Data Integration Solutions: An Engineer’s Shortlist for 2026

Updated May 2026. Rewritten away from a generic platform list. This is the working shortlist we use when picking data integration solutions, with what each category is for and where each one breaks.

Data integration solutions used to be a single category: an ETL tool that moved data on a schedule from operational systems into a warehouse. In 2026 the category has split into at least four sub-types, and picking the wrong one is more expensive than picking a slightly weaker tool inside the right one.

We are a small AI and automation consultancy in Brisbane and we build and maintain data integrations for clients in healthcare, recruitment, and professional services. This piece is the shortlist we work from when a team asks which data integration solutions to consider, organised by the problem each one is trying to solve rather than alphabetical platform reviews.

If you want the upstream context, our note on system integration best practices covers the design patterns. This piece is about the tools.

The Four Types of Data Integration Solutions

There are four categories that cover almost every real data integration job. Every shortlist we build starts by asking which of these you actually need, because the answer kills two-thirds of the platform list immediately.

• ELT (load first, transform later). Move raw data into a warehouse, transform it inside the warehouse. The default for analytics in 2026.
• iPaaS (event-driven workflows between SaaS apps). When the CRM updates, push a record to the billing system. Real-time, transactional, business logic.
• Reverse ETL (warehouse to operational systems). Take enriched data out of the warehouse and push it back into Salesforce, HubSpot, or Marketo where teams act on it.
• Streaming (real-time event pipelines). Kafka, Kinesis, or Pub/Sub when sub-second latency or high throughput matters and a daily batch will not do.

The old “ETL platform” category still exists, mostly inside Informatica, Talend, and the bigger MS-stack tools. We rarely recommend it for new projects in 2026. Modern warehouses are so cheap to compute against that the case for transforming data outside the warehouse has almost disappeared.

ELT Data Integration Solutions: The Default for Analytics

If your goal is “get all our data into Snowflake/BigQuery/Databricks/Redshift and run analytics on it”, you want an ELT tool. The category is mature. The tools we shortlist:

Fivetran

The market leader. Hundreds of pre-built connectors, automatic schema handling, very low maintenance. Pricing is based on Monthly Active Rows and gets expensive fast above mid-sized data volumes. For a team under 50 people with standard SaaS sources, you are looking at roughly $1,000 to $5,000 AUD per month all-in. Above that, the bill scales aggressively. The classic “just works” tool until invoicing catches up with you.

Airbyte

Open source, deployable locally, with a cloud-hosted option. The connector library is now competitive with Fivetran for popular sources, weaker for long-tail systems. Self-hosted is the right answer if you need data residency in a specific region or if your Fivetran bill has crossed $4,000 AUD per month. The trade-off is that someone on your team owns the deployment, which is one to two days a month of attention.

Stitch

Owned by Talend, often cheaper than Fivetran at small volumes, fewer connectors. We see Stitch in shops that started with it years ago and have not had a reason to migrate. Not what we usually pick for new builds in 2026.

Hevo and Matillion

Hevo is positioned between Fivetran and Stitch, with reasonable pricing and good support. Matillion is a heavier ETL tool that fits if your team prefers a visual transformation layer outside dbt. Both are credible picks if Fivetran’s pricing model does not work for you and Airbyte feels like too much operational overhead.

The hidden cost in every ELT tool: dbt. You will need a transformation layer once data lands in the warehouse. dbt Core is free, dbt Cloud starts at $100 USD per developer per month, and learning it well is two to four weeks for a competent SQL person.

iPaaS Data Integration Solutions for Event-Driven Workflows

If your goal is “when X happens in this app, do Y in that app”, you are in iPaaS territory. The tools here look superficially similar to ELT tools but solve a different problem: low-latency business automation rather than analytics.

n8n as an Integration Platform

Self-hosted, code-friendly, fair-code licensed. The tool we reach for most often for client integration work. About 500 built-in nodes, easy custom nodes when something is missing, and a clean way to handle long-running workflows. Self-hosting on a $20 to $50 AUD per month VPS covers most teams. If you want a deeper look at why, see our self-host guide.

Zapier and Make

The two big hosted iPaaS tools. Zapier has the biggest connector library and the simplest UX. Make (formerly Integromat) is more powerful and cheaper at scale. Pricing for both works out around $50 to $500 AUD per month for a small team, escalating quickly with operation counts. We use them for quick prototypes and stop using them when an integration starts handling money or compliance-sensitive data.

