Build vs Buy: Why Companies Stop Building Internal ETL Pipelines

Building an internal ETL data pipeline looks cheap on the first invoice and turns into the most expensive recurring line on the data team's budget by month three. When a CTO totals developer hours, API breakage fixes, monitoring, retries, OAuth refresh and the reports that still need to be built on top of the data, the build path runs several times the cost of a ready data platform. This guide gives CTOs, technical decision-makers and operations leaders a framework to compare the two paths using total cost, time-to-first-report and operational risk.

Building a custom data pipeline costs far more than the initial code: the recurring cost comes from API breakages, retries, monitoring, schema evolution and reports that need to be built from scratch. Kondado delivers 80+ ready data sources, included report templates, and bilingual support in English and Portuguese, freeing the engineering team for product and revenue work. The real choice is not "code vs platform"; it is "where does the expensive hour of your team go". The 5-question framework at the end of this guide makes the trade-off explicit.

Why do so many teams stop building internal data pipelines?

Because the cost shows up in layers that do not appear in the first sprint budget. When a team builds the first data pipeline with Google Ads, Meta Ads or an ERP, they solve the easy part: call the API, save the response. The price tag comes from what happens after:

1. API breakages ship without warning. Meta deprecated post_impressions in Graph API v24.0 in November 2025; anyone running an in-house data pipeline had to rewrite the extractor under deadline pressure to avoid blank reports the next morning.
2. Endpoint windows and rate limits vary by data source. The Bling API caps the gap between dataInicial and dataFinal at one year; teams that fail to handle this lose historical records silently.
3. OAuth tokens expire and need an automated refresh flow per connected account. Multiplied across dozens of clients (for agencies) or business units (for retailers), this becomes a system of its own just to manage credentials.
4. Retries, deduplication and idempotency need to be coded in every data source. A job that fails mid-run without a checkpoint leaves a partial dataset; the person who finds out is the operator who looks at the report on Monday.
5. Schemas evolve. Today's ERP returns category as a string; tomorrow it may return a nested object. Without contract testing, the report breaks silently.

Together these items are the recurring cost of the build path. They are also what separates a production data pipeline from a proof of concept.

What are the cost components of in-house data pipelines?

The estimate varies by team size, but the structure is stable. For five active data sources (a typical mix of one ERP, one e-commerce platform and two paid media channels), the minimum annual cost components are:

1. Initial development: weeks of a senior engineer to cover extraction, transformation, loading and basic error handling.
2. Corrective maintenance: a permanent fraction of the senior engineer's time to fix API breakages, schema drift and edge cases.
3. Monitoring and observability: alerts, pipeline health reports, structured logs, plus third-party tools.
4. Building the reports on top of the data: Power BI, Looker Studio or equivalent, with additional time per new report every time a business area asks for one.
5. Opportunity cost: hours spent on maintenance do not go into product. This cost is silent, rarely shows up on the planning sheet, and is where the ROI of the build decision leaks most.

Together these components push the build path to a different order of magnitude than a ready platform, with most of the cost coming from recurring maintenance, not from the initial build.

What does Kondado deliver in its place?

Kondado is a no-code data platform that replicates from 80+ data sources to destinations such as Google Sheets, Excel, BigQuery, PostgreSQL, MySQL, SQL Server, Redshift and Amazon S3, and to visualization tools such as Power BI, Looker Studio, Tableau, Metabase and Qlik Sense through two paths: direct connection Via Kondado with no warehouse required, or reads from the supported destinations.

The full catalog is in the Kondado data sources catalog and covers Google Ads, Meta Ads, Pinterest Ads, TikTok Ads, RDStation, HubSpot, Pipedrive, Salesforce, Shopify, VTEX, Tray, Nuvemshop, Mercado Libre, Bling!, Tiny ERP, Omie, ContaAzul, PostgreSQL, MySQL, MongoDB, BigQuery, and more across CRM, e-commerce, ERP, paid media, social, support, analytics and databases.

