"How do enterprise infrastructure teams monitor and manage observability across hybrid on-premises and cloud environments in 2026?"

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now the thing that's occupying the most time is unifying our telemetry layer across environments. We've got some workloads that aren't moving off-prem anytime soon — mission-critical stuff where the business isn't willing to take the migration risk — and then obviously a lot running in AWS. Getting coherent observability across both is harder than the vendors make it sound. The specific tension I'm sitting with is the build vs. buy question on the aggregation layer. The managed observability platforms want you fully in their ecosystem, and that works great until you have on-prem infrastructure where their agents either don't fit well or the data egress costs become a real constraint. So we end up with this patchwork — one tool doing better on the cloud side, something else covering on-prem, and the correlation between the two is manual in ways it shouldn't be. We're also dealing with what I'd call ownership ambiguity. When you've got platform engineers embedded with product teams, it's not always clear who's responsible for the observability stack as a whole versus just the slice that touches their service. That organizational question is honestly as hard as the technical one right now.

What's the most important thing you need to understand or solve here?

For us, the core problem is maintaining a coherent view across environments that weren't designed to talk to each other. We have workloads on-prem that aren't moving — some mission-critical stuff that the business won't touch — and then we have cloud-native services expanding on AWS. Getting consistent telemetry across both is harder than vendors make it sound. The secondary thing is that committing too deeply to any one provider's observability tooling creates lock-in I'm not comfortable with. If I build everything around CloudWatch or whatever managed service AWS is pushing this year, I've traded one problem for another. So we're trying to thread that needle between OSS flexibility and the operational overhead of self-managing everything.

What does 'good' look like to you — and how far are you from that today?

For us, "good" means a single pane of glass where I can see our on-prem Kubernetes clusters and our AWS workloads with the same level of fidelity — correlated traces, metrics, and logs without having to context-switch between tools. That's the north star. How far are we from it? Closer than we were two years ago, but not there. We're running Grafana as the primary UI layer with OpenTelemetry doing the collection, and that covers maybe 75% of the surface area. The gaps are mostly in our legacy on-prem services that predate our instrumentation standards — getting those to emit structured telemetry consistently is still a work in progress. The other gap is ownership. We have the tooling mostly sorted, but when something crosses the boundary between on-prem and cloud, it's still not always clear which team is holding the incident. That's more of an organizational problem than a technical one.

What would change your perspective on this entirely?

That's a fair question. If someone showed me a unified observability platform that genuinely worked across our on-prem VMware clusters and our AWS environment without requiring us to fully commit to one ecosystem, I'd pay attention. Right now the choice feels like — either you go all-in with the cloud provider's native tooling, or you self-manage a stack of OSS that requires people and effort you may not have. The other thing that would shift my thinking is ownership clarity. We've had conversations about consolidating tooling, but who actually owns the observability layer when you've got platform engineers, SREs, and dev teams all touching it? If a vendor could make that governance model less painful, that would be meaningful to me.

What question are you not being asked that you wish someone would ask?

The tooling fragmentation question, honestly — not in a generic way, but specifically around how many different agents you're running on a single host. We've got environments where we're looking at a logging agent, a metrics agent, a tracing agent, sometimes a security agent on top of that. Each one has its own config management story, its own update cadence, its own resource overhead. Nobody asks about the operational cost of the observability layer itself.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now the thing I keep coming back to is the ownership model question. We've got a mix — some legacy workloads that genuinely can't move to cloud yet, and a growing footprint in AWS — and the observability tooling that works well in one environment tends to fall apart in the other. So we end up with two or three different dashboards depending on where the signal is coming from. The practical problem is that nobody fully owns the seam between those environments. The infra team knows the on-prem side, the platform team leans cloud-native, and observability kind of lives in this uncomfortable middle. We're not in crisis, but it creates latency when something actually goes wrong. The other thing is vendor consolidation pressure. We've evaluated a few of the major observability platforms and the pitch is always "one pane of glass," but when you get into the integration specifics for our on-prem stack, it gets complicated fast. I don't have a clean answer on whether we build more of that connective tissue ourselves or just accept the lock-in that comes with committing to one vendor's ecosystem.

