schedule 12-min read

Build vs Buy: Should an MSP Build Its Own DMARC Platform?

DMARC AI editorial team · Last updated

Building a DMARC report parser is easy. Running a multi-tenant DMARC platform is not. A cost model, comparison matrix and decision tree for MSPs.

01

Introduction

An MSP should build its own DMARC platform only when DMARC software is meant to become part of its product. Writing a parser for aggregate reports takes days; running a reliable multi-tenant reporting and enforcement service for every client domain takes permanent engineering capacity. For most MSPs selling DMARC as a managed service, a platform plus their own integrations is the better use of engineering time.

This article is published by DMARC AI, a DMARC platform provider for MSPs. The evaluation criteria are designed to be useful when comparing our platform, another vendor, open-source tooling, or an internal build.

The question usually arrives the same way. An engineer parses a few aggregate (rua) reports with a script, the output looks useful, and someone asks why the MSP would pay for a platform at all. This article separates the part that is easy from the part that is expensive (the protocol itself is covered in the DMARC guide), and gives you a cost model you can fill in with your own numbers. If you are looking at vendors rather than at building, the checklist for comparing DMARC tools for MSPs is the better starting point.

02

Key takeaways

  • A report parser is a small project. A platform that ingests, verifies, deduplicates, stores, explains and alerts across hundreds of client domains is an ongoing operation.
  • The 2026 standards (RFC 9989, RFC 9990 and RFC 9991) are a recent example of the maintenance a build commits you to.
  • Building is defensible when DMARC tooling is your product, you have specific residency or integration needs, and you can staff security engineering long term.
  • Buying is usually right when DMARC is a service line and your engineers are better spent on onboarding, integrations and client delivery.
  • Keep the exit open either way: own your DNS, your reporting address and your exports.
03

What are the five realistic options?

There are five ways an MSP can handle DMARC data, and most practices move through several of them as they grow:

  1. Manual processing. Someone opens the XML attachments and reads them. Workable for one or two domains, unworkable beyond that.
  2. General-purpose AI. An engineer pastes a report into an AI chat and asks what it means. Useful for learning and one-off questions; whether AI can analyse DMARC reports safely covers where that stops.
  3. Open-source or self-hosted parser. A community parser and dashboard on your own infrastructure. Good visibility for a technical team; multi-tenancy, client reporting and workflow are left to you.
  4. Fully custom build. Your own ingestion, storage, analysis, tenant model and client output.
  5. Commercial multi-tenant platform. A vendor runs the platform; you run the service on top of it.

The honest short answer: options 3 and 4 make sense when the software itself is strategic to you. When the goal is a profitable managed service, the effort usually belongs in the service layer, which is where running DMARC as a managed service for clients is won or lost.

04

What does a DMARC platform actually have to do?

A production DMARC platform has to do far more than parse XML. Each component below is something you would design, build, test and then keep running:

Component What it involves
Report intake Mailboxes that accept reports from any sender, at volume, without losing messages
External report authorisation Publishing and maintaining the _report._dmarc records that let receivers send reports for client domains to your address (RFC 9990, Section 4)
Decoding MIME parsing, ZIP and GZIP attachments, malformed archives, oversized files
Standards compliance The RFC 9990 schema, plus the older RFC 7489 format some receivers will keep sending for a while
Validation Reports with missing fields, wrong date ranges, unknown values or invalid XML
Deduplication The same report delivered twice, overlapping reporting windows, resent reports
Storage and history Months of per-domain data, retention rules, trend queries that stay fast
Sender identification Mapping source IPs and DKIM domains to the services that send for each client
Alignment analysis SPF, DKIM and DMARC results per sender, relaxed versus strict alignment
Indirect mail context Forwarding, mailing lists, security gateways and ARC, which all produce failures that are not abuse
Alerting Detecting a new sender, a sudden failure rate, a record change or a missing report feed
Enforcement workflow Moving each domain from p=none to p=quarantine to p=reject with evidence and sign-off
Multi-tenancy Tenant isolation, role-based access control (RBAC) and audit logs
Client output White-label reports and summaries a client can read
Integrations Ticketing, finance and export routes into the rest of your stack
Privacy and residency Data minimisation, retention, location of storage, handling of failure reports
Operations Monitoring, incident response, backups, security patching and on-call

Most internal projects reach the first four rows quickly. The remaining rows are where the time goes, and they never finish: receivers change their output, standards change, and clients add senders every month.

