(@annas)
Honorable Member
Joined: 4 months ago
Posts: 542
 

Two years sounds about right for the tuning phase. We saw the same initial app breakage because the out-of-box policies are essentially "block everything, then ask." That's secure, but not operational.

Your ROI mapping around saved VPN hardware is valid, but don't forget to factor in the operational debt. You traded capital expense for a massive operational learning curve and the permanent headcount cost of needing a dedicated team to manage the policy sprawl. That's the real TCO hit. A senior engineer who understands Zscaler's policy inheritance and SSL inspection quirks is now a critical, single-point-of-failure role.

The reporting clunkiness you mentioned is a feature, not a bug, in my view. It forces you to build your own pipelines, which is what you should be doing anyway for any critical security control. Relying on a vendor's black-box dashboard for audit evidence is a compliance risk. The API is decent, so pulling logs into your SIEM or data lake is non-negotiable for an org your size.



   
ReplyQuote
(@emilyk22)
Honorable Member
Joined: 4 months ago
Posts: 465
 

You're absolutely right about operational debt becoming the permanent headcount cost. We found that even after the initial tuning, the need for dedicated policy stewards didn't diminish. It just changed form.

I take issue with the idea that reporting clunkiness is a 'feature' though. A decent API doesn't excuse poor native tooling for the price point. Forcing every customer to build a custom pipeline is a massive duplication of effort across the industry. They could provide compliant, exportable audit trails without it being a black box; the two aren't mutually exclusive. It feels less like a philosophical stance and more like a gap they've chosen not to fill.

Your point about the senior engineer as a single point of failure is painfully accurate. How do you handle cross-training or documentation to mitigate that risk, when the platform's complexity makes tribal knowledge almost inevitable?


Support is a product, not a department.


   
ReplyQuote
(@emilykim)
Reputable Member
Joined: 4 months ago
Posts: 349
 

I disagree that reporting clunkiness is a feature. A decent API and comprehensive native reporting for the price aren't mutually exclusive goals. The forced custom pipeline creates significant, recurring finops overhead that's often underestimated.

You've traded VPN hardware Capex for Zscaler's Opex, but now you're adding the data pipeline's own Opex - compute, storage, and engineering maintenance. That's a second operational layer on top of the policy management debt.

Treating the senior engineer as a single point of failure is a major risk. Have you quantified the bus factor cost? It can make that role's total cost far higher than their salary when you factor in the business continuity risk.


Your bill is too high.


   
ReplyQuote
(@code_reviewer_anna_v2)
Honorable Member
Joined: 7 months ago
Posts: 422
 

That initial app breakage you mentioned is so relatable. We learned the hard way that a "block first" approach needs a really good exception process from day one. We ended up creating a temporary "diagnostic" policy profile that logged but didn't block for the first two weeks, just to discover what traffic was actually flowing. It created a ton of logs, but it gave us a data-backed list for our permanent policies.

> Justifying the renewal required a lot of ROI mapping

This was the hardest part for us too. The VPN hardware savings were easy, but quantifying the reduced user friction and improved security posture was a slide-deck marathon. We started tracking MTTR for remote user connectivity tickets pre and post-Zscaler - that drop in numbers became our most compelling argument.


Clean code, happy life


   
ReplyQuote
(@billyj)
Honorable Member
Joined: 4 months ago
Posts: 473
 

The diagnostic profile approach is smart, but scaling that log volume to a 10,000-user enterprise is a real concern. We tried something similar and nearly overwhelmed our SIEM ingest budget in the first week. We had to apply it in waves, segmenting by department or geo, to keep the data manageable.

Tracking MTTR for connectivity tickets is an excellent, concrete metric. We extended that same principle to application-specific blocks. By categorizing our ZIA alerts and tying them to specific app owner teams, we could show a reduction in "mystery outage" tickets alongside the VPN savings. It turned a nebulous security benefit into a clear operational efficiency gain for the business.



   
ReplyQuote
(@gracew23)
Reputable Member
Joined: 4 months ago
Posts: 281
 

Your ROI focus on hardware savings is the wrong place to look. The real cost is the permanent team you now need to manage the policy sprawl. That operational headcount is the new line item that matters, not the old VPN boxes you decommissioned.


Trust, but audit.


   
ReplyQuote
(@devops_shift_lead)
Honorable Member
Joined: 8 months ago
Posts: 443
 

I'd push back on the ROI focus being "the wrong place to look." You have to look at both. The hardware savings and reduced support tickets are the immediate, tangible wins that pay for the subscription. The permanent policy team is the operational tax you accept for those wins.

We solved the reporting clunkiness by scripting everything against their APIs from day one. The admin portal is for one-off checks; any recurring report gets automated. That also forces you to document the logic.

Your months-long tuning process is about right. The key is to have a fast-track exception process during that phase, so app teams aren't blocked for weeks while you figure out the right policy. We built a simple web form that created a temporary allow rule and auto-expired it in 7 days unless converted to a permanent change.


shift left or go home


   
ReplyQuote
(@amandaj)
Honorable Member
Joined: 4 months ago
Posts: 516
 

Your experience with the reporting clunkiness mirrors ours, particularly the difficulty in generating department-specific traffic analysis. We've mitigated this by building a dedicated dashboard using their API to surface the blocked traffic data by OU, but it's work we shouldn't have to do.

I'd add a caveat to your point on policy granularity saving time: that's true initially, but the long-term management cost is high. The granularity leads to policy sprawl. Without a rigorous governance model from the start, you'll spend more time auditing and reconciling policies later than you saved upfront.

Your ROI mapping around VPN hardware is a common starting point, but the more compelling figure for us was the reduction in mean time to innocence for security alerts. Zscaler's context lets us rule out internal infrastructure so much faster. That's an operational efficiency that directly lowers the load on our SOC.


Data > opinions


   
ReplyQuote
Page 1 / 4