You’ve decided to modernize reporting on your ERP. Leadership wants Power BI dashboards, finance wants numbers that tie back to the general ledger, and someone has already asked when the AI tools will be able to answer questions about the business. The budget is approved.
Then comes the build vs. buy question for analytics, the one that decides how the next two years go: build the analytics layer in-house, or buy one that is already built for your system.
QuickLaunch sells the second option, so read this with that in mind. We’ve tried to make the case the way an IT or finance leader would test it: every figure is attributed, the one cost story is a customer’s own, and there’s a section on when building is the right call.
The decision matters most on systems with complex data. JD Edwards, Vista, Spectrum, NetSuite, OneStream and Salesforce each keep business meaning in codes, tables and logic that a generic BI build has to learn from scratch. We call that knowledge Application Intelligence: the business logic that turns raw system data into measures the business can use. Without it, a build moves unusable data from one place to another.
Why does building analytics in-house look like the right choice?
The case for building seems compelling at first glance. You have talented developers. You know your business better than any vendor. You want complete control over the solution, and paying for something your team could build feels hard to justify.
AI coding assistants have made that case look stronger. An analyst with Power BI and an AI assistant can stand up a dashboard in an afternoon. Fast is easy now. Fast and right, across the whole business, is the hard part, because the dashboard is only the part you can see. The ERP logic underneath decides whether its numbers tie out.
The report layer looks buildable because it is. The rest of this post is about what sits under it, and that is where most builds stall.
What makes ERP analytics harder to build than it looks?
Building Application Intelligence takes more than technical skill. It takes specific knowledge of how each application stores and processes data, and that gap catches experienced teams off guard.
Consider what it takes to turn JD Edwards data into reliable analytics:
- Converting Julian date integers into standard dates
- Reading decimal placement so financial values display correctly (otherwise millions become pennies)
- Translating user-defined codes across modules into business descriptions
- Knowing how JD Edwards uses journaling and handles deleted records, and reproducing that logic in the build
- Reshaping transactional tables into dimensional structures built for analysis
- Applying the structure finance expects: income statement and balance sheet hierarchies, fiscal calendars, multi-company and multi-currency consolidation
None of this is in standard technical documentation. It builds up over years of hands-on work with the application, and even certified Power BI consultants and experienced data architects struggle without it.
Every system has its own version of the problem. Vista needs construction-specific WIP calculations and cost code hierarchies. Spectrum’s job cost and equipment data need the same construction knowledge on a different schema. OneStream needs multi-dimensional cube structures flattened without losing the financial logic. NetSuite’s large multi-purpose transaction tables need restructuring before they perform well in Power BI. Salesforce needs historical pipeline snapshots built from Opportunity History, which most custom builds skip.
People who combine deep application knowledge with data architecture skills are rare and expensive to hire, and harder to keep.
The wider record is poor too: 73% of enterprise data initiatives fail to meet expectations (Fivetran Enterprise Data Infrastructure Benchmark, 2026). Builds that also have to encode Application Intelligence start from behind.
What happened when Washington Companies tried to build it?
Steve Schmidt, Business Intelligence Architect at The Washington Companies, was brought in 18 years ago to bring business intelligence to the group’s companies. His team supports about 25 operating companies running seven JD Edwards instances. When the group moved to Power BI, building its own version made sense on paper: the team knew JD Edwards well, and Microsoft introduced them to two Platinum-level partners.
Both were strong partners on the technical side. Across two attempts over a year and a half, they tried to build a simple general ledger solution from JD Edwards. Neither attempt worked. Steve explained why in our webinar:
“They were very good partners. Technically, but they didn’t understand E1, and what we learned is the knowledge of E1 is a big deal.”
Steve Schmidt, Business Intelligence Architect, The Washington Companies
The gaps were specific: how E1 uses journaling, how it handles deleted records, and how to reproduce both in a data mart. Then he compared the cost:
“What we paid for these two failed attempts to build just a single, simple GL solution was way more than we paid for an entire QuickLaunch solution across all the modules, fully implemented.”
Steve Schmidt, Business Intelligence Architect, The Washington Companies
Washington Companies installed the full QuickLaunch solution across all modules in 12 weeks, after two builds with two different Microsoft partners failed over an 18-month period.
What does a custom build cost beyond the budget line?
The build costs more than the buy, and most of that cost shows up as rework and delay rather than as a line on the budget. Washington’s two failed builds are the clearest example we have, and the hidden costs go further.
