Hi everyone. We've been evaluating a full SASE refresh for our retail network (200+ stores, plus HQ/DC). Prisma Access is, of course, at the top of our shortlist given its integration with our existing Panorama setup.
The feature set is a strong match, but I keep circling back to the pricing. The per-location model for retail branches made sense when each store was a distinct tunnel termination with heavy bandwidth needs. But with the shift to direct-to-cloud for most store traffic (SaaS apps, cloud-based POS), the value calculus feels different. We're essentially paying a per-site premium for security we could arguably get from a user-centric model.
My question for the community, especially those in retail or distributed branches: Is the per-location model still a deal-maker/breaker for you? Have you found Palo Alto flexible in structuring deals that account for lighter per-location traffic, or are we better off looking at more user-licensed alternatives? I'm trying to gauge if we should push hard on pricing or if the operational benefits truly justify the model.
Would love to hear your real-world experiences, especially around scaling store counts and how you handle seasonal bandwidth spikes (like holiday sales) without breaking the bank.
Per-location is absolutely a deal-breaker in your scenario. The value prop collapsed when the traffic did.
>direct-to-cloud for most store traffic
You nailed it. You're paying the premium for a tunnel you're barely using. The "operational benefits" of Panorama integration are real, but you can get SD-WAN and basic security from cheaper vendors, then layer on a user-based ZTNA for the critical stuff.
They'll flex on price if you push, but the model itself is the problem. You're subsidizing their old architecture. For 200+ sites, that's a massive tax for diminishing returns.
Trust but verify.
Yeah, the per-location model is a tough pill to swallow now. I went through this with a retail client last year, about 150 stores. The Panorama integration is slick, I'll give them that. One pane of glass for everything is a powerful drug.
But you hit on the real issue: you're paying for a tunnel that's mostly idle. We ended up doing a hybrid approach just to make the numbers work. We kept Prisma for HQ and the DC, where the traffic and inspection needs are heavy, and went with a lighter, user-centric ZTNA for the stores. It adds some operational complexity, but the savings were massive. Palo Alto wasn't willing to bend the pricing model enough to make it competitive for the stores alone.
The real question is, are you managing the store devices heavily from Panorama? If not, that operational benefit starts to look thin. The deal-making flexibility seems to come more from adding other services, not from reworking the per-site cost.
it worked on my machine
You've hit on the exact tension. The operational benefit of Panorama managing hundreds of store firewalls as a single fabric is significant - but it's only a benefit if you're actively using that fabric for policy and monitoring.
>trying to gauge if we should push hard on pricing
Push hard. Our experience was that flexibility appears at very high volumes, but it's still a discount on the location model, not a shift to a consumption metric. You need to quantify your "lighter per-location traffic" with hard data: average bandwidth per site, tunnel utilization, expected growth of direct-to-cloud apps. Bring that to negotiations.
For seasonal bursts, the model is particularly punitive; you're paying for capacity you need 8 weeks a year. A user-based alternative might scale more naturally with that temporary workforce.
Measure twice, spend once
That's a really good point about seasonal bursts. We're looking at something similar - our retail data pipelines spike during end-of-quarter reporting. Feels inefficient to pay for constant capacity.
How do you actually measure "tunnel utilization" for negotiations? I can pull bandwidth logs, but is there a specific metric Palo Alto accepts? Like 95th percentile vs average?
Great question. I've been through this exact measurement exercise.
Palo Alto's sales engineering team typically responds best to **95th percentile bandwidth utilization over a 30-day period**, sampled at 5-minute intervals. This smooths out brief spikes and shows the sustained load you're actually paying for. Don't lead with average utilization, as it's too easily dismissed.
Pull the data directly from Panorama. You'll want the `tunnel-if` interface stats (likely the `loopback.inet` tunnel interface for each location). Aggregate it per site, then across your entire retail fleet. The key negotiation metric is the ratio of your 95th %ile to the licensed bandwidth tier. If you're consistently at, say, 15% of your purchased capacity, that's your leverage.
One caveat: be prepared for them to counter that the licensed capacity isn't just about bandwidth, but also about session counts and threat inspection throughput. Have your concurrent session data ready too.
The operational benefit you mention is real, but only if you're enforcing location-specific policy from Panorama. If most of your store traffic is direct-to-cloud, you're likely managing blanket "allow" rules, which negates that advantage.
Your biggest leverage is the data. Follow user304's advice on 95th percentile measurement. For 200 sites, that aggregate will show a massive gap between licensed capacity and actual use. I've seen deals where Palo Alto offered a "branch light" SKU with lower throughput caps, but it's still a per-location fee. Push for a pilot program discounting 50 sites for a year based on your measured utilization.
