Apexnode TechEnterprise Wealth Tech
Security

Eight SIEM Buying Mistakes Banks Make, From Runaway Log Bills to Unread Alerts

Most regrets about SIEM for banks trace back to eight decisions made before contract signature. Here's each mistake, the fix, and the question to put to every vendor on your shortlist.

By the Apexnode Tech editorial team · November 2026 · 8 min read

Pricing the SIEM on This Year's Log Volume

The most expensive mistake when buying a SIEM for banks is sizing the contract on the log volume you have today. Ingest-based pricing looks predictable in the proposal. Then you add EDR telemetry, cloud audit logs, a new digital channel and a second data center, and daily volume can double within 18 months.

Here's a hypothetical. A regional bank ingests 200 GB a day. At an effective $2.50 to $6 per GB, a typical 2026 market range for analytics-tier ingest after discounts (it varies widely by contract and region, so confirm with vendors), that's roughly $180,000 to $440,000 a year. If volume grows 40% a year, year three costs about twice what year one did, and the budget line was set in year one.

Advertisement

What to do instead: Model three years of volume by source, including systems already on your roadmap. Get pricing for each volume band in writing, plus a rule for what happens on days you go over.

Ask the vendor: "If we go over our commitment by 30% for a quarter, what exactly do we pay, and can we true down at renewal?"

Paying Analytics Rates for Logs You Only Keep for Auditors

Banks keep many logs because a regulator or the card schemes require it, not because anyone will query them in real time. Paying premium analytics rates for firewall allow logs or verbose DNS data is how SIEM pricing for financial services runs away from you.

Most major platforms now separate hot analytics storage from cheaper archive or data-lake tiers. How they do it matters more than list price in a Microsoft Sentinel vs Splunk evaluation.

FactorMicrosoft SentinelSplunk (Cloud Platform / Enterprise Security)
Main pricing modelPer-GB ingest, with commitment tiers that discount at higher volumesIngest-based or workload-based (compute) licensing
Low-cost tierLower-cost log tiers and a data lake for long retention, with limits on how you can query themArchive and federated search options; cost depends on licensing model
Strongest fitBanks already heavily invested in Microsoft 365, Entra ID and AzureMixed estates, heavy on-premises footprints, mature search-driven SOC teams
Watch forSeparate charges for retention, queries and automationCompute sizing under workload pricing, and premium app add-ons

Google Security Operations, Elastic, CrowdStrike Falcon Next-Gen SIEM and Securonix price differently again. Some use fixed tiers or other non-ingest metrics, which can suit high-volume banks.

Ask the vendor: "Which of our log sources would you put in the low-cost tier, and what can't we do with them once they're there?"

Assuming One Retention Policy Fits Every Jurisdiction

Log retention requirements for banks differ by country, regulator and data type. A group running a single global retention setting is probably breaking a local rule or overpaying everywhere else. Residency is the bigger trap: some rules decide where logs live, not just how long you keep them.

Regime (examples)What it generally expects
PCI DSS (global, card data environments)At least 12 months of audit log history, with the most recent three months immediately available for analysis
CERT-In directions (India)ICT system logs kept on a rolling 180-day basis and maintained within Indian jurisdiction
DORA (EU)Documented logging and retention procedures tied to ICT risk management; firms must be able to justify the periods they choose
MAS TRM (Singapore), SAMA CSF (Saudi Arabia), central bank cyber rules in Brazil and elsewhereLogging, monitoring and retention controls, and sometimes local hosting or outsourcing conditions

On top of these, local record-keeping laws, data-protection laws such as GDPR, and supervisors' own expectations often push retention to five years or longer for particular records. Have compliance counsel in each jurisdiction confirm the periods that apply to you.

Ask the vendor: "Can we set retention and storage region per log source and per legal entity, and show an examiner proof of both?"

Buying a Detection Library Nobody Will Tune

Vendors love to cite how many out-of-the-box rules they ship. That number tells you almost nothing. Untuned, those rules fire on your scheduled batch jobs, service accounts and vulnerability scanners, and within weeks the team stops reading the queue.

What to do instead: Pick the 25 to 40 detections that map to threats you actually face: account takeover, privileged access abuse, changes to payment templates, SWIFT operator anomalies, and mass data exports from the core. Name an owner for each and review it quarterly. Track false-positive rate per rule, not just total alert count.

Good security analytics for banks comes from a small set of rules someone maintains. A big catalog nobody owns won't get you there.

