Home › Topics › How to Choose a Network Monitoring Tool › Needs vs. Demo Glitter

Separating Needs from Demo Glitter


Forty minutes into the demo, the presenter zooms into an animated map of a data centre you do not own. Traffic pulses along the links. A red glow spreads across a rack as a simulated fault cascades and is neatly diagnosed by something with "intelligent" in its name. It is genuinely impressive. It is also beside the point, because you came with a question. Will this thing tell you, reliably and without shouting, when the warehouse switch goes down at two in the morning?


This guide is about telling the two apart. Glitter is the set of features that demonstrate beautifully in an hour. Needs are the features you will open every working day for years. They overlap less than anyone in the room would like, and the demo format is built to show you the first kind. You will learn the standard glitter catalogue and what each item is worth. You will also learn the unglamorous features that decide whether a tool earns its keep. You will learn a fifteen-minute exercise called the day-ninety test, and how to steer a demo instead of being carried along by it. You will also learn the red flags that should make you slow down.


In this guide

  • What glitter is, and why it works on you
  • The glitter catalogue
  • The daily-use features
  • The day-ninety test
  • Demo choreography, and how to steer it
  • Red flags
  • The glitter-versus-need table
  • Where to go next

What glitter is, and why it works on you


Glitter is not fake, and it is not a trick. An animated map is a real feature that some organisations use every day. The problem is the format, not the feature. A demo is short, visual, and driven by the presenter. It naturally favours things that are quick to show, look good on a projector, and do not depend on your data. The boring parts of a monitoring tool are invisible in an hour and only reveal themselves over months. Those are whether the poller keeps polling, whether the alert rules are sane, and whether an upgrade goes smoothly. Nobody demos them because nobody can.


It works on you because your brain rewards novelty and visual motion, and because you are a professional who appreciates clever engineering. You are not being foolish when the map impresses you. You are being human, in a room designed for it. The defence is not cynicism. It is a short list of what you actually need, held in your hand. It is also a habit of asking "when would I open this?" about everything on the screen.


The glitter catalogue


These five show up in nearly every demo. Each has a real use somewhere; the question is whether that somewhere is your network.


Animated maps

Topology maps with live links, traffic animation, and glowing faults. They are good for one thing: explaining an outage to somebody who does not know the network, which happens occasionally and matters when it does. They are less good as a daily tool, because on a healthy network the map is a picture of nothing happening. On a broken network you want the list of what is broken, not a puzzle to find it in. A map that is easy to build and keep accurate is a modest plus. A map that took a professional-services engineer a week to lay out for the demo is a preview of your own week.


Features with intelligence in the name

Anomaly detection, automatic root cause, self-learning baselines, and their relatives. Some of this is useful, particularly baselining, which is simply learning what "normal" looks like so that "abnormal" can be flagged. The demo version is trained on the vendor's data and shows a tidy diagnosis of a staged fault. Ask instead what it does on day one with your data and how long it takes to learn. Ask how often it is wrong and what it looks like when it is wrong. The guide on translating monitoring marketing terms decodes the vocabulary. Here the point is simpler. If the help desk cannot explain what the feature is saying, they will stop reading it, and then stop trusting the tool.


Hundreds of integrations

A scrolling wall of logos. The number is the glitter; the two or three you will actually use are the need. Ask to see the ticketing integration create a ticket with the fields your ticketing system requires. Ask to see the chat integration post an alert that a sleepy on-call person could act on. If those two work well, the other hundreds are pleasant. If they do not, the other hundreds are irrelevant.


Mobile apps

A phone app with a dashboard looks great in the demo and is opened, in practice, mainly to acknowledge an alert. That is a legitimate need, and worth checking. Can you acknowledge from the notification? Does it show enough context to decide whether to get up? Does it work over your VPN? The rest of the app — graphs, maps, reports on a five-inch screen — is rarely used by anyone with a laptop nearby. I have opened the graphs on my phone twice, both times to show someone at a party.


Fancy reports

Glossy executive reports with gauges and colour. Management does want a report, and that is a real need. The guide on requirements before demos suggests you ask them what they actually want to see. It is usually one page with uptime per site or service, which almost any tool can produce. The glitter is the report builder with forty templates. The need is the one report your manager will read, produced automatically, on schedule, without anyone remembering to run it.


The daily-use features


The features below never get a gasp. They are what you will judge the tool by after the first quarter. And each one comes with a way of testing it, because they cannot be assessed by watching.