What's the most important thing you need to understand or solve here?

For us, the core problem is maintaining a coherent view across environments that weren't designed to talk to each other. We have workloads that are staying on-prem — not because of inertia, but because they genuinely belong there for compliance or latency reasons — and then cloud workloads on top of that. Getting consistent telemetry across both is harder than vendors make it sound. The secondary issue is tooling sprawl. Every layer of the stack wants to own observability, and you end up with five dashboards that each tell you a slightly different story. That's the thing I'm actually trying to solve right now — not the data collection part, but the normalization and correlation across environments.

What does 'good' look like to you — and how far are you from that today?

For us, "good" is a single pane of glass where I can see latency, error rates, and infrastructure health across both our on-prem data center and our AWS footprint without context-switching between four different consoles. Correlated signals, not just aggregated metrics. How far are we? Closer than two years ago, but still not there. We've got decent coverage on the cloud side — that part's relatively easy when you lean into managed services. The on-prem side is still patchy, and stitching those two together with consistent tagging and trace context is where most of our active work is right now.

What would change your perspective on this entirely?

That's a fair question. If I saw a vendor actually solve the context problem — where the observability platform understands the relationship between my on-prem Kubernetes clusters and my AWS workloads without me having to manually wire that together — that would shift my thinking. Right now I'm spending engineering time building those bridges myself, which is basically the worst of both worlds: I'm paying for the managed service *and* doing integration work that should be the vendor's problem. If someone genuinely collapsed that gap, I'd care a lot less about the build-vs-buy calculus.

What question are you not being asked that you wish someone would ask?

That's a fair question. I'd say it's something like: "How do you actually manage the ownership model when you're running observability tooling across teams that have different levels of cloud maturity?" We have groups that are deep into AWS-native services and groups still running stateful workloads on-prem that predate my tenure here. The observability tooling that works beautifully for one side creates friction for the other. Nobody asks how we govern that seam — they ask about the tools, not the organizational structure underneath them.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now the thing I'm spending the most mental energy on is the ownership question — when you have infrastructure that spans our on-prem data processing layer and the cloud side, it gets murky fast. Like, who actually owns the observability stack for a service that touches both? In our environment we have some legacy on-prem components that aren't going away anytime soon, and the teams that built the cloud-native pieces have very different tooling assumptions than the folks managing the older infrastructure. The secondary thing is just consistency of instrumentation. Getting every service to emit logs and metrics in a format that a unified dashboard can actually consume is harder in practice than it sounds in planning docs. I don't have a great answer for it yet — we're still iterating.

What's the most important thing you need to understand or solve here?

For us, the core problem is visibility across the seam between on-prem and cloud. You end up with data living in two different places — your legacy infrastructure tooling on one side, your cloud-native observability stack on the other — and stitching those together coherently is where things get messy. In our environment we're not running hybrid at that scale, but I've talked to infra teams where the ownership question is the real blocker. Like, who actually owns the unified view? Is it a platform team, is it embedded in each product team? That governance piece tends to get underestimated relative to the tooling.

What does 'good' look like to you — and how far are you from that today?

For us, "good" is when your observability setup actually shifts you from reactive to proactive — you're catching issues before users do, not scrambling after the fact. That framing resonates with how I think about it. How far are we from that? Honestly, we're closer on the cloud side than the on-prem side. The cloud tooling is more mature, easier to instrument, easier to iterate on. Where we still feel the gap is in getting a unified view across both environments — right now there's some manual correlation happening that I'd like to see automated or at least better structured. I don't have a strong view on exactly what the industry benchmark is, but for our team, the practical measure is: are on-call engineers spending their time investigating real problems or chasing noise? We've made progress there, but it's not where I'd want it to be yet.

What would change your perspective on this entirely?