Why do the 2026 standards matter for a build?

They are a concrete example of standards maintenance. In May 2026, RFC 9989 replaced RFC 7489 as the core DMARC specification, and RFC 9990 and RFC 9991 now define aggregate and failure reporting. RFC 9989 marks the pct, rf and ri tags as historic, adds the np, psd and t tags, and replaces the Public Suffix List lookup with a DNS tree walk to find the organisational domain. An internal platform has to absorb every one of those changes, and keep handling old-style records and reports at the same time. The RFC 9990 explainer for MSPs covers what changed in reporting.

05

How do the options compare?

Here is a qualitative comparison. It deliberately avoids prices and hour counts, because those depend on your rates, your stack and your clients:

Criterion General AI only Open-source / self-hosted Fully custom build Commercial MSP platform
Initial effort Very low Moderate: deploy, configure, secure High Low: onboarding and DNS changes
Recurring maintenance None, but repeated manual work Moderate to high: upgrades, parser fixes, hosting High and permanent Low for the platform; service work remains
Multi-tenancy None Usually limited or none Whatever you build Built in
Auditability Poor: chats are not a record Depends on your logging Whatever you build Usually built in; verify in a trial
Client reporting Manual copy and paste Mostly manual Whatever you build Built in, often white-label
Operational risk High at scale: missed reports, inconsistent answers Moderate: you own uptime and security Moderate to high: you own everything Shared with the vendor; assess their security
Customisation High per question, none as a system High, if you code it Total Limited to the vendor's options and integrations
Time to value Immediate for one report Weeks Months Days
10 domains Possible, tedious Workable Hard to justify Workable
50 domains Not workable Workable with dedicated time Hard to justify Workable
100 domains Not workable Strain on multi-client workflow Justifiable only if strategic Workable
500 domains Not workable Needs significant engineering Justifiable if it is your product Workable; check pricing at this scale

The matrix shows the pattern most MSPs find in practice: open source is a sound way to learn and to run a handful of domains, and a custom build becomes rational only at the point where the software is a product in its own right.

06

How do you calculate the total cost?

Use three formulas and fill them in with your own figures. The point is to compare like with like, including the costs that internal builds tend to leave out:

Year-one build cost = architecture + development + testing + security review
                    + deployment + operational documentation

Annual run cost     = hosting + storage + monitoring + maintenance
                    + support + standards updates + opportunity cost

Platform alternative = active domains x platform price per domain
                     + onboarding and remediation labour

The inputs worth making explicit:

  • Engineering rate. Use the loaded cost of the engineers who would build it, not their salary alone.
  • Opportunity cost. The billable or project work those engineers would otherwise do. This is usually the largest line and the one most often left out.
  • Standards updates. Budget a recurring allowance, not zero. The 2026 change shows why.
  • Support. Who answers when a client's report feed stops at 2 a.m.? That person's time is a cost.
  • Onboarding and remediation labour. This is the same in both scenarios: someone has to find every sender and fix SPF and DKIM. Do not count it only against the platform.

Then run the formulas at 10, 50, 100 and 500 active domains. The build cost is mostly fixed, so it looks worse at 10 domains and better at 500. Platform pricing is per domain, so it scales with the portfolio. For reference, DMARC AI's general MSP pricing starts from €1 per domain per month (£1 in the UK), volume-based, with the per-domain price falling as you add domains. Other vendors price per domain, per message volume or per tier, so compare at your real domain count, not at the entry price. The economics of what you charge clients are a separate question, covered in how to price DMARC services as an MSP.

07

When does building your own DMARC platform make sense?

Building, or running self-hosted software, is a defensible choice in these situations:

  • DMARC software is your product or strategic IP. You intend to sell the tooling itself, or it is central to how you differentiate.
  • You have specific integration or data residency requirements that no vendor meets, for example a sovereign hosting obligation from a public sector client.
  • You have security engineering capacity, not just development capacity, to harden an internet-facing mail intake and a multi-tenant data store.
  • You accept permanent maintenance of parsers, receivers' quirks and standards changes.
  • Your volume and economics justify it when you run the cost model honestly, opportunity cost included.