  • Reliable polling. The collector asks every device the same questions on schedule, every time, and tells you when it cannot. Test it by running the trial for a couple of weeks and looking for gaps in the graphs. A gap with no explanation is the tool failing silently, which is the one failure a monitoring tool must never have.
  • Sane alerting. Warning and critical thresholds, a delay before an alert fires so a single missed ping does not page anyone, and suppression during maintenance windows. Test it by counting the alerts the trial sends in its second week, after tuning. If the number is still high, imagine it in year two.
  • Fast search. Type part of a device name or an address and land on the device in a second. You will do this thirty times a day. Test it by trying to find the odd devices, the ones with inconsistent names, from memory.
  • Dependency handling. When the core switch fails, one alert, not forty. Test it by disconnecting something during the proof of concept and counting what fires. The guides under running a fair proof of concept describe how to stage this safely.
  • Clean upgrades. A new version installs in a known amount of time, preserves your configuration, and can be rolled back. Test it by asking to see the upgrade documentation and, better, by asking a reference customer what their last upgrade was like.

Not one of these can be evaluated from a slide. Every one of them requires the tool running on your network for a while. That is why a demo, however good, can only ever be the first filter. A slide has never once had a polling gap.


The day-ninety test


The day-ninety test takes fifteen minutes and cuts through most glitter on its own. Imagine it is ninety days after the tool went live. The deployment excitement is over, the vendor's engineer has gone, the map has been admired and forgotten. It is a Tuesday morning. You arrive, you make coffee, you sit down. What do you open? Write it down, honestly, screen by screen.


For most administrators the list is short and unglamorous. It includes a list of what is currently down or warning, sorted by how long it has been that way. It includes last night's alerts, to see what fired and whether it was noise. Then come a couple of graphs for the things you are keeping an eye on this week. One graph might show the link that has been creeping toward capacity. There is also the search box, because somebody has just emailed about a device you cannot immediately place. That is the tool you are actually buying. Everything else is occasional.


Now repeat the exercise for the help desk technician and the on-call person. The help desk opens one status view and asks "is it us?" The on-call person opens nothing until their phone buzzes. Then they want the alert to tell them enough to decide whether to get out of bed. Three people, perhaps five screens between them, used hundreds of times a month. Score those five screens in every demo, and score them hard. If the presenter has not shown one of them by the halfway mark, ask for it.


Northgate Retail chose its tool in the meeting straight after the demo. That was largely on the strength of a staged fault that the intelligent feature diagnosed in four seconds. On Northgate's own network the feature spent six weeks learning what normal looked like. It then flagged the month-end stock count as an anomaly every month, with the same confidence each time. The help desk switched it off in the second quarter. The animated map was never built. The engineer who had laid out the demo's map turned out to be a professional-services line item nobody had ordered. Ninety days in, the team opened the same three screens they had opened in the old tool.


Tip: the day-ninety test has a second half. On that same Tuesday, how many alerts arrived overnight, and how many of them did you act on? If the honest imagined answer is "a dozen, and none," you are looking at a tool that will be muted. That will happen by day one hundred and twenty, whatever the map looked like.


Demo choreography, and how to steer it


A standard demo follows an arc. It starts with an overview of the company, the big dashboard, the map, and a staged fault with a tidy diagnosis. Then come the integrations wall, a report, and questions with five minutes left. It is a good arc for the presenter and a poor one for you, because everything you need to see is missing from it. The fix is to change the choreography before the demo starts, politely and in writing.


Send your scenarios ahead. If you followed the requirements guide you have a one-page summary with numbered scenarios. In that case, attach it and say you would like the demo to walk through them. Ask to drive: request that the console be shared with you for part of the session. That way, you can click, search, and get lost the way you would on your own. Set the timebox: ten minutes for the overview, the rest for your scenarios. Good presenters welcome this, because it tells them what wins the deal; presenters who resist it are telling you something too.


Then ask to see the boring parts, by name. These are the requests that separate a demo from a product:


  • Add a device by hand. Not by discovery, not from a prepared list. Type in an address and the SNMP credentials and watch what it takes to get from nothing to a graph. This is the task you will repeat for years.
  • Tune a noisy alert. Pick an alert on the screen. Ask the presenter to make it fire less often, with a delay and a higher threshold, without restarting anything. Time it.
  • Show the admin interface. Users, permissions, the polling settings, the log that says what the collector has been doing. The admin screens are where you will live when something is wrong.
  • Restore a backup. Ask how the configuration and history are backed up, and ask to see the restore procedure in the documentation. A tool that cannot be restored is a tool you will rebuild from scratch one day.
  • Walk through an upgrade. What the last one involved, how long it took, and what happened to customers who had customised things.