Workato and Tray.io

Enterprise iPaaS. Strong governance features, audit logs, deployment environments. The pricing starts at five figures annually, which puts them out of reach for most teams below 200 people. If you do have the budget and your compliance team is breathing down your neck, Workato is the safe pick.

Microsoft Power Automate

The default if your stack is heavily Microsoft and you already pay for Power Platform licences. Connector library is big, the UX is functional. We see it work well inside finance and operations teams that live in Microsoft 365. It struggles when you need to integrate non-Microsoft systems at scale.

Reverse ETL: The Newest Data Integration Category

The pattern: you have built a beautifully modelled customer table in your warehouse, with churn risk scores and lifetime value and segments. Your sales team lives in Salesforce. Reverse ETL is the layer that pushes that warehouse data back into the operational tools where someone will actually act on it.

• Census. The category-defining tool. Connectors to every CRM and ad platform, good observability. Pricing scales by destination row counts.
• Hightouch. Direct competitor to Census, often cheaper for similar use cases, very strong sync logic. The other tool we shortlist by default.
• Rivery. Bundles ELT and Reverse ETL, which is convenient if you want a single bill but means you are committing to their ELT side too.

The honest answer most teams need to hear: you probably do not need Reverse ETL yet. It becomes valuable when you have a real data team modelling customer-level metrics. If your warehouse currently has three tables and a chart of last quarter’s revenue, skip this category for now.

Streaming Data Integration: When You Actually Need It

This is the smallest category and the most over-recommended. Streaming data integration solutions like Kafka, Kinesis, and Google Pub/Sub are the right answer for high-throughput, sub-second-latency use cases: trading systems, fraud detection, real-time personalisation, IoT telemetry.

They are the wrong answer for most analytics. A team that says “we want real-time dashboards” almost always means “we want dashboards that are not 24 hours stale”. A 15-minute batch on Fivetran solves that. Kafka would cost ten times more and demand a dedicated platform engineer.

When streaming is actually right, we use:

• Confluent Cloud for hosted Kafka if the team does not have platform engineers to spare.
• AWS Kinesis if the rest of your stack is already on AWS.
• Self-hosted Kafka on Kubernetes if scale and cost justify the operations team. The pricing inflection point is around 100MB/s sustained throughput.

How We Actually Choose Among Data Integration Solutions

The decision tree we walk a client through is shorter than the platform list suggests. Four questions, in order:

1. What is the destination? If it is a warehouse, ELT. If it is another operational app, iPaaS. If both, you may need two tools.
2. What is the latency requirement? Sub-minute is streaming or iPaaS. Hourly to daily is ELT. Real-time analytics requirements are usually about a 15-minute batch dressed up in marketing language.
3. What is the data volume? Under 100M rows per month is firmly in hosted ELT territory. Above 1B rows per month, self-hosted or enterprise tier starts paying for itself.
4. What is the compliance posture? Health data, financial data, anything covered by APP or APRA in Australia, or HIPAA or GDPR elsewhere. Self-hosted or a tool with the right regional residency story.

Those four answers almost always narrow the shortlist to two tools. Then you pilot both for two weeks against the hardest source you have, and pick whichever survived contact with reality. Skipping the pilot is the single biggest mistake we see clients make.

When Not to Buy a Data Integration Tool

The most expensive integration mistakes we have helped clients fix were not tool choices. They were buying a platform when the right answer was custom code or no integration at all. A few patterns where the tool is the wrong answer:

• One-time data migrations. A scheduled job written in Python, run twice, then deleted. Fivetran is overkill. The Python file is 80 lines.
• Integrations with one obscure system. If you need to push data from a single niche legacy app and no platform has a built-in connector, you are buying a platform to solve one connector. A 200-line custom integration is cheaper.
• Integrations that should not exist. Sometimes teams want to integrate two systems because process is broken. The right fix is moving work into one system, not building a sync between two. Cheaper, more reliable, fewer late-night alerts.

For most clients, the integration platform is the right answer. But we have seen enough $30,000 platforms purchased to solve a problem a $0 process change would have killed that we now ask the “should this integration even exist?” question first. If you want to think through that decision before buying anything, book a call and we can walk through it.