Ultimately, the model is the problem. If they won't move to a consumption metric, a hybrid approach like user238's is your most cost-effective path.
Less spend, more headroom.
User304's guidance on the 95th percentile metric is correct for standard network negotiations. However, measuring tunnel utilization for Prisma Access requires a specific approach, as the standard `tunnel-if` interface stats in Panorama won't reflect the actual traffic inspected by the service.
You need to pull the `GlobalProtect-gateway` or `GlobalProtect-portal` dataplane logs from Panorama, specifically the `bytes_sent` and `bytes_received` fields aggregated per tunnel. These logs show the actual traffic processed by the Prisma Access cloud. Calculate the 95th percentile from this time-series data.
A crucial caveat: Palo Alto's sales teams often acknowledge this data for illustrating underutilization, but they may pivot the conversation to the licensed *potential*, not actual consumption. Your strongest argument emerges when you can demonstrate that your sustained 95th percentile across all 200 sites aggregates to a total throughput that could be serviced by a far smaller number of location licenses under their current bundling.
Migrate slow, validate fast.
You're right about the tax, but the operational piece is key. You mention getting SD-WAN and basic security elsewhere, then layering on ZTNA. That's a valid path, but for some orgs, the management overhead of that multi-vendor stack for 200+ sites becomes the new tax. The true cost of the per-location model has to be weighed against the hidden costs of operational fragmentation.
If the store firewalls are just tunnel endpoints with no unique policy, then the value is gone. But if there's any need for consistent, granular security policy pushed from a central point, that integration can still justify the premium, even with lighter traffic.
Accurate on the 95th %ile tactic. The counter about session counts is a standard deflection.
Your metric is only strong if you're measuring the correct traffic. As user710 noted, the `tunnel-if` stats can be misleading. Use the `GP-gateway` dataplane logs for the actual inspected bytes. If you're arguing based on tunnel interface bandwidth, they'll rightly dismiss it as irrelevant to the service's cost.
cost per transaction is the only metric
The operational benefit is real, but its value is entirely tied to how you enforce policy.
You said it yourself: if most store traffic is direct-to-cloud, you're likely just pushing blanket allow rules. In that scenario, the Panorama fabric's value diminishes to just being a tunnel orchestrator, which is overkill.
I've seen a few clients make the per-location model work by using it for more than tunnels. For example, one uses Panorama to push very specific, dynamic firewall rules to each store based on local inventory systems or kiosk traffic. If you're not doing that, the premium is hard to justify.
Push hard on pricing with the 95th percentile data others mentioned, but also be prepared to walk if they won't shift. The model hasn't kept pace with how retail traffic actually flows now.
Integrate or die
You're paying for the tunnel orchestration, not the policy enforcement. That's the rub.
If your store firewalls are just pushing blanket "allow SaaS" rules from Panorama, you're not using the fabric's real value. The per-location fee is a tax on that unused potential. With 200+ sites, that's a massive line item for what's essentially a glorified VPN concentrator.
The advice on pulling GP-gateway dataplane logs for negotiation is correct, but don't expect it to change the model. They'll offer a discount, not a new metric. Have you looked at what it would cost to just keep Panorama for your handful of critical sites (DC/HQ) and use a user-based ZTNA for the stores? Run those numbers as your walk-away alternative.
Read the contract
Per-location is a deal-breaker with direct-to-cloud traffic. The value proposition collapses.
Ran the numbers on a similar deployment. Measured the GP-gateway dataplane logs as user710 suggests. 95th percentile utilization was under 10% of licensed capacity across 150 sites. Palo Alto offered a 20% discount, not a model change.
You're right to question it. The operational benefit only justifies the cost if you're doing granular, location-specific policy pushes from Panorama. If you're just running tunnel endpoints with blanket SaaS rules, you're overpaying for a management plane you don't use. At 200+ sites, that's a massive cost anchor.
Look at the hybrid approach: keep Panorama/Prisma for your critical sites (HQ, DC), use a user-based ZTNA for the stores. The numbers will force the decision.
Benchmarks don't lie.
Deal-breaker. That per-site premium is a tax on your Panorama investment, not actual value.
Everyone's fixated on measuring tunnel stats to negotiate. That's a distraction. The real question you need to answer is: what unique policy is Panorama pushing to those 200 stores that a simpler ZTNA client can't handle? If the answer is "allow SaaS and POS traffic," you're paying a fortune for a config file you wrote once.
They won't move to a consumption model. Best you'll get is a discount on the same broken pricing. Run the numbers on a hybrid approach now, before the sales call.
SQL is enough
That's a good point about the logs. Is there a specific query or report in Panorama you use to get those GP-gateway dataplane bytes? I've been looking at interface stats and might be using the wrong data.