For scale, a from-scratch analytics build on an ERP such as JD Edwards or Vista typically runs 12 to 24 months and $500K to $2M. Treat that as a generic industry range. AI tools have brought some build costs down, and your own number depends on your team, your modules and how many attempts it takes.
Time and Opportunity Cost
Every month without trusted reporting means another close on spreadsheets, more manual work, and decisions made without data support. Over a build that runs a year or two, those months add up.
Organizational Impact
Talented team members spending months on struggling projects experience frustration and burnout. When projects fail to deliver, trust in IT’s ability to execute erodes, making future initiatives harder to launch and sustain across the organization.
Strategic Paralysis
While a team works on basic reporting, competitors with working analytics are improving operations and margins and taking market share.
Washington’s experience follows a pattern we hear across industries and company sizes, and its lesson can help other teams avoid the same costly detour.
Why does a finished build keep costing money?
Suppose the build works. The commitment has only started, and three costs keep coming.
Version Compatibility: The Never-Ending Upgrade Cycle
Enterprise applications change. ERP vendors release new versions with changed data structures and modified business logic, and each upgrade can break a custom analytics build.
Teams that build in-house face a recurring project with every upgrade. Tables change. Logic shifts. Fields are added or deprecated. Each change means rebuilding part of the analytics layer. Pipelines carry the same burden: in the Fivetran Enterprise Data Infrastructure Benchmark (2026), 53% of engineering time goes to pipeline maintenance.
Washington Companies went through this across JD Edwards World, then E1 8.0, 8.1, 8.2, 9.0, 9.1 and 9.2, and moved its back end from the i-Series to SQL. With a custom build, each step would have meant rework. With a vendor-supported solution, the team asked for the updated pipeline, connector and semantic model and installed it.
When one subsidiary had about 90 days’ notice to upgrade JD Edwards to meet a governance deadline, the team made the deadline the same way. A custom build would have needed its own upgrade project inside that window. As Steve put it, “from a business continuity perspective, you just can’t beat it.”
The Knowledge Retention Problem
Picture the day your lead developer, the person who understands the custom build, takes a job elsewhere. Their knowledge walks out the door. You’re left with code that’s difficult to maintain, business logic that’s hard to decipher, and consultants billing by the hour to reverse-engineer what the previous team built.
This affects every custom system, and it’s sharper with Application Intelligence, where the domain expertise is already scarce. Replacing someone who understands both your ERP and your analytics architecture can take months, and the analytics stand still in the meantime.
Scope Creep and Endless Enhancements
Custom builds suffer from constant scope expansion. Every department wants something slightly different. Every executive has unique reporting needs. Without a defined product scope, the analytics build becomes a never-ending development project that never reaches a stable state.
Pre-built solutions set clear boundaries. Core functionality is defined, proven, and supported. Extensions are possible and go through a structured process. Your team spends its time on insight instead of rebuilding infrastructure.
Why do custom analytics programs run long?
If you’ve talked with peers who attempted an ERP analytics build, the timelines they report are rarely the ones they planned. Four patterns drive the overrun.
1. Trying to “Boil the Ocean”
Teams try to build all of the ERP’s modules into the data model before delivering any value. They build massive enterprise data models upfront. They wait for “perfect” master data and governance. The result: months with no usable output, stakeholder fatigue, and relentless scope creep.
2. Leading with Technology Instead of Outcomes
Organizations choose tools first and define business value later. They build infrastructure, pipelines, and schemas without a tight business anchor. The analytics team operates disconnected from Finance and Operations realities. The result: a beautiful platform that nobody trusts or uses.
3. Underestimating ERP Translation Effort
This is especially true for JD Edwards. Cryptic tables, user-defined codes, and hard-coded business logic create layers of complexity that only surface during development. Business meaning lives in people’s heads and in spreadsheets, not in documentation. The result: endless rework, reconciliation fights, and the familiar complaint, “your numbers don’t match mine.”
4. Under-Resourcing the Team
Many mid-market companies run analytics with a few data engineers and a small analytics team. That’s thin for ingestion, data modeling, data quality, observability, security, and semantic model development across several business areas. Internal teams burn out, timelines slip, and the project quietly becomes a permanent initiative instead of a finished deployment.
The people on these programs are usually talented and committed. The approach is what stretches them: the wrong things happen in the wrong order, and the result is predictable.
What do you get when you start from a pre-built foundation?
After a build stalls, many teams look at pre-built Application Intelligence and find a different starting point. Steve’s advice to anyone at the start of the journey, from the same webinar:
“If you’re doing product selection, or wondering how you’re going to do it, don’t try building it. Buy.”
