Question 1

How does DATFID differ from black-box AI models for business forecasting?

Accepted Answer

DATFID is interpretable by design. Instead of a neural network you cannot audit, every DATFID forecast returns a transparent mathematical equation that shows exactly how each driver — price, promotion, seasonality, etc. — influences the result, with coefficients and statistical significance. Every number is defensible to a CFO, a regulator, or a board. Black-box models give you a prediction; DATFID gives you a prediction plus the reason.

Question 2

What is interpretable AI, and how is it different from explainable AI?

Accepted Answer

Explainable AI (xAI) — techniques like SHAP or LIME — tries to explain a black-box model after the fact by approximating its behaviour. Interpretable AI builds transparency into the model itself from the start: the math is the explanation. DATFID is interpretable by construction. You never need a separate "explanation layer" because the model output already is a human-readable formula.

Question 3

Is DATFID a B2B data analytics tool?

Accepted Answer

Yes. DATFID is a B2B data analytics tool purpose-built for teams that forecast revenue, demand, and risk on panel data — multiple products, stores, regions, or customers tracked over time. It is used by CEOs, CDOs, Heads of BI, product managers, and data teams in retail, finance, energy, insurance, and SaaS.

Question 4

How does DATFID compare to AutoML platforms like DataRobot, H2O.ai, or Google Vertex AI?

Accepted Answer

AutoML platforms optimise for predictive accuracy by searching over opaque model ensembles. DATFID optimises for interpretability while maintaining competitive accuracy — it placed in the Top 5% of the M5 Walmart Forecasting Challenge against 5,000+ teams with zero model tuning or training time. You gain auditability and speed; you lose nothing on accuracy for typical B2B forecasting problems.

Question 5

How accurate is DATFID compared to deep learning models?

Accepted Answer

On the M5 Forecasting Challenge — the industry-standard benchmark covering 30,049 Walmart products, 10 stores, and 1,941 days of sales data — DATFID placed in the Top 5%, outperforming most Amazon and Meta baselines. It achieved this with no model training time. See the full report at datfid.com/m5-report.

Question 6

Do I need a data science team to use DATFID?

Accepted Answer

No. Business teams can run DATFID directly from Excel or the free playground at datfid.com/playground — no code required. Engineering and data teams can automate via the Python SDK or REST API. DATFID is designed so a Head of BI or product manager can get results the same day without waiting on a data science backlog.

Question 7

What data does DATFID need, and which file formats does it accept?

Accepted Answer

DATFID works with panel time-series data: one row per entity (product, store, customer) per time step. Accepted formats are CSV, Excel (.xlsx), and direct API upload. No preprocessing, feature engineering, or normalisation is required — DATFID handles raw business data as-is.

Question 8

Which industries and use cases is DATFID built for?

Accepted Answer

DATFID has been applied to retail demand forecasting, revenue forecasting, insurance claims pricing, energy and electricity load, loan default probability, payments analytics, and venture capital portfolio scoring. Full worked examples are available at datfid.com/docs/examples.

Question 9

Does DATFID integrate with Excel, Python, BI tools, and REST APIs?

Accepted Answer

Yes. DATFID ships a Python SDK, a REST API, and native Excel workflow. Model outputs — transparent coefficient tables — drop directly into Power BI, Tableau, or any ERP. Documentation and API key setup are at datfid.com/docs.

Question 10

Is DATFID GDPR-compliant and where is my data processed?

Accepted Answer

DATFID is built and operated in the EU — headquartered in Munich, Germany. All data processing is GDPR-compliant. Your data is never used to train shared models. For data processing details, see our Privacy Policy at datfid.com/privacy.

Frequently Asked Questions

Still have questions?