When a Real SIEM (or Managed Detection) Is Worth It
The do-it-yourself approach in this pillar works. Six sources, a dozen rules, a small kit of scripts, and a weekly hour of attention will catch most of what matters on a small network. It will do so for the cost of a virtual machine and your time. But "your time" is the phrase to watch. There is a point at which the lite version stops being a clever economy and becomes a slow leak. At that point, the hours you spend joining logs by hand would buy a tool that does it in seconds. The things you are not catching overnight are worth more than the subscription you are avoiding.
This guide is about recognizing that point honestly. Too early means buying a platform you will never tune. Too late means discovering an intrusion from an outside phone call. It describes the signs you have outgrown DIY and explains the three ways to move up in plain language. It lists what to prepare before you talk to anyone. And it spends real time on the most common failure of all: the expensive tool that nobody owns, nobody tunes, and nobody looks at.
It is written for the person who has been doing the lite version and is starting to feel its edges. If that is you, the good news is that you are far better prepared to buy than most first-time buyers. You already know your sources, your rules, and what normal looks like. Those are the things a platform cannot ship in the box.
In this guide
- The signs you have outgrown DIY
- The three options, in plain language
- What to prepare before buying
- Avoiding shelfware
- Questions to ask any vendor
- The decision table
- Where to go next
The signs you have outgrown DIY
None of these alone means you must buy something. Two or three of them together, on a typical week rather than a bad one, mean the conversation is overdue.
Hours per week joining logs
The lite version budgets about an hour a week for reviewing alerts and checking sources. You may be spending that hour every day: running pulls, chasing addresses across DHCP leases, re-checking a rule that keeps misfiring. If so, correlation has become a job rather than a task. Count the hours for a month. If the total would pay for a modest platform, the platform is cheaper, and it will also do the work at 3 AM.
Missed detections found later
The most uncomfortable sign. A colleague mentions that their account did something odd last month. You pull the logs, and there it is. It is a pattern one of your rules should have caught, or one you had not thought to write. One such discovery is a tuning lesson. A second, within a few months, is evidence that your dozen rules and weekly review are not enough coverage for what your network is now facing.
Compliance asking for reports you cannot produce
An auditor, an insurer, or a customer's security questionnaire asks for something specific. It might be a monthly summary of privileged-access events. It might be proof that logs from a given source have been retained for a stated period. Or it might be a report of every alert and how each was resolved. You can answer from text files and scripts, but each answer costs an afternoon, and the requests are becoming quarterly. Platforms produce these reports as a matter of course. The value is not the report. It is the afternoon.
More sources than your scripts can handle
Six sources is comfortable for a kit of normalizers. The number may pass ten or twelve: a second firewall, a cloud identity provider, a hosted mail platform, a second office. Each new source adds a normalizer to maintain and a new set of quirks. The merged files grow past the point where a search returns in seconds. The moment you find yourself not adding a source you know you need because of the effort, the kit has become the constraint.
Nobody looking at alerts at night
This one is a fact about staffing, not tooling, and no self-hosted platform fixes it. If a critical alert fires at 2 AM on Saturday and the first human sees it Monday morning, whoever caused it had fifty-four hours to work. For some organizations that is an acceptable risk and they say so in writing. For others, it is the single strongest argument for the managed option below. Those organizations hold data whose loss would be an existential event. What you are buying there is not software; it is the person who is awake.
Tip: keep a short log of the DIY pain itself. Note each time a pull took too long, a rule misfired again, a report ate an afternoon, or a detection was found late. Write one line with the date and the cost in hours. After three months, that file is the business case, and it is far more persuasive than any vendor's slide.
A five-question self-assessment
Answer each honestly for a typical month, not your worst one.
- Did log correlation take more than four hours in total?
- Did you find any event after the fact that a rule should have caught?
- Did anyone outside IT ask for a log-based report you could not produce the same day?
- Is there a source you know you need but have not added because of the effort?
- Would a critical alert on Saturday night wait until Monday to be seen?
Zero or one "yes": stay with the lite version and keep tuning. Two or three: start preparing the material below and begin conversations. You have a few months. Four or five: the cost of waiting is now higher than the cost of moving. The question is only which option, not whether.
The three options, in plain language
Beyond DIY there are three broad shapes. Each does the five jobs — collect, normalize, correlate, alert, retain. But they differ sharply in where the effort goes, what the bill is tied to, and who is watching.
