Business

7 Common Mistakes to Avoid When Submitting a 510K to the FDA

Published

4 years ago

January 21, 2022

The FDA deserves credit for ensuring high patient safety standards. However, there is no ignoring the hassle medical device manufacturers go through when submitting 510K applications. They spend hours collecting documents and data from multiple departments only to face a 36% prospect of having their application rejected.

While there is no formula to always getting your submissions cleared by the FDA, you can increase your chances of approval and avoid delivery delays and unnecessary stoppages by ironing out things on your end. Here are some of the most common mistakes manufacturers make that you can easily avoid:

1. Losing track of your product’s regulatory history

Your company ought to know its product’s regulatory history in the U.S., since that’s what 510Ks are based on. Unfortunately for most companies, poor data-keeping leads to loss of important information resulting in a bitter clash with the FDA. No matter the history of your product, it’s good to keep data where you can access it and not likely to lose it. A dedicated clinical metadata repository software tool, such as Formedix Ryze can help you take control of the key challenges associated with keeping and organizing data.

2. Using incorrect FDA templates

Up in the FDA checklist is the correct use of their templates. The agency requires that each section of all 510K submissions be based on an FDA-issued template. Most manufacturers remember this but then forget how rapidly the FDA updates these templates. While using an older template doesn’t automatically render your submission void, it increases your chances of leaving out some data, which you can’t get away with. For this reason, it’s good to confirm that the template on your hands is the latest issued by the FDA before drafting your application.

3. Data irregularity

The FDA requires that you be consistent with the information you provide if it appears multiple times in your application. If there is a discrepancy in your wording, your application will likely be flagged and even rejected. So while keeping your intent consistent, make a point of doing the same with your wording for the sake of your application’s approval.

4. Skipping sections

A typical 510K application form has 20 sections, some of which may not apply to your device. For most manufacturers, irrelevant sections include Electromagnetic Compatibility and Electrical Safety, Performance Testing and Proposed Labeling, Disclosure Statement or Financial Certification, and Class 3 Summary and Certification. If any of the sections don’t apply to you, it is required that you confirm it in writing.

5. Choosing an incorrect predicate (comparison) device

The FDA will treat your device like they did a previously cleared one, meaning you have to identify a device whose parameters match those of yours. Your predicate of choice should bear similarity in design, size, materials, packaging, indications for use, and other considerations, failure to which you will draw out the review process, and even risk rejection. For instance, if your device requires sterilization before use, while the predicate is supplied sterilized, the FDA will ask for more information before getting on with the review process.

6. Failing to comply with the Refusal to Accept provisions

Nearly 90 percent of all rejected submissions are tossed out before being reviewed by a human. This is because they don’t tick off the Refusal To Accept (RTA) checklist, which outlines across-the-board prerequisites. Meeting the RTA requirements simply means your device is worthy of an FDA review and has a realistic chance of being cleared.

7. Misunderstanding the point of a 510K submission

The 510 (k) has evolved quite remarkably over the years. Some time back, it was an endless series of paperwork submissions; now, it’s a streamlined affair that makes maximum use of mainstream contemporary technology. In all that, one thing remains the same: the purpose of the 510K, which is clearance through association or clearance for devices similar to other previously cleared devices.

Failure to understand that can have you wondering why the FDA is hard on you. As stated above, you should have a predicate device at the ready or even model your device on an existing one. That is not to say you should shy from being creative. However, if you want it easy with the FDA, you have to make it easy for them first.

Endnote

If you have been struggling to meet the requirements for a 510K clearance, you’re in good company. The process requires time, manpower, data, and a ton of resilience. It doesn’t have to be a hassle, though. By avoiding the above mistakes, you can massively simplify the process and speed up the review process.

Business

AI in Asset Management Explained: How Leading Firms Apply It

Published

1 day ago

April 16, 2026

Matthew Wiles

AI in asset management explained at its most basic level is this: using machine learning, data modeling, and automation to make faster and more accurate investment decisions. The applications vary widely across asset classes, fund strategies, and operational functions. Understanding where AI creates real value separates productive adoption from expensive experimentation.

Asset managers now face a data environment far larger than any human team can process manually. Market signals, company filings, macroeconomic indicators, alternative data sources, and portfolio monitoring all generate information continuously. AI tools process that information at scale. They surface patterns that traditional analysis would miss or find too late.

