Build Log #1: Why We Built PowerPoll (Because Everything Else Sucks)

Every department that touches the data center has its own tool, its own spreadsheet, its own version of the truth — and none of them talk to each other. PowerPoll is the first platform built from a single data pipeline that serves all of them. This is the story of why.

The Tuesday Night — Five People, One Problem, Zero Shared Truth

The Director of Operations is in the NOC at 11 PM. Again. The CFO wants to know why revenue is down 4% when occupancy is up, and she needs the answer by morning. She's staring at PDU readings that don't match the billing system. She knows the gap is real — she can feel $40K in annual leakage hiding somewhere between Ubersmith and reality — but proving it means cross-referencing three tools that speak different languages. She's been here before. She'll be here again.

The billing manager spent all day reconciling Ubersmith invoices against PDU readings in an Excel spreadsheet. He found three customers with unbilled overages totaling $840/month — and that's just the ones he got to today. There are 200 more cabinets he hasn't audited this quarter. He does the math in his head: if the error rate holds, they're leaking north of $5,000 a month. He'll never finish the audit because next week he'll be doing it again.

The sales director lost a deal this afternoon. A prospect asked a simple question: "What's your available power capacity on Floor 2?" It took 45 minutes to get an answer — three different people checking three different systems, arguing about whether a held cabinet should count as committed or available. The prospect went with the competitor who could answer in real time. That deal was worth $180K over three years. Gone, because nobody could answer a basic inventory question without a conference call.

The account manager has a customer threatening to churn. They asked for a real-time power dashboard — their other colo gives them a live portal where they can see their own consumption. The best she could offer was a monthly PDF, manually generated, usually delivered a week late. The customer's $96K annual contract is up for renewal in 60 days, and she has nothing to counter with.

The NOC tech got an alert that a CRAC was running hot — 45 minutes after customers had already called because their servers were throttling. His monitoring tool saw the temperature creep. Nobody else's did. Operations didn't know. Account management didn't know. The customers knew before the facility did, and now they're questioning whether the SLA means anything.

They're all looking at fragments of the same data. The same SNMP poll that could tell Operations about capacity could also tell Billing what to charge, tell Sales what's available, tell Account Management what to show the customer, and tell the CFO whether the facility is profitable. But nobody built a system that does all of that. So instead: five departments, five tools, five spreadsheets, zero shared truth.

This wasn't a bad night. This was a Tuesday.

We've Tried Everything. Everything Was Built for Someone Else.

Between us, we've collectively spent about 20 years in data center operations. We've worked at mid-market colos, enterprise facilities, and managed hosting shops ranging from 50 racks to 5,000. Along the way, we've used damn near every tool on the market — and built a few of our own out of desperation.

Here's what we learned: every tool on the market was built for one department. The problem is cross-departmental.

Ubersmith — Billing's Tool

If you've run a colo, you know Ubersmith. It's the billing and provisioning platform that most colocation operators end up on — built in the early 2000s, and the UI hasn't changed much since. To be fair, it handles billing and provisioning decently. Customer contracts, rate schedules, invoicing — it works. It's not pretty, but it works. Finance loves it. Kinda.

Here's the problem: Ubersmith has zero real-time power monitoring. Zero. It knows what a customer's contract says. It has no idea what their PDUs are actually reading. Operations has never logged in. The NOC tech doesn't have an account. Sales can't check available capacity. Account management can't show a customer their own consumption. Ubersmith is Billing's tool, and everyone else is invisible to it.

So you end up with Ubersmith for billing, a separate DCIM or homegrown tool for monitoring, and — here's the part that should make you angry — an Excel spreadsheet to reconcile between them. Three systems. Two humans. One prayer that the numbers match at the end of the month.

Every colo operator has a love-hate relationship with Ubersmith. It's not bad enough to justify a six-figure migration to something else. It's not good enough to stop the bleeding. It's the operational equivalent of a leaky faucet — you know it's costing you money, you just can't quite justify ripping out the plumbing.