Things That Have Broken for Us in Real Data Integration Builds

A few specific gotchas worth knowing before you sign a contract:

• Source API rate limits. Most managed ELT tools handle rate limiting gracefully. Most iPaaS tools assume your source can handle one request per workflow execution. Bulk operations into the wrong tool will get you throttled.
• Schema drift. The source system adds a column. The destination does not know. Fivetran handles this well. Most custom integrations do not, and you find out three days later when a dashboard quietly stops being correct.
• Soft deletes. Most CRMs do not hard-delete records. They mark them deleted. ELT tools handle this differently. We have lost half a day chasing a “why are these rows still in our warehouse” question more than once.
• Time zone handling. A surprising number of integration platforms do not normalise time zones consistently. If you are seeing 1.5-hour offsets in your data, the integration tool is doing weird things with daylight saving on top of a half-hour time zone.
• OAuth refresh tokens. The integration is fine for 90 days, then silently breaks because the refresh token expired and nobody noticed. Always set up an alert for missed runs, not just failed runs.

Regional Notes for Australian and EU Teams

If your data has residency requirements, three things to check before committing to a hosted tool: which region the data is processed in, which region it is stored in, and where the support team can see it from. Several major data integration solutions process in the US even if they store in ap-southeast-2 (Sydney) or eu-west-1. For APP or GDPR purposes, that matters. Self-hosted Airbyte or n8n in a Sydney or Frankfurt region sidesteps the question entirely.

AUD pricing on the major tools: Fivetran roughly $1,000 to $5,000 per month for small teams scaling fast. Airbyte Cloud about $300 to $1,500. Self-hosted Airbyte or n8n is mostly the underlying VPS cost, $20 to $200 per month for typical workloads.

Frequently Asked Questions About Data Integration Solutions

What is the best data integration solution for a small business?

For a team under 30 people doing basic analytics, Fivetran or Airbyte Cloud paired with a warehouse like BigQuery or Snowflake will cover almost every use case for $200 to $1,000 AUD per month. If most of your integration needs are app-to-app rather than analytics, self-hosted n8n on a small VPS is the cheapest credible option at around $30 AUD per month.

How much do data integration solutions cost in 2026?

For small teams, $200 to $2,000 AUD per month covers most ELT and iPaaS tools. For mid-sized teams with multiple sources and meaningful volume, $2,000 to $10,000 AUD per month is the realistic range. Enterprise tier with full governance, compliance, and dedicated support starts at $50,000 per year and goes up.

Should we build or buy our data integration?

Buy for the first five integrations. Once you are running 20 or more pipelines and the costs are eating into the team’s budget, self-host the open-source tools (Airbyte, n8n) on your own infrastructure. Pure custom code is rarely the right answer unless the connectors do not exist in any platform.

What is the difference between ETL and ELT?

ETL transforms data before loading it into the destination. ELT loads raw data first, then transforms it inside the destination warehouse using SQL or dbt. ELT is the modern default because warehouses are cheap to compute against and storing raw data lets you re-derive metrics later when the business definition changes.

Do I need Fivetran or can I use Airbyte?

Airbyte Cloud is competitive with Fivetran for the top 30 or so popular connectors. Fivetran is still ahead on the long tail of less common sources and on connector reliability under heavy load. If your sources are mainstream SaaS apps, Airbyte is fine. If you have weird sources or you are running pipelines critical enough that a half-day outage matters, Fivetran is worth the extra cost.

What data integration solution handles real-time best?

For true sub-second real-time, you want Kafka (via Confluent Cloud or self-hosted) or AWS Kinesis. For “near real-time” use cases (1 to 15 minutes), Fivetran and Airbyte both offer sync schedules that are good enough. Be honest about what “real-time” means in your case before paying streaming prices.

How long does a data integration project take?

For a single source-to-warehouse pipeline using a hosted tool, two to five days including testing. For a custom integration with non-standard sources, two to four weeks. For a full data platform build with five to ten sources, modelling, and dashboards, six to ten weeks for a small team.

What data integration solution supports AI and ML workloads?

Any of the ELT tools paired with a warehouse that supports vector operations (Snowflake, BigQuery, Databricks, Postgres with pgvector) will support modern AI workloads. The integration layer matters less than the warehouse and the transformation layer. dbt has good model patterns for feeding ML pipelines and is worth learning regardless of which integration tool sits upstream.

If you are picking between two or three data integration solutions and want a second opinion before committing to a year-long contract, get in touch. We have built and broken enough of these to have strong opinions on what will hold up.