If a presenter genuinely cannot show one of these in the session, that is fine. In that case, ask for it to be shown in the proof of concept, and write down that it was deferred. What matters is that it was asked, and that the answer is recorded next to the score. The answer should not be replaced by the memory of the map.


One last piece of choreography belongs to you: the debrief. Book fifteen minutes immediately after the demo, before anyone checks email, and fill in the scoring sheet while the room is still warm. Split what you record into two columns — things you saw happen and things you were told. By the next morning the two will have merged into a single warm impression. And the impression will be mostly map. Note what was deferred to the proof of concept. Note which of your scenarios were skipped. And note anything that surprised you in either direction. The debrief is where the demo stops being an experience and becomes evidence.


Red flags


None of the following means a tool is bad. Each one means you should slow down and look harder, because it usually points at something the demo is built to steer you around.


  • They will not show the admin interface. "That is for our engineers" or "let us configure that for you" is a polite way of saying the administrative side is hard. You will be paying for the hours either way.
  • Everything you ask for is coming in the next release. One or two deferred items are normal in any product. If most of your scenarios are answered with the roadmap, you are evaluating a promise. Promises cannot be scored above the bottom of the evidence scale.
  • References only from organisations far larger than yours. A tool that thrives with a dedicated monitoring team may be a burden for a team of one. Ask for a reference your own size and, if none exists, treat that as data.
  • The demo only runs on their data. A prepared environment behaves the way it was prepared to. Insist on a trial on your own devices before any score goes above the demo cap.
  • Nobody can say how many hours a month it takes to run. Every vendor should know this figure from their customers. An evasive answer means it is high, or nobody asked, and neither is comforting.
  • The pricing conversation keeps sliding. Add-on modules, counting units, and renewal terms may only become clear after the decision. That is the mechanism by which a good price becomes a surprising one. Ask for the full catalogue and the counting units in writing.
  • The good parts need professional services. The map, the dependencies, or the alert rules may all have been built by an engineer for the demo. In that case, budget for the same engineer on your network, or accept that you will not have those parts.

The glitter-versus-need table


Keep this next to your scoring sheet during demos. The left column is what you will be shown. The right columns are what to ask about instead and which daily-use feature it maps to.


What you are shownWhat it usually isWhat to ask insteadDaily-use need behind it
Animated topology map with pulsing linksA prepared layout on the vendor's dataHow long to build and keep accurate on our network?Dependency handling; a clear "what is down" list
Automatic root-cause diagnosisA staged fault with a scripted answerWhat does it do on day one with our data, and how often is it wrong?Sane alerting; dependencies that suppress noise
Wall of integration logosBreadth, mostly untested by youShow the ticketing and chat integrations creating a real itemAlerts reaching the right person with context
Mobile app dashboardA phone view of the big screenCan I acknowledge from the notification over our VPN?Acknowledging an alert at 2 AM
Report builder with dozens of templatesFlexibility you will use onceProduce our one-page uptime report on a scheduleThe one report management reads
Big-screen dashboard with gaugesA good-looking defaultShow me the list of what is down right now, sorted by durationThe morning screen
Discovery finding hundreds of devices instantlyA clean lab with known credentialsAdd one odd device by hand with its own credentialsReliable polling of the devices you actually own
Features with intelligence in the nameBaselining with a marketing coatHow long does it learn, and can the help desk understand its output?Knowing what normal looks like

The right-hand column keeps landing on the same handful of things. Those things are a clear list of what is down, alerts that reach the right person without noise, and polling you can trust. They also include the ability to add and tune things yourself. That is not a coincidence. There are only a few daily needs, and every piece of glitter is an elaborate route back to one of them. Ask for the direct route.


Where to go next


Glitter is not the enemy; misplaced weight is. A map or a clever diagnosis is a pleasant bonus on top of a tool that polls reliably, alerts sanely, searches fast, handles dependencies, and upgrades cleanly. It cannot stand in for any of them, and a demo is built to make you forget the difference for an hour. Take the day-ninety list into every demo. Steer the choreography toward your scenarios and the boring parts, and note the red flags without drama. And let the scoring sheet and the proof of concept do the deciding. I still lean forward when the rack turns red; I just no longer write it down. The rest of How to Choose a Network Monitoring Tool covers requirements, cost, and the end-to-end playbook for a small team.


Put this into practice with SysPeep Monitor

The day-ninety test only works on a tool you have actually run for a while. So the candidates worth shortlisting are the ones that let you do that easily. SysPeep Monitor is one you can put through it on your own network. It provides automatic discovery, SNMP and WMI collection, and dependency-aware alerting from a single web console on a Windows machine. Its free trial is long enough to find out what you would really open every morning.

Download Free TrialLearn About SysPeep