Sunbird dcTrack — Operations' Tool

$33,000 a year for 100 racks. And it still doesn't generate an invoice. Let that sink in. You pay thirty-three thousand dollars annually for a tool that monitors your power beautifully — gorgeous dashboards, great visualizations — and then you alt-tab to Excel to actually bill for it.

Sunbird was built for enterprise IT departments that manage their own infrastructure. They don't bill tenants. They don't care about rate schedules. They just want to know if Rack J-14 has enough power for another 2U server. Operations loves it. Billing has never seen it. Sales can't pull available capacity from it. Account management can't share dashboards with customers. For colocation, Sunbird solves maybe 40% of the problem and costs more than most mid-market colos spend on their entire NOC staff.

Nlyte — The "Do Everything" Enterprise Promise

$100,000+ to deploy. Eighteen months to implement. By the time the consultants finish their "discovery phase" and the platform is finally configured, half your rack assignments have changed, two customers have churned, and you've hired a new technician who wasn't part of the training sessions.

We watched one facility go through a full Nlyte deployment. Eighteen months from contract to "production." The operations manager who championed the purchase had moved to a different company by the time it went live. The person who inherited it used it for asset tracking and nothing else. A hundred grand for a fancy spreadsheet.

Nlyte promises to serve everyone. In practice, it becomes one person's tool that everyone else avoids. The billing team still uses Ubersmith. Sales still asks Operations for capacity numbers. Account management still sends monthly PDFs. The fragmentation doesn't go away — it just gets more expensive.

Device42 — IT/Asset Management's Tool

Device42 is what mid-market shops land on when Nlyte is too expensive and OpenDCIM is too janky. And honestly? It's pretty good at what it does — asset tracking, dependency mapping, IPAM. The API-first architecture is genuinely nice. If you need to know what's in Rack B-12 and what depends on it, Device42 delivers.

But billing? "Here's our API. Go build something." So the IT team has a great asset management tool. Billing still has Ubersmith. Sales still doesn't know what's available. Account management still can't show customers anything useful. Device42 is better than OpenDCIM and cheaper than Nlyte, but it still leaves the hardest problem — the cross-departmental problem — as an exercise for the reader.

Schneider / Vertiv / Eaton — Engineering's Tools

Schneider EcoStruxure works beautifully if you're an all-APC shop. Vertiv Trellis (now LIFE) works great if you're all Vertiv. Eaton Brightlayer is fine if you're all Eaton. In those single-vendor universes, the integration is tight, the monitoring is solid, the alerting works. Engineering and facilities love them.

But God help you if you have a mixed environment. And every real colo has a mixed environment. You've got the APC PDUs that came with the building, the ServerTech units your biggest customer insisted on, the couple of Raritan CX2s that someone ordered by mistake three years ago, and that one weird Eaton UPS in the back room.

These are walled gardens. They serve engineering and facilities — and only with their own hardware. Billing doesn't touch them. Sales can't query them. Account management can't share them. And the moment you need to integrate a PDU from a different vendor, you're writing custom scripts or going without. Vendor lock-in isn't a monitoring strategy — it's a fantasy.

OpenDCIM / NetBox — The "We Can't Afford Real Tools" Tools

We love NetBox. It's a fantastic tool for what it is — network documentation and IPAM. But it's not a DCIM in any operational sense. It doesn't monitor power in real time. It doesn't generate alerts. It definitely doesn't generate invoices. It doesn't serve any department's actual workflow — it documents what other tools should be managing.

OpenDCIM is the same story with a different coat of paint. A wiki with rack diagrams. Documentation systems pretending to be DCIM. Useful for the engineer who maintains them. Invisible to everyone else.

The Pattern

Every tool was built for one department. The problem is cross-departmental. Ubersmith serves Billing. Sunbird serves Operations. Device42 serves IT. Schneider serves Engineering. NetBox serves the network team. And nobody serves the Director who needs all five to work together.