A self-hosted platform
Software you install on your own servers, whether commercial or open source. You get a real correlation engine and a search interface that returns in seconds over months of data. You get built-in parsers for common sources, a rule library, and reporting. What you also get is a new system to run. It needs sizing, patching, backups, and storage that grows with your log volume. It needs someone who learns its rule language well enough to tune it. It is the most flexible option and the most effort-heavy one. It suits teams that have at least one person with real time to own it. These teams want to keep logs entirely on premises. They are comfortable being their own support desk.
A cloud SIEM service
The same jobs, hosted by someone else. You send your logs to the provider — usually through a small forwarder on your network — and the platform lives in their data center. Upkeep drops sharply: no servers, no patching, no storage planning, and parsers are maintained by the provider. The bill is typically tied to how much log data you send and how long they keep it. That makes your source list and retention window directly financial questions. A chatty firewall allow log can cost more than every other source combined. It suits teams that want the capability without the second job. These teams can accept their logs living off site and are disciplined about what they send.
Managed detection and response
Often shortened to MDR. Here you are buying people as well as software. A provider runs the platform, writes and tunes the rules, and staffs a team that watches the alerts around the clock. When something real happens, that team contacts you. Or, by prior agreement, it takes containment actions such as isolating a host. This is the only option that answers the "nobody at night" sign. It is also the most expensive per month and the one where you hand over the most control. The provider's analysts will not know your network as well as you do. And the quality of the service depends heavily on how well you brief them. It suits teams that cannot staff after-hours coverage. These teams hold data they cannot afford to lose to an overnight event. They are prepared to treat the provider as a colleague rather than a vendor.
A fourth shape is worth naming because it is common in practice. Keep the lite version for what it does well and add one of the above only for the gap that hurts. Some teams keep their own rules and kit for daytime investigation and add managed detection purely for overnight coverage. Others move only the high-volume sources to a cloud service for search and retention. They continue to run the dozen rules themselves. Mixing is legitimate as long as one person owns the whole picture.
Whichever shape you choose, budget for a cost that never appears on the quote: the first ninety days. Every option needs your sources pointed at it, your rules ported or re-tuned, and your exclusions re-learned. Your team needs to be taught to read a new interface. During that period you are running the old and the new side by side. Plan for it to consume more of your time than the lite version did, not less. Do not schedule it to overlap with a migration or an audit. Avoid the one month of the year when half the team is on leave. The effort curve drops sharply after that, but only if the ninety days were actually spent.
What to prepare before buying
A first-time buyer walks into a vendor conversation with a vague sense of need and walks out with a quote sized to the vendor's imagination. You can do better, because the lite version has already produced everything you need. Bring the following, written down:
- Your source list. The inventory from the logs that matter most: every source, how it is collected, and — critically — its daily volume, measured from your collector. Volume is what cloud pricing and self-hosted sizing are both built on. Having the real number prevents both over-buying and a nasty surprise three months in.
- Your retention needs. How long each source must be kept, and why: which requirement, which contract, which insurer. Separate what must be searchable quickly from what merely has to exist somewhere restorable; the two are priced very differently.
- The rules you already run. Your dozen detections from the starter set, with the tuning notes and exclusions. A platform's built-in library will overlap with most of them. And your tuning notes are exactly what turns those built-ins from noise into signal on day one.
- A worked incident timeline. One real or rehearsed investigation from the correlation guide, saved with its commands and output. Ask every vendor to show you how their product would reproduce it. This single test tells you more than any feature list.
- Who will own it. A name, and a number of hours per week that name has. If the honest answer is "nobody, really," you are not shopping for a self-hosted platform. In that case, you are shopping for managed detection, or you are not ready to buy yet.
- What "response" means for you. Name who may be called at 3 AM, and in what order. Also state what a provider is permitted to do without asking — isolate a workstation, disable an account, block an address. Decide this before signing, because the contract will ask.
Avoiding shelfware
Shelfware is the tool that was bought, installed, briefly admired, and then left to send unread alerts to a shared mailbox until the renewal invoice arrives. Security platforms are unusually prone to it, because the box does nothing useful until someone teaches it what normal looks like. And the teaching is the part that feels optional. It is not. A SIEM that nobody tunes is an expensive way to store logs. And it is measurably worse than the lite version it replaced, because the lite version at least had a person reading it every week.