That's a fair question. I think if I saw a fintech our size — not a hyperscaler, not a legacy enterprise — actually running a clean, unified observability layer across hybrid infrastructure without significant ongoing engineering overhead, that would shift things for me. Right now my mental model is that the operational cost of maintaining that consistency is high enough that most teams quietly accept gaps. If someone showed me a real ownership model that held up when the on-prem side had a bad week and the cloud side kept scaling, I'd reconsider how much of the friction is structural versus just bad tooling choices.

What question are you not being asked that you wish someone would ask?

That's a good prompt. I think the one that doesn't come up enough is around ownership clarity in hybrid environments — specifically, when something goes wrong at the seam between on-prem and cloud, who actually owns that incident? In our setup we don't have dedicated infra teams the way a large enterprise would, so it tends to fall on whoever noticed the problem, which isn't a great system. I imagine at bigger orgs with embedded platform engineers that question gets even messier. It's less about tooling and more about the organizational model, and I don't hear researchers asking about that directly.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now the thing I keep bumping into is that we're a fintech startup, so our infrastructure is predominantly cloud-native, but we do have some on-prem data processing requirements because of compliance constraints around certain transaction data. And bridging those two worlds from an observability standpoint is genuinely messy. The specific thing I'm wrestling with is ownership. Like, who actually owns the observability layer when you've got infra engineers responsible for the on-prem side and a separate cloud platform team? We don't have clean lines there, and when something breaks across that boundary, the investigation gets slow because neither team has full context. That feels like the more fundamental problem than any tooling gap we have.

What's the most important thing you need to understand or solve here?

For us, the hardest part is actually the ownership question. When you're running hybrid — some stuff on-prem, some in cloud — it's not really a tooling problem, it's a "who's responsible for what layer" problem. Like, if a latency spike shows up, is that the network team's problem, the platform team's, the app team's? That boundary gets blurry fast. From a PM perspective, I spend a lot of time thinking about how observability tooling decisions reflect and sometimes entrench those organizational boundaries. You can instrument everything perfectly and still have teams pointing fingers at each other because ownership wasn't defined upfront.

What does 'good' look like to you — and how far are you from that today?

For us, "good" is when your observability layer actually tells you something actionable before a customer notices a problem. Not just "service X is down" but enough context that whoever's on call can trace it back to a root cause without spending 45 minutes correlating logs from three different systems. How far are we from that? Closer on the cloud side than on anything legacy. Our newer services have reasonable tracing and metrics pipelines, but the moment something touches older infrastructure the visibility drops off pretty quickly. I'd say we're maybe 60-70% of the way there for our critical paths, and then it gets spottier from there.

What would change your perspective on this entirely?

For us, I think the thing that would really shift my perspective is seeing a genuinely unified data plane that works across on-prem and cloud without requiring a team of engineers to maintain the connective tissue between the two. Right now a lot of what I observe from adjacent infrastructure teams is that the tooling story is still pretty stitched together — Terraform for IaC, some combination of cloud-native monitoring, and then separate tooling for anything on-prem. If ownership models got cleaner — like, if there was a clear answer to who actually governs the observability layer when you've broken up traditional silos — I think the conversation would change substantially. That question of ownership is the one that seems to slow things down more than the tooling itself.

What question are you not being asked that you wish someone would ask?

For us, the question I'd want someone to ask is: "Who actually owns observability outcomes, not the tooling?" Because in a hybrid environment you can stand up Datadog or whatever across your on-prem and cloud layers, but if nobody has clear accountability for acting on what those tools surface, you're just collecting expensive dashboards. At a fintech startup we don't have the siloed team structure that larger enterprises deal with, but even at our scale I see the ownership question get fuzzy fast — especially at the boundary between what engineering owns and what the platform or SRE function owns. I imagine at a bigger org with dedicated infrastructure layers, that gets significantly harder to resolve.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

For us, it's less about the observability tooling itself and more about what the data actually tells us from a cost and attribution standpoint. We're a Series A company, so we're running a hybrid setup — some things still on-prem from before we fully committed to cloud, some workloads on AWS — and the honest challenge is that nobody owns the full picture. Engineering has their monitoring stack, I'm trying to understand what infrastructure costs are doing to our unit economics, and those two worlds don't talk to each other cleanly. The consumption-based pricing on some of our cloud tooling is the thing that keeps me up at night a little. It's genuinely hard to govern. Teams spin things up, usage compounds, and by the time it hits a budget conversation it's already happened. I don't have a strong view on the observability tooling layer specifically — that's more in engineering's lane. But from where I sit, the CAC and cost-per-pipeline-dollar calculations get muddier when you can't cleanly attribute infrastructure spend to specific workloads or teams.