The Problem Nobody Talks About

Here's the thing that took us way too long to articulate: the real problem isn't that individual tools suck. Some of them are actually decent at their narrow job. Ubersmith is fine at invoicing. Sunbird is genuinely good at power monitoring. Device42's asset tracking is solid.

The problem is that colocation is inherently cross-departmental. The same PDU reading that tells Operations about capacity also tells Billing what to charge, tells Sales what's available to sell, tells Account Management what to show the customer, and tells the CFO whether the facility is profitable. One SNMP poll, five departments, five different questions — and no existing tool was architected to answer all of them from a single data pipeline.

So every facility builds a Frankenstein stack. Ubersmith for billing. Sunbird or a homegrown tool for monitoring. Excel for reconciliation. A custom script for capacity reports. A manually generated PDF for customer dashboards. And prayers that the numbers match at the end of the month.

The fragmentation isn't a minor annoyance. It's a direct line to revenue leakage (billing can't see what PDUs read), lost deals (sales can't answer capacity questions), customer churn (account management can't offer live dashboards), and operational failures (alerts don't reach the right people). Every one of those Tuesday Night scenes we described? That's a fragmentation problem wearing a different department's badge.

The Breaking Point

The real breaking point wasn't one night. It was the accumulation of a thousand small failures — each one happening because the right data existed in the wrong department's tool:

• The time Billing discovered a customer had been 4kW over their commitment for five months because the billing spreadsheet referenced the wrong PDU. Total unbilled overage: $2,100. Operations knew the PDU reading. Billing didn't. Nobody connected the dots.
• The time a NOC tech read "18.3" off a PDU display and wrote "13.8" on his clipboard. That number flowed into Billing's reconciliation spreadsheet. Nobody caught it until the customer called to dispute a credit they shouldn't have received.
• The time Sales lost a $40K/year customer because Account Management asked for a real-time power dashboard and the best anyone could offer was "we'll email you a PDF every month." The customer's other colo — the one that got the deal — had a live portal.
• The time the DCIM alerted Operations that a CRAC was running hot — 45 minutes after customers had already called the NOC because their servers were throttling. The monitoring tool had the data. The phone system had the complaints. Nobody had both.
• The time Billing spent an entire weekend reconciling Ubersmith invoices against PDU readings because someone in Operations had updated the rate schedule in one system and not the other, breaking three formulas in the reconciliation spreadsheet nobody tested.

Every one of these was preventable. Not with better training or better processes — with a single data pipeline that serves every department from one source of truth.

So We Built PowerPoll

Not because we wanted to be a software company. Honestly, the last thing any of us wanted was to write code. We wanted to run data centers. But we were tired of watching revenue leak through the gap between tools, tired of losing deals because nobody could answer a simple question, and tired of watching the industry pretend that the only options are $100K enterprise platforms or duct-tape-and-spreadsheets.

Here's what we're building — and who it serves:

Monitoring → Billing pipeline. Operations gets real-time visibility into every PDU, every circuit, every watt. Billing gets invoice-ready numbers — no more reconciliation spreadsheets, no more "what does the PDU actually say?" phone calls. Finance gets accurate revenue reporting, automatically. From the same SNMP poll. One pipeline, not three tools. The same data that tells the NOC tech a circuit is at 80% capacity generates the overage line item on the customer's invoice and feeds the CFO's revenue dashboard. One poll. Three departments served.

Customer dashboards. Account managers get a branded portal they can share with customers — live power consumption, historical trends, SLA metrics. No more monthly PDFs. No more "I'll get back to you with those numbers." Sales gets a demo tool that shows prospects exactly what their experience will look like. "What does your customer portal look like?" stops being an awkward question and starts being a competitive advantage.