Steve Schmidt, Business Intelligence Architect, The Washington Companies
Starting from a pre-built foundation changes four things.
Proven Expertise Embedded in the Product
Pre-built packs carry years of domain expertise in production-ready semantic models. For JD Edwards, that means 3,900+ pre-built measures and KPIs, 3,200+ business-friendly dimensions and 35 business perspectives, from a team with 250+ implementations behind it. The depth runs in several directions at once. Julian dates, decimal precision and user-defined codes are decoded for the business. Business logic is built in, from gross profit and sales backlog to income statement and balance sheet hierarchies. One semantic model spans the modules, with multi-company, multi-currency and multi-instance structure, row-level security that mirrors JD Edwards, and descriptions AI tools can read. With those definitions in the semantic model, Microsoft 365 Copilot, Claude and Databricks Genie answer from the same trusted definitions as your reports. Every pack has its own depth, so check the figures for your system.
Rapid Deployment Timelines
Because the semantic model, pipelines and governed lakehouse already exist for your system, a pre-built foundation is production-ready in weeks, built right the first time. Timelines vary by system and scope. IGI Wax went live in 8 weeks against the 3 years vendors had quoted. Its finance team cut credit dispute resolution from 2 to 4 hours to 5 minutes, and it then ran machine learning on the same foundation: $8-10M increased annual profit, with waste in one production unit down from 8.5% to under 3%.
Vendor-Managed Version Compatibility
This may be the most underrated benefit. When your ERP upgrades, the vendor updates the semantic models to match, and does the same as Power BI, Fabric and Databricks change. No rebuilding required. No emergency projects. No disrupted reporting.
A Proven Path That Grows With You
Departmental BI projects can show early wins, but they rarely extend across the business on their own. A pre-built foundation follows a path proven across 250+ implementations, and it grows with you: add companies, systems and AI on the same foundation as you go.
How do the costs of building and buying compare?
A fair comparison counts more than a license fee against developer salaries. It counts every cost over the life of the solution: the team you hire and keep, the rebuild on each ERP upgrade, the rework when numbers don’t tie, and the months of decisions made without the data.
We don’t publish a percentage for the difference, because it depends on your team, your modules and your timeline. Build your own number from those three. Washington Companies is the reference point with a customer’s own words on it: two failed builds cost more than the full solution.
The table sums up the comparison this post has made.
| What you’re weighing | Build it yourself | Start from a pre-built foundation |
|---|---|---|
| Time to trusted ERP analytics | 12 to 24 months for a from-scratch ERP build, and builds frequently overrun | Production-ready in weeks, built right the first time |
| Up-front cost | $500K to $2M for an ERP such as JD Edwards or Vista (generic industry range) | A subscription, set against your own build estimate |
| Business logic on day one | What your team builds, one measure at a time | Tested logic for your system (for JD Edwards, 3,900+ measures and KPIs) |
| ERP upgrades | Your team rebuilds what breaks | Semantic models kept current for you |
| Know-how | Held by the people who built it | Held by the vendor’s team and shared with yours |
| What you own | All of it, and all of the upkeep | Your lakehouse, your data and the reports you built (the semantic model content is licensed) |
| Execution risk | High: builds stall, fail or depend on one person | Low: a path proven across 250+ implementations |
| Where your team spends its time | Pipelines, decoding and upgrade rework | Reports, analysis and AI on top of the foundation |
Ownership belongs in the comparison too. A build is yours to own and yours to maintain. With QuickLaunch, this is the ownership answer we give every buyer: “It’s your environment. You own the lakehouse, you own your data, and you own the reports you built. What lapses is the licensed semantic model content, plus the subscription benefits around it: Community access, professional services, and the ongoing upkeep that keeps the semantic model current as your ERP schema changes.”
When is building in-house the right call?
Building is the right call in some situations, and they are narrow. Be honest about whether yours is one of them:
- Your processes are proprietary: workflows that set you apart from competitors and that no vendor addresses, beyond “we do things slightly differently.”
- You have long-tenured staff with both deep application knowledge and data architecture skills, who plan to stay in maintenance roles for years.
- Your systems never upgrade, migrate, or change significantly. This scenario barely exists in modern IT.
- Stakeholders accept that analytics returns may take two years or more, and the budget can absorb overruns.
- You need a handful of dashboards from one module, and no governed semantic model across the business. The report layer is buildable, and a good analyst with Power BI can do it well.