What's the most important thing you need to understand or solve here?

For us, the biggest thing is honestly just visibility — knowing what's actually happening across the stack at any given moment. We're not a massive enterprise with 150k employees, but we're at that awkward Series A stage where we've got some on-prem legacy stuff and we're pushing workloads into AWS, and the two don't talk to each other cleanly from a monitoring standpoint. The attribution problem I deal with in marketing has a close cousin in infrastructure: you're trying to figure out what caused what, and the data you need to answer that question is scattered across different tools that weren't designed to work together. That's the core tension I see from my vantage point — it's less a technology problem and more a "who owns this and how do we govern it" problem.

What does 'good' look like to you — and how far are you from that today?

For us, "good" means having a single place where I can see pipeline influence across every channel — paid, organic, outbound — and actually trust the attribution model behind it. Right now we're pretty far from that. We've got data sitting in HubSpot, some in our ad platforms, some in Salesforce, and stitching it together into something coherent takes manual work every week. On the CAC side, I want to know cost per opportunity by segment, not just blended CAC, and understand which channels are actually converting versus which ones are generating engagement that goes nowhere. That gap between "marketing looks successful" and "sales can't close anything" is real for us in certain channels right now. So directionally I'd say we're maybe 60% of the way there — the data mostly exists, but the tooling and process to make it reliable and fast isn't quite in place yet.

What would change your perspective on this entirely?

That's a fair question. Honestly, if I started seeing attribution data that clearly tied observability tooling decisions back to pipeline velocity or deal cycle length on our end — like, if the infrastructure teams we're selling to could show us that their monitoring setup was directly influencing how fast they could evaluate and commit to new vendors — that would shift how I think about positioning this whole category. Right now from where I sit, observability feels like a cost center conversation, not a revenue conversation, and that makes it harder to build campaigns around. If someone showed me compelling evidence otherwise, I'd reconsider that framing pretty quickly.

What question are you not being asked that you wish someone would ask?

That's a fair question. For us, the thing that rarely comes up is how observability decisions made by engineering actually create downstream CAC problems for marketing. Like, if we can't instrument our trial or freemium funnel well because the infra team hasn't prioritized telemetry on those touchpoints, I'm flying blind on where people drop off. Attribution becomes a guessing game, and I'm making channel spend decisions on bad data. Nobody asks marketing leaders what they actually need from observability infrastructure. It's always framed as a DevOps or platform engineering conversation.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

For us, the observability piece isn't something I own directly — that sits with our infrastructure and DevOps folks. But I end up caring about it from a cost and attribution angle. We're on a consumption-based model with our cloud provider, and honestly governing that spend is a real pain point. Teams spin things up, tools get adopted outside of a formal procurement process, and by the time finance or I see it, the bill is already there. The hybrid piece specifically — where we've got some on-prem legacy stuff alongside cloud — the visibility across both environments feels patchy based on what I hear from our engineering lead. It's not a clean picture. I don't have a strong view on the tooling side of that, but from where I sit, the cost unpredictability that comes out of it is what actually lands on my radar.

What's the most important thing you need to understand or solve here?

For us, the core question is really about attribution and cost governance across channels — but I recognize that's my lens, not the infrastructure team's problem. If I'm trying to put myself in the shoes of the teams you're actually studying, I'd guess the hardest thing is just having a single coherent view when your environment is split between on-prem and cloud. The tooling doesn't naturally unify those two worlds, so you end up with visibility gaps or you're stitching together data from multiple sources manually. I don't have strong firsthand experience with hybrid observability specifically — my world is pipeline and demand gen, not infra. So I'd be cautious about overstating what I know there.