The defenses are unglamorous. Ownership before purchase: a named person with protected time, written into the project, before the contract is signed. Start from your own rules, not the library: switch on your dozen first, tuned with the notes you brought. And add built-in rules one or two at a time as you understand their noise. Turning on the whole library on day one is how the shared mailbox fills up. Keep the weekly hour: the same fixed slot the lite version needed. Read the week's alerts, check every source is still sending, adjust one rule. The platform automates the joining; it does not automate the attention. Measure the boring number: what fraction of alerts were acknowledged by a human within a day. If it drifts toward zero, the tool is becoming shelfware, and you will see it months before the renewal. Retire rules: any rule that has fired fifty times without a single real finding is a rule to fix or delete. Do not keep it for completeness.
For managed detection, shelfware takes a different form: a provider that nobody briefs. The analysts watching your alerts know only what you told them. If your onboarding call was the last time anyone described your network, they are tuning against a picture that is months stale. Schedule a short review with them every quarter, send them your change log, and treat their escalations as conversations rather than tickets.
Questions to ask any vendor
These are deliberately generic, and any serious provider of any of the three options should answer them plainly. Evasive answers to the first four are a warning.
- Here is our source list with daily volumes. What will this cost per month, and what happens to the bill if volume doubles?
- Here is our incident timeline. Show us, live, how your product or team would reconstruct it from these sources.
- Which of our sources do you parse out of the box, and what happens when one of those vendors changes its log format?
- What retention is included, what is searchable versus archived, and how do we get our own logs back out if we leave?
- How many hours per week does a customer of our size typically spend on tuning and upkeep after the first three months?
- For managed detection: who exactly is watching at 3 AM? How do they contact us? What are they permitted to do before they reach a human on our side?
- How are alerts prioritized, and can we see the actual rule logic rather than a description of it?
- Can we run our own rules alongside yours, and will you help us port the ones we already have?
- What does a typical first ninety days look like, and what do you need from us to make them go well?
The conversation may be running beyond what one person can reasonably evaluate. The requirements may be regulatory, the stakes high, or you may simply need a second opinion on a contract. That is a normal moment to bring in outside help. The guide to when to get help covers how to do so without losing control of the decision.
The decision table
| Your situation | Stay with lite | Self-hosted platform | Cloud SIEM service | Managed detection |
|---|---|---|---|---|
| Under ten sources, one person, hour a week is enough | Yes | Premature | Premature | Only for after-hours cover |
| Correlation eating hours every day | No | Good fit if someone can own it | Good fit, less upkeep | Good fit, highest cost |
| Detections being found late | Tune first; then move | Helps if tuned | Helps if tuned | Strongest, someone else tunes |
| Compliance reports demanded quarterly | Painful | Good | Good | Good, often included |
| Twelve-plus sources, multiple sites | Breaking | Good, needs sizing | Good, watch volume cost | Good |
| Nobody available overnight | Accept and document, or move | Does not solve it | Does not solve it | The only option that does |
| Logs must stay on premises | Yes | Yes | Usually no | Depends on provider |
| Nobody can own a new system | Yes, honestly | Will become shelfware | Risky | Best fit, if budget allows |
Read the table by rows, not columns: find the two or three rows that describe you, and see which column keeps saying yes. If the answer is "stay with lite," that is a legitimate outcome and not a failure. In that case, write down why, revisit in six months, and keep the pain log going. If the answer points at a platform, you are walking in with a source list, real volumes, tuned rules, a worked timeline, and a named owner. That puts you ahead of most buyers before the first meeting.
Where to go next
Whichever column wins, the work you did in the lite version is not wasted. It is the foundation the next step is built on. The five guides in Security Log Monitoring on a Budget (SIEM-lite) fit together as a path. The guide to what a SIEM does gives you the vocabulary. The source list and detections give you the substance. The correlation guide gives you the method. And this one tells you when the method has done its job and it is time for something bigger.
Put this into practice with SysPeep Monitor
Whatever you choose for security logs, the servers, firewall, and collector behind them still need someone watching their health. That part does not require a security budget. SysPeep Monitor discovers the devices on your network. It tracks uptime and performance over SNMP and WMI. It alerts you when a log source goes down or a disk fills. All this comes from one console on a single Windows machine. The free trial installs in minutes.
Download Free TrialLearn About SysPeep