- No pre-built option covers your core system, so part of the work has to be built either way.
For most companies running a major ERP, the first four don’t apply. Many land in between: buy the foundation and build on top of it. If you work with a partner, keep them. They build the reports, processes and first AI agents on the governed semantic model instead of starting from zero.
How should you make the decision?
The build versus buy decision shouldn’t rest on technical pride, perceived control, or proving what your developers can do. It should rest on delivering business value as quickly, reliably, and cost-effectively as possible.
Put these questions to your own team:
- What’s your cost of delay? Every month without trusted reporting is a month of decisions made with incomplete information: missed opportunities, manual work, and margin you can’t see in time.
- How much risk can you absorb? Think about a stalled project, a rebuild with each ERP upgrade, and the chance that one person holds the know-how.
- What’s your full total cost of ownership? Count maintenance, version compatibility, staff turnover, failed attempts, and the cost of waiting, then compare that total to predictable, vendor-supported costs.
- Where should your talent focus? Building pipelines and maintaining compatibility, or finding the opportunities in the numbers?
- Will finance trust it on day one? Ask whether the numbers tie back to the general ledger, with drill-down to the transaction, before the first close on the new reports. A build earns that one measure at a time.
- What’s your track record? If your organization hasn’t delivered a similar project before, what makes this one different, and do you have the application expertise or only general technical skill?
One practical test: ask any builder, internal team or partner, for their estimate to decode the ERP’s structures and build the semantic model, not only to land the tables. That estimate tends to change the plan.
For most organizations, honest answers to these questions point toward pre-built Application Intelligence.
What’s the next step?
An analytics program on an ERP needs Application Intelligence either way. The choice is how you get it: through a year or more of custom development with uncertain results, or by starting from a foundation already built for your system and having your team extend it.
The organizations that do this well buy the part that every company running the same ERP needs, and spend their own people on what sets them apart.
See it on demo data
See a pre-built semantic model for your ERP, from pipelines to reports
Frequently asked questions
What is Application Intelligence?
Application Intelligence is the expert translation layer that turns raw enterprise system data into business-ready measures. It encodes how applications like JD Edwards, Vista, Spectrum, NetSuite, OneStream and Salesforce work, including data structures, business logic and system-specific quirks that generic BI tools cannot handle without significant custom work.
Why do custom BI projects fail so often?
Data projects have a poor record in general: 73% of enterprise data initiatives fail to meet expectations, according to the Fivetran Enterprise Data Infrastructure Benchmark (2026). ERP analytics builds add more causes: underestimating Application Intelligence, the scarcity of people with both application and data architecture skills, scope creep, knowledge that leaves with key team members, and the ongoing work of keeping up with ERP upgrades.
How long does it take to build analytics in-house versus buying a pre-built solution?
A from-scratch analytics build on an ERP such as JD Edwards or Vista typically takes 12 to 24 months, and builds frequently overrun. A pre-built foundation is production-ready in weeks, with timelines that vary by system and scope. IGI Wax went live in 8 weeks against the 3 years vendors had quoted.
What is the total cost difference between building and buying?
A from-scratch ERP analytics build typically runs $500K to $2M, a generic industry range, before ongoing upkeep. A pre-built foundation lowers total cost by removing the build team, the rebuild on each ERP upgrade, and the rework from numbers that don’t tie. The difference depends on your team, modules and timeline, so build your own number. At Washington Companies, two failed builds cost more than the entire QuickLaunch solution across all modules.
When does building in-house make sense?
Building can make sense when your processes are proprietary and no vendor covers them, when you have long-tenured staff with both application and data architecture skills, when you only need a few dashboards from one module, or when no pre-built option covers your core system. For most companies running a major ERP, the better path is to buy the foundation and have your team or partner build on top of it.
What happens to custom-built analytics when our ERP system upgrades?
An ERP upgrade can break custom analytics. Tables change, logic shifts, and fields are added or deprecated, which can mean months of rebuilding. With a vendor-supported pre-built solution, you receive updated pipelines and semantic models that keep pace with the upgrade, without emergency projects or disrupted reporting.
Keep learning
- The Data Silo Tax: Calculating the Hidden Costs of Data Silos
- Why 80% of AI Projects Fail Before They Start: It’s Your Data Foundation
- Ultimate Guide to Connecting JD Edwards to Power BI
- Excel Limitations: Why Spreadsheets Can’t Scale for Enterprise Analytics
- The Definitive Guide to Enterprise BI Integration