What does 'good' look like to you — and how far are you from that today?

For us, "good" means I can trace a dollar of pipeline back to a specific channel or campaign without spending three hours in spreadsheets reconciling UTM data against CRM records. That's the north star — clean attribution, clear CAC by channel, and confidence in where we're investing next quarter. How far are we? Honestly further than I'd like. We've got decent top-of-funnel tracking, but the moment a lead touches a sales conversation or goes through a multi-touch ABM sequence, the attribution story starts to fall apart. We know roughly what's working, but "roughly" isn't good enough when I'm making budget decisions.

What would change your perspective on this entirely?

That's a fair question. I'm not deep in the infrastructure and observability weeds day-to-day — that's more of an engineering and DevOps concern for us. But from where I sit on the demand gen side, what would shift my perspective is if I started seeing observability vendors actually prove out business impact in terms I care about — like how downtime or monitoring gaps translate into pipeline or retention risk. Right now when I'm evaluating tools or even just co-marketing with vendors in adjacent spaces, it's hard to connect their value prop to something tangible on my side of the house. If that bridge got clearer, I'd probably pay more attention to this space.

What question are you not being asked that you wish someone would ask?

That's a fair question. For us, the thing that doesn't come up enough is how observability tooling decisions made by engineering or DevOps teams actually create downstream CAC problems for go-to-market. Like, when an infrastructure team commits to a consumption-based observability platform without looping in finance or procurement, we end up with cost structures that are genuinely hard to model when I'm trying to justify pipeline spend. Consumption-based models are notoriously difficult to govern, and when cloud costs balloon unpredictably, the first budget conversation that happens is usually about marketing headcount or program spend — not the observability contract that nobody flagged. I don't have a strong view on the technical side of hybrid monitoring architectures, that's not my lane. But I do wish more research explored how those infrastructure buying decisions propagate into broader budget constraints across the org.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now my biggest challenge is that I'm sitting between two worlds — we've got some legacy on-prem infrastructure that supports certain compliance-sensitive workloads, and then the rest of our stack is spread across AWS and a bit of GCP. The observability story across those two environments is pretty fragmented. What I'm actually wrestling with is the tooling consolidation question. We've got one set of dashboards and alerting for the cloud-native side, and a separate monitoring setup for what's on-prem, and neither team really has full visibility into the other's environment. It creates blind spots when something breaks across that boundary, which happens more than I'd like. The other piece is that our infra team keeps telling me they need more headcount to manage this, but leadership's expectation is that we automate our way out of it. I'm skeptical that "intelligent automation" solves this as cleanly as vendors pitch it — the actual work of integrating these environments requires organizational change, not just a new tool. That disconnect between what the platform promises and what it actually takes to implement is where a lot of our friction lives right now.

What's the most important thing you need to understand or solve here?

For us, it's really about getting a coherent picture across environments without having to stitch it together manually. We have some workloads still on-prem, some in cloud, and the visibility gaps between them create real operational headaches — you end up with different teams owning different parts of the stack and no single source of truth. The secondary piece is making sure whatever tooling we adopt actually reduces the coordination overhead, not just adds another dashboard someone has to babysit. A lot of what I see pitched in this space promises unified observability but still requires significant configuration work to get there — and that organizational change piece is usually harder than the technology itself.

What does 'good' look like to you — and how far are you from that today?

For us, "good" means a single pane of glass where we can see signal across both our on-prem data center and whatever we're running in AWS or Azure — with enough context that the ops team isn't just reacting, they're getting ahead of issues. Correlation between layers, not just raw telemetry dumped into a dashboard. Where we are today? Probably somewhere in the middle. We've got decent coverage on the cloud side, but the on-prem piece still has gaps — some of it's legacy tooling that doesn't integrate cleanly, some of it's just organizational inertia around what gets instrumented. The hybrid piece is genuinely harder than vendors make it sound.