If three or more of these are true, a build deserves a serious proposal. If only the last one is, re-run the numbers with the opportunity cost included.

08

When is a platform the better choice?

A platform is usually the better choice when:

  • DMARC is a service you sell, not software you sell.
  • Time to market matters, for example because clients are hitting Google, Yahoo or Microsoft sender requirements now.
  • You need multi-tenancy, client reporting and auditability from day one, not after a development roadmap. See why multi-tenant DMARC management matters for MSPs.
  • Your engineering time is worth more in integrations, onboarding and service delivery than in maintaining an ingestion pipeline.
  • You want predictable per-domain economics that map cleanly onto what you bill clients.
09

Should an MSP build integrations instead?

Often, yes. The most valuable custom work for an MSP is usually the layer that connects DMARC to its own operations: ticket routing, client onboarding checklists, change records for DNS edits and billing. That work is specific to your business, so no vendor will do it for you, and it compounds. Check which integration routes a platform supports before you plan this. In DMARC AI's case, Autotask is supported through DMARC AI notification emails and finance through Application Link.

10

A short decision tree

Work through these questions in order:

  1. Is DMARC software part of what you sell? If yes, evaluate a build or a self-hosted base seriously. If no, go to 2.
  2. Do you have a residency or integration requirement no vendor meets? If yes, scope a build or self-hosted deployment for that requirement only. If no, go to 3.
  3. Do you manage more than a handful of client domains, or plan to? If yes, use a multi-tenant platform and build integrations around it. If no, a self-hosted parser or a platform's trial are both reasonable while you learn.

Whichever way you go, compare candidates with the same criteria. The DMARC tool comparison checklist lists them. If you want to test the platform route, the quickest check is to start a free trial with one real client domain and compare what you see with what your scripts show. A demo with a realistic client domain answers the multi-tenant and reporting questions faster than a feature list.

11

FAQ

Can an MSP build its own DMARC report parser?

Yes. Parsing RFC 9990 aggregate reports is a well-defined task: decode the attachment, read the XML and store the rows. The hard part is everything around it: reliable intake, deduplication, sender identification, alerts, tenant isolation and keeping up with receivers and standards for years.

Is open-source DMARC reporting suitable for multiple clients?

It can be, with work. Most open-source tools focus on parsing and dashboards for one organisation. Serving many clients adds tenant separation, access control, client-facing reports and workflow, which you would build and maintain yourself. Many MSPs use open source to learn, then move clients to a multi-tenant platform.

At what scale does a commercial platform become more economical?

There is no universal break-even point, because it depends on your engineering costs and the platform's pricing. Run the three formulas above at your current and expected domain counts. Include opportunity cost and standards updates; leaving those out is the most common reason an internal build looks cheaper than it is.

How do we avoid vendor lock-in?

Keep control of the parts that are expensive to move. Your clients' DNS stays with you or the client, so records can change. Consider pointing rua at an address on a domain you control and forwarding it to the platform, so a vendor switch does not require editing every client's DMARC record. Confirm the platform lets you export data.

What data and configuration must remain exportable?

At minimum: the list of domains and their current records, each tenant's sender inventory (which services send for which domain), historical aggregate data or summaries, alert rules and enforcement history. Those are the records you need to keep running the service elsewhere and to show a client what changed and why.

Should an MSP build integrations instead of the reporting platform?

In most cases, yes. Integrations with your ticketing, documentation and billing are specific to your business and give you an operational advantage. A reporting platform is general-purpose infrastructure. Building the specific part and buying the general part is usually the better trade.

12

Final thoughts

The build versus buy question is rarely about whether your team can write a parser. It is about whether you want to run a mail-processing, multi-tenant data platform for as long as you sell DMARC. If the software is your product, build it properly. If the service is your product, put your engineering into the service and keep your exit options open.

Related articles

Related tools

Ready to Implement?

Get authenticated mail moving in minutes: start free, book a guided demo, or talk to the team about your stack.