Capacity planning. Sales knows available inventory in real time — power, space, cooling — without calling three people and waiting 45 minutes. Operations knows when to plan expansion before they're scrambling. Executives get utilization metrics that actually match financial reporting. One capacity model, three departments, zero conference calls.

Remote hands auto-invoicing. NOC techs log work, the system auto-generates invoice line items. Billing doesn't chase timesheets. Finance gets accurate service revenue. No more "oh, I forgot to log that four-hour cabling job from last Tuesday." The work gets done, the invoice appears. Two departments connected by a workflow instead of a prayer.

ML anomaly detection + cost optimization. Operations gets early warnings: "CRAC-3's return air temp has been creeping up 0.3°F per week for the last month — check the compressor before it fails on a Saturday night." Finance gets waste identification: "Your facility is spending approximately $4,800/month on ghost devices and cooling inefficiencies. Here's the list." Directors get actionable savings numbers they can take to the board. One ML pipeline, three different value propositions.

Discovery that works with real hardware. If it speaks SNMP, we'll find it, profile it, and start pulling data. APC, ServerTech, Raritan, Liebert — whatever's in your facility. Engineers don't have to worry about vendor lock-in. Operations doesn't have to exclude half the floor from monitoring because the PDUs are the wrong brand. No walled gardens. Your facility, your hardware, our problem to integrate.

Who This Is For

The fragmentation problem exists at every scale. Enterprise facilities just throw more money and more bodies at it — they don't solve fragmentation, they budget for it. PowerPoll is for anyone who refuses to accept that as a strategy:

• Mid-market colos (50–500 racks) — the sweet spot. Too big for spreadsheets, too sensible for six-figure DCIM contracts. This is where the pain is sharpest and the ROI is fastest. Five departments trying to share data through Excel and hope.
• Enterprise facilities tired of paying Nlyte or Sunbird prices for tools that serve one department while the other four figure it out on their own. You shouldn't need board approval and an 18-month implementation to give your billing team the same data your operations team already has.
• Managed hosting providers who need power monitoring and billing in the same system — because having them in two systems connected by API calls and optimism is how you leak $5K/month and never notice.
• Multi-site operators who need a single pane of glass across facilities — one data pipeline, one set of dashboards, one version of the truth — without deploying a separate DCIM instance at each location and reconciling between them.

PowerPoll delivers what enterprise DCIM promises — unified visibility across every department — at a price and deployment speed that doesn't require a capital expenditure committee. Built for operators from 50 racks to 5,000.

What This Blog Is For

This blog documents the build. All of it. The wins. The failures. The architectural decisions that seemed brilliant at midnight and looked questionable at 9 AM. The moments where we discovered something that made us genuinely angry at the existing tools — the "why the hell didn't anyone build a single pipeline for this" revelations.

We're writing this for the people who live in the gap between departments. The Directors who are still in the NOC at 11 PM because the CFO needs answers that live in three different systems. The billing managers reconciling spreadsheets that should be automated. The sales directors losing deals over questions that should take seconds. The account managers sending monthly PDFs to customers who expect live portals. There is a better way. We're building it.

We'll share the technical details — and we already have. Check out Build Log #2: Automating the Billing Workflow From Hell for the invoicing pipeline, Build Log #3: Auto-Discovering 41 Devices in 90 Seconds for our SNMP discovery engine, and Build Log #4: Teaching an ML Model What 'Normal' Looks Like for the anomaly detection system. Not because we're showing off, but because we've been on the receiving end of enough "proprietary secret sauce" marketing to last a lifetime. If we're going to ask you to trust us with your facility's data, the least we can do is show you how the sausage gets made.

The Bottom Line

The data center industry doesn't have a monitoring problem. It doesn't have a billing problem. It has a fragmentation problem. Five departments, five tools, five spreadsheets, zero shared truth. The same PDU reading that Operations needs, Billing needs, Sales needs, Account Management needs, and the CFO needs — trapped in five different systems that don't talk to each other.

We built the one platform that fixes that.