What would change your perspective on this entirely?

That's a fair question. Probably if we saw a genuinely unified control plane that worked across on-prem and cloud without requiring a team of engineers to maintain the integrations — that would shift how I think about the category. Right now a lot of the "hybrid observability" pitch is stitched together under the hood, and you end up owning the complexity anyway. If a vendor could actually demonstrate that at scale in a real enterprise environment, not just in a sandbox demo, I'd reconsider some of my skepticism about consolidating onto a single platform.

What question are you not being asked that you wish someone would ask?

That's a fair question. For me, it's probably something like: "How do you get infrastructure and marketing teams to agree on what 'good' observability data looks like when they're optimizing for completely different outcomes?" We're increasingly consumers of that observability data — pipeline attribution, product usage signals, campaign performance tied to infrastructure uptime — and the handoff between what the infra team tracks and what we actually need to make decisions is messier than most vendors acknowledge. I don't have a strong view on the technical architecture side, but the organizational alignment piece feels underexplored in most of these conversations.

Tell me what's top of mind for you on this topic right now — what are you wrestling with?

Right now, from where I sit in marketing, the observability conversation mostly touches me when we're trying to understand pipeline attribution across our stack — what's happening in our CRM, our MAP, our ad platforms — and whether the data we're pulling is actually reliable. That's where I feel the pain directly. But more broadly, the bigger organizational wrestling match I see is around hybrid environments. We've got workloads that haven't fully moved to cloud, and the teams managing those are dealing with two different monitoring paradigms that don't talk to each other cleanly. It's less a tools problem and more a process and data consistency problem — like, who owns the source of truth when an incident spans on-prem and cloud? I don't have strong technical depth on the infrastructure side specifically, so I'd caveat that. But from a business operations standpoint, the friction I hear about is getting coherent visibility without having to stitch together three different dashboards manually.

What's the most important thing you need to understand or solve here?

For us, the core problem is really unified visibility. When you've got workloads split across on-prem data centers and two or three cloud providers, the default is that every environment has its own native monitoring tooling, and they don't talk to each other cleanly. So you end up with your infrastructure team context-switching between dashboards and trying to manually correlate what's happening where. That's the thing we're trying to get in front of — not just "can we see everything" but "can we act on what we're seeing without needing five different logins and tribal knowledge about which system tells you what."

What does 'good' look like to you — and how far are you from that today?

For us, "good" means unified visibility — you're not toggling between three dashboards to understand whether a performance issue is on the infrastructure side, the application side, or somewhere in the network between on-prem and cloud. One pane of glass is the cliché, but the actual requirement is correlation: when something degrades, you can trace it end-to-end without manually stitching together logs from different systems. How far are we from that? Closer than we were two years ago, but not there. We've made progress consolidating some of the cloud-native tooling, but the on-prem environment still has its own monitoring stack that doesn't talk cleanly to the rest. That gap is the main friction point right now.

What would change your perspective on this entirely?

That's a fair question. If we saw a genuinely unified control plane that actually worked across on-prem and cloud without requiring a team of engineers to maintain the integrations, that would shift how I think about the space. Right now a lot of the "single pane of glass" pitches don't survive contact with a real hybrid environment. If someone could demonstrate that at scale — not in a demo, but in a reference customer with a similar architecture — I'd take a harder look. The proof point is really what's missing.

What question are you not being asked that you wish someone would ask?

That's a fair question. I'll be upfront — I'm coming at this from a marketing angle, not infrastructure or DevOps, so there's probably a lot of the technical depth here that's outside my lane. But the question I don't hear enough in these conversations is around organizational alignment — specifically, who actually owns the observability strategy when you've got infrastructure teams, DevOps, and security all with overlapping stakes. In my experience working cross-functionally, the tooling decisions often get made before anyone's figured out the ownership model, and then you end up with three teams each running their own dashboards and nobody talking to each other. That seems like a real coordination problem that doesn't get surfaced in conversations that focus purely on the technology stack.