Merge Unified

Merge is primarily used for change data capture to identify inserts, updates, and deletes for transactions in data warehousing. This is the place where I implement Merge, which is the most efficient solution through SQL rather than using stored procedures to identify inserts and updates separately.

In one of my data warehousing projects, when we load data, it is in medallion layer form with different layers including a raw layer, staging layer, and data mart modeling layer. We generally call these the bronze, silver, and gold layers. When it comes into the gold layer, that is where we generally implement change data capture. The raw layer contains all raw data as we receive it from the source. We then bring the incremental data into the silver layer with basic data cleansing activities. In the gold layer, we perform more transformations and keep history and report-related data. When we make changes in the gold layer, this is where change data capture comes into the picture for identifying changes. When we have a huge table, we do not want to reload everything on a daily basis but instead want to identify what changes have occurred and load only that data.

We can accomplish this in multiple ways. One approach is using an ETL tool to identify the source and target, compare both, and figure out what the incremental changes are, including inserts and updates. Another approach is using a stored procedure, but in that case we do not know how many rows were updated or inserted. The third approach is using SQL statements, which is where Merge comes into the picture. This is one of the most efficient ways of doing things. When working with Snowflake, instead of ETL, ELT became a more optimized way of doing things. During this process with Snowflake, I have tried multiple times using Merge. With Merge, I am able to figure out what the inserts and updates are, capture those records, and load them into the target.

The main use case is finding data changes. There is nothing specific about using a particular project to only use Merge, but for most projects where we do CDC changes, instead of using an ETL tool, we can use native SQL and implement this solution.

We can perform inserts and updates, and at the same time when we do these changes, we can compare using hash joins. That is the main advantage of using Merge.

When I use hash joins, I create a hash key. Whenever I want to do a Merge statement with a table having 10 key combinations and 40 non-keys, I do not want to handle all that complexity. I create hash keys for the key combinations and create another hash key for the non-keys. When I join between the source and target table, I can use a hash key join to figure out what are inserts and what are updates.

One significant advantage is the ability to go with open source and not depend on proprietary ETL tools. Tools may be helpful, but they create dependency without much knowledge sharing. If we use open source solutions such as SQL, that will give more exposure and is a common technology skillset. I do not need to look for a skilled person who knows only specific ETL features. If someone has knowledge of data warehousing concepts, I can hire them with basic SQL skills and they can perform the work in SQL itself. This way, I do not need to limit my hiring of resources to a particular skill set. Open source solutions can have a broader category where I can bring in people. Additionally, tool-specific solutions are licensed products that charge millions for an enterprise, so this is also a good cost saving advantage.

Regarding optimization, I have observed that on the Snowflake side, with Merge being used on a table that has micro partitions and has been well maintained, the statement execution will be very fast. This helps in both cost savings and performance-wise optimization.

It would be beneficial if there is a common public community so that people who may not be aware of this will become more aware. Additionally, when we do joins within the Merge statement, there is always a chance of getting duplicates. Snowflake added a feature called QUALIFY that allows us to filter based on ranking and eliminate duplicate records. If such a feature were available in regular Merge statements within any database, that would be helpful. This would eliminate the need to rebuild another subquery inside the Merge statement.

I have been working in the current field for almost 13 plus years.

In one of my data warehousing projects, when we load data, it is in medallion layer form with different layers including a raw layer, staging layer, and data mart modeling layer. We generally call these the bronze, silver, and gold layers. When it comes into the gold layer, that is where we generally implement change data capture. The raw layer contains all raw data as we receive it from the source. We then bring the incremental data into the silver layer with basic data cleansing activities. In the gold layer, we perform more transformations and keep history and report-related data. When we make changes in the gold layer, this is where change data capture comes into the picture for identifying changes. When we have a huge table, we do not want to reload everything on a daily basis but instead want to identify what changes have occurred and load only that data.

We can accomplish this in multiple ways. One approach is using an ETL tool to identify the source and target, compare both, and figure out what the incremental changes are, including inserts and updates. Another approach is using a stored procedure, but in that case we do not know how many rows were updated or inserted. The third approach is using SQL statements, which is where Merge comes into the picture. This is one of the most efficient ways of doing things. When working with Snowflake, instead of ETL, ELT became a more optimized way of doing things. During this process with Snowflake, I have tried multiple times using Merge. With Merge, I am able to figure out what the inserts and updates are, capture those records, and load them into the target.