What that means for the build-vs-buy decision:

1. Time to first report. Replicating a data source in Kondado and seeing the data land in the destination takes minutes. Building the same from scratch, on the schedule you choose, takes weeks.
2. Zero data source maintenance. When the Meta or Bling API changes, Kondado keeps the data source in production. The internal engineering team stops receiving the "the report is broken" ticket.
3. Ready report templates. Kondado ships report templates for popular data sources (Google Ads, Meta Ads, ERPs, e-commerce), removing the "now someone has to model and design everything" step.
4. Predictable costs. Entry pricing is USD 19 per month, with a 14-day trial covering up to 30 pipelines and 10 million records, no credit card required. The build-path spreadsheet has no comparable cap.
5. Bilingual support in English and Portuguese via chat. When there is a question about API behavior, savepoint configuration or a specific ERP field, the support team replies directly. The internal engineering team stays focused on what creates competitive advantage.

When does it still make sense to build in-house?

The build choice still wins in three specific scenarios:

1. Proprietary or internal data sources with no data source available in any platform and no expected reuse. A one-off build is cheaper than paying a platform for something it does not cover.
2. Highly custom transformation logic that needs to run inside the source system before extraction (for example, proprietary financial calculations executed inside the company's own ERP).
3. Petabyte-scale workloads with sub-second latency requirements and contractual SLAs, where the platform cost exceeds the cost of a dedicated team. This profile is uncommon across mainstream business operations of any size.

Outside these three, the math is consistent: Kondado costs a fraction of a senior engineer and delivers 80+ data sources on day one of the subscription. The Kondado pricing page lists usage tiers, with USD or BRL billing options.

A 5-question decision framework

Before you approve the internal build ticket, answer:

1. Is the data source in the catalog? If yes, the direct cost comparison already says buy. If no, check with support whether the data source is on the roadmap.
2. Is there a report template for that data source? If yes, time to first report drops from weeks to hours.
3. What is the opportunity cost of the engineering team? Every hour spent on pipeline maintenance is one less hour on product.
4. How many breakage points can the operation absorb? Each custom data source is a failure point the team has to monitor, alert on and patch.
5. Does the platform support the refresh frequency the business needs? Kondado runs on the schedule you choose, configured per plan.

When 4 of the 5 answers point to "buy", the ROI is positive in month 1.

Frequently Asked Questions

What is the practical difference between building a custom API data pipeline and using Kondado?

Building means coding extraction, authentication, retries, monitoring, transformation and load for each data source, and then creating reports on top of the data from scratch. Using Kondado means picking the data source from the catalog, configuring the destination and receiving the data on the schedule you choose, with the option to use ready report templates or build visualizations in the BI tool of choice.

What are the cost components of an in-house data pipeline?

Initial development of extraction and transformation, permanent corrective maintenance for API breakages and schema evolution, monitoring and observability, building the reports on top of the data, and the opportunity cost of the engineering team on other projects.

Does Kondado serve large companies or only small ones?

Kondado serves companies of all sizes, from SMB to enterprise, in any market. The platform covers SMBs, agencies, e-commerce operations, ERP-driven teams and mid-to-large companies that want to combine marketing, sales and operations data without allocating engineering to the job.

Which data sources does Kondado connect?

More than 80 data sources across paid media (Google Ads, Meta Ads, Pinterest, TikTok), CRM (HubSpot, Pipedrive, RDStation, Salesforce), e-commerce (Shopify, VTEX, Tray, Nuvemshop, Mercado Libre), ERPs (Bling!, Tiny ERP, Omie, ContaAzul), databases, social media, support and analytics.

What does Kondado deliver beyond data extraction?

Beyond loading data into destinations such as Google Sheets, Excel, BigQuery, PostgreSQL, MySQL, SQL Server, Redshift and Amazon S3, Kondado ships ready report templates for popular data sources and a direct connection Via Kondado to visualization tools including Power BI, Looker Studio, Tableau and Metabase.

How does the trial work?

Fourteen days with up to 30 pipelines and 10 million records, no credit card required. Long enough to replicate three to five data sources and validate time-to-first-report on the real business flow.

How often does Kondado refresh the data?

Replication runs on the schedule the plan allows, on the cadence you choose. For typical reporting use, this covers the business need without the cost of a streaming architecture.

To validate the decision on your own operation, start the Kondado free trial and replicate your first data sources in minutes.