Ask the vendor: "After go-live, who tunes detections, how many hours a month does that include, and what happens when we run out of hours?"

Leaving Core Banking and Payment Systems Out of Scope

Many SIEM rollouts cover Windows, firewalls and cloud consoles well, then quietly drop the core banking platform, the card management system, the SWIFT interface or the instant payments gateway. Those are exactly where insider fraud and payment manipulation show up. Teams skip them because they need custom parsers and the application team won't commit the time.

What to do instead: Make the business-critical sources dated milestones in the statement of work. Name them: Temenos, Finastra, FIS, Oracle FLEXCUBE, Infosys Finacle, your payment hub, SWIFT Alliance Access. Get each application owner to agree which events matter: user and limit changes, beneficiary additions, manual overrides, and after-hours batch edits.

Ask the vendor: "Do you have a maintained parser for our core and payment platforms at our version, or will we be building and maintaining it ourselves?"

Running the Proof of Concept on Clean Lab Data

A proof of concept fed sample data proves the dashboard works. It doesn't show how the platform copes with your messy reality: mainframe logs, odd timestamps, high-volume network devices, and a branch network sending syslog in three different formats.

What to do instead: Run a four-to-six-week POC on production log copies from at least one core or payment source. Score each vendor on parsing accuracy, how fast queries return over 90 days of data, and how many hours your own staff spent.

Ask the vendor: "Will the engineers running our POC be the ones who deliver the implementation?"

Budgeting for Software but Not for the People Reading Alerts

A SIEM doesn't watch itself. Round-the-clock alert coverage typically takes somewhere around five to eight analysts once you account for shifts, leave and turnover. Many mid-size banks can't hire or keep that team, especially in tight labor markets such as the Gulf, Singapore and London.

What to do instead: Decide early whether to build the SOC, outsource to a managed detection and response (MDR) provider, or run a hybrid. Indicative 2026 MDR pricing for a mid-size bank often falls between $150,000 and $600,000 a year, depending on scope and region. Confirm with providers, and check each is acceptable under your local outsourcing rules.

Ask the vendor: "What's the median time from alert to human triage for your financial services customers, and is it written into the contract?"

Signing Without an Exit Plan for Your Data and Detections

SIEM contracts commonly run three to five years, and moving off a platform is painful. If your detections are written in a proprietary query language and your archived logs can only be read inside the vendor's tier, you'll have little leverage at renewal.

What to do instead: Negotiate bulk export of all your data in open formats at no extra charge, plus read access to archives for a set period after termination. Keep detection logic documented in a vendor-neutral format such as Sigma where you can. Your exit plan also feeds the third-party risk register that DORA and many other regimes expect.

Ask the vendor: "If we leave, how do we get seven years of archived logs out, in what format, and at what cost?"

Before you issue the RFP, send all eight questions to every vendor on the shortlist. If a vendor won't put its answers in writing, drop it.

Questions buyers ask

How much does a SIEM cost for a mid-size bank?

In typical 2026 market ranges, a bank ingesting 100 to 500 GB a day might pay roughly $150,000 to $1.2 million a year in licensing. Managed monitoring, implementation and storage come on top. Pricing varies with contract terms, region and how much data goes into low-cost tiers, so get quotes based on your own volume forecast.

Is Microsoft Sentinel or Splunk better for banks?

Neither wins everywhere. Sentinel tends to fit banks already standardized on Microsoft identity and Azure, because integrations and some data sources are cheaper. Splunk often suits banks with large on-premises estates and experienced search-driven SOC teams. Run a proof of concept on your own core and payment logs before deciding.

How long do banks need to keep security logs?

It depends on jurisdiction and data type. PCI DSS expects at least 12 months of audit history for card environments. India's CERT-In directions call for 180 days of logs held in India. Many local record-keeping rules push certain records to five years or more. Have compliance counsel in each country where you operate confirm your obligations.

Can a bank store SIEM data in a public cloud region outside its home country?

Sometimes, but not always. Some regulators require local hosting or prior notification for material outsourcing, and data-protection laws can restrict cross-border transfers. Check your regulator's outsourcing guidance and consult legal counsel before choosing a storage region.

Should a bank build its own SOC or use an MDR provider?

Smaller and mid-size banks usually do better with MDR or a hybrid model, because 24/7 coverage takes a sizable analyst team. Large banks with stable security teams often keep detection engineering in-house and outsource overflow work. Either way, contract for measured triage times and regular access to the provider's analysts.

Written by the Apexnode Tech editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.