With open source solutions, we get pay-as-you-go pricing as we use the service. Additionally, when we have a major issue, we pay for the service we may need to take. This results in very different cost perspectives. We do not need to pay just for the license cost. On the cost optimization side, and from a hiring perspective, if I want a data engineer, I do not need to look for an ETL-specific pipeline developer role. In the current world with hundreds of ETLs available, I do not need to look for a specific ETL person who knows the features of that tool to develop as an expert from day one. I would either need to take an ETL developer and give them leverage of days or weeks to make them feel expertise in this particular ETL tool before starting development. Instead, if I go with an open source solution such as Snowflake or Databricks where I can use SQL or regular Python coding, I can have a very wide variety of people to hire. Anyone who comes with a basic SQL background can directly jump into the work. I do not need to look for a specific tool skill that will take longer to hire. This way I can save time in hiring people. I would rate this solution an 8 out of 10.

Hybrid Cloud

Amazon Web Services (AWS)

What do you like best about the product?

Merge's support team is always really helpful and often go above and beyond, and the product works great.

What do you dislike about the product?

There are nuances with the setup process for some integration partners that can add friction to the user experience.

What problems is the product solving and how is that benefiting you?

It helped us get integrations off the ground much faster than we could have if we built them ourselves, and it also requires less maintenance work on our end.

What do you like best about the product?

The level of support from the Merge team is exceptional, specifically our Account Manager, Jemma, who is proactive, professional, and incredibly quick in responding. The integration with Pylon for support works fine. Beyond day-to-day help, Merge is genuinely invested in our success, actively seeking feedback to help shape their product roadmap, making us feel like a true partner.

What do you dislike about the product?

One area for improvement would be adding input validation for the ISU field within the Merge Link UI. Implementing this during the initial onboarding process would help flag incorrect values at the point of entry, streamlining the setup for end-users.

Also, provide better, less generic error messages from the /employees endpoint (or create a dedicated validation endpoint). This would allow users to realize their ISU is incorrect and self-correct the issue without needing to contact support

What problems is the product solving and how is that benefiting you?

Merge solves the high cost of integration maintenance and data fragmentation. For us, this means our engineers spend more time on innovation and our sales team rarely has to turn away a customer due to a missing HRIS integration.

What do you like best about the product?

Intuitive API, flexible Admin console, great support

What do you dislike about the product?

Programmatic support of custom field mapping

What problems is the product solving and how is that benefiting you?

removes the complexity from syncing with different HRIS api providers

What do you like best about the product?

I think the core model of Merge Unified is great, especially how it consolidates every vendor into a single API for us to integrate with. They have a solid understanding of their position in the marketplace and offer an ecosystem of people, documentation, and services that support us in building and delivering a new product offering. There's never a question we can't answer or a resource they haven't created for both internal and customer-facing documentation and enablement. The initial setup was as straightforward as a backend user account integration could be, and we could self-serve through the proof of tech and implementation with human support from solutions architects when needed.

What do you dislike about the product?

Some features we need and are excited about are paywalled behind add-ons or more expensive packages. It's the only part of the customer experience with any friction.

What problems is the product solving and how is that benefiting you?

I use Merge Unified to integrate different IdP and HRIS platforms with my SaaS platform for enterprise user account management, which was impossible to manage at scale before. It consolidates vendors into a single API, saving time and effort for our small engineering team.

What do you like best about the product?

I really appreciate the large collection of integrations they offer. They make the setup easy with thorough documentation and dedicated support.

What do you dislike about the product?

Sometimes partner requirements can be better streamlined. Beyond build, HRIS providers may have additional review processes or requirements. I find that process to be often time-consuming and unclear. It would be great if Merge was more informed or involved in that process.

What problems is the product solving and how is that benefiting you?

Solving a long tail of HRIS integrations which can take up a lot of engineering time without Merge

What do you like best about the product?

We love Merge. Swantide uses Unified API to power our ticketing integration (Jira, Asana, Linear, etc)  which supports our amazing partnerships with Hubbl, Provar, Moonox, etc

What do you dislike about the product?

There is nothing that we dislike about Merge Unified

What problems is the product solving and how is that benefiting you?

Merge doesn't just help, they power our Swantide integrations. This allows us to move at the speed of business and satisfy our customers.

What do you like best about the product?

The integration quality is objectively the best in the industry. We sync a ton of data and their team and product has been able to help us land some of our biggest customers. Highly recommend.

What do you dislike about the product?

We’d always love more categories so we can sync more data!!

What problems is the product solving and how is that benefiting you?

Everyone knows integrations are a huge pain. This lets us focus on our product.