AI in Asset Management Explained Across Core Investment Functions

AI delivers the most measurable results when applied to specific investment functions rather than deployed as a general capability. The clearest applications sit in portfolio construction, risk management, and credit analysis.

Portfolio Construction and Factor Modeling With AI

Traditional portfolio construction relies on return and correlation assumptions built from historical data. AI-driven portfolio tools go further. They process real-time market data, alternative signals, and macroeconomic inputs simultaneously. This surfaces factor exposures that static models miss.

Machine learning models in portfolio construction can:

Identify non-linear relationships between asset classes that correlation matrices do not capture
Adjust factor weightings dynamically as market conditions shift rather than on a quarterly rebalancing schedule
Flag concentration risks before they appear in standard risk reports
Model tail scenarios using a broader range of historical stress periods than traditional value-at-risk models allow

James Zenni, founder and CEO of ZCG with over 30 years of capital markets experience, has built the platform’s investment approach around the principle that better data and faster analysis produce better outcomes. That view shapes how AI capabilities get deployed across ZCG’s private equity, credit, and direct lending strategies.

Credit Analysis and Private Markets AI Applications

Credit analysis in private markets has historically depended on periodic financial reporting and relationship-based deal intelligence. AI changes that model. Lenders using machine learning tools now monitor borrower health continuously rather than waiting for quarterly covenant tests.

Specific credit applications include:

Cash flow pattern analysis that identifies revenue deterioration weeks before it shows up in reported financials
Supplier and customer relationship mapping that flags single-source dependencies and concentration risks
Covenant monitoring automation that tracks hundreds of credit agreements simultaneously and alerts teams to early warning signs
Loan pricing models that incorporate current market spread data and comparable transaction history

These capabilities compress the time between identifying a problem and taking action. In credit, that time advantage directly affects loss rates and recovery outcomes.

AI in Asset Management Explained Through Risk and Compliance Applications

Risk management and regulatory compliance represent two of the highest-value AI applications in asset management. Both functions involve processing large volumes of structured and unstructured data under time pressure.

How AI Transforms Risk Monitoring in Asset Management

Traditional risk monitoring produces reports at set intervals. AI-powered risk systems run continuously. They flag anomalies in position data and monitor correlated exposures across a portfolio. Alerts fire when market conditions shift beyond defined thresholds.

The practical risk management applications include:

Real-time portfolio stress testing against live market inputs rather than end-of-day snapshots
Liquidity modeling that accounts for position size relative to market depth across multiple scenarios
Counterparty exposure monitoring that aggregates risk across instruments, custodians, and trading relationships
Regulatory reporting automation that reduces manual preparation time and lowers the risk of filing errors

ZCG applies these capabilities across its approximately $8 billion in AUM. The platform was founded 20 years ago. It built its investment infrastructure around systematic data analysis and operational discipline.

AI for Operational Efficiency in Asset Management Firms

Beyond investment decisions, AI delivers significant value in fund operations. Back-office functions like reconciliation, reporting, and compliance documentation consume substantial resources at most asset management firms.

AI tools applied to fund operations include document processing systems. These extract and verify data from offering documents, side letters, and subscription agreements automatically. Reconciliation tools flag breaks between custodian records and internal systems automatically. Investor reporting platforms generate customized materials from structured data inputs, reducing the manual production time significantly.

ZCG Consulting (“ZCGC”) advises operating companies across more than a dozen sectors on operational improvement programs, including technology-driven process redesign. Those operational efficiency principles translate directly to asset management back-office functions.

Applying AI to Asset Management: Limitations Firms Must Address

AI in asset management explained fully must include the limitations. Models trained on historical data perform poorly when market regimes change. Overfitting produces tools that work in backtests but fail in live environments. And AI outputs require experienced interpretation to avoid acting on statistically significant but economically meaningless signals.

The ZCG Team approaches AI adoption with the same discipline it applies to investment underwriting. Every tool requires a defined use case and a measurable success metric. A review process keeps experienced judgment in the decision chain. That framework prevents the common failure mode where AI adoption generates activity without improving outcomes.

Firms that treat AI as a capability layer on top of sound investment processes generate sustainable advantages. Those that treat AI as a replacement for process discipline find the technology amplifies existing weaknesses. It rarely corrects them.