Building is cheap now. Knowing what to build is the job.
A Fast MVP pairs an expert product analysis with a working first version you can build on: a new feature, app, model, or dashboard, with a researched strategy and execution plan. Four to eight weeks, scoped to the metric that defines success.
From the shiproom. We've seen how products ship, stall, and get shut down.
Expedia and Microsoft
Close to 15 years behind the scenes, in the shiproom with the metrics when key features shipped, got held back, or were shut down. Grew Microsoft Word online and mobile 300% to more than 150M Monthly Active Users in 18 months by pairing strategic analytics and A/B experimentation.
Navigate360
Reported to the Chief Product Officer and owned products. Architected and shipped an agentic AI platform on Databricks in 3 months, unifying six legacy products behind a single AI interface and data platform.
TNS Call Guardian
Led data science and ML engineering for a voice spam and fraud detection platform ranking billions of calls in near real time. Built full MLOps that moved model releases from every 18 months to monthly, lifting fraud detection from near coin-toss to >90% precision and recall within 18 months.
Anyone can build it. Few can say if it should exist.
AI tools have made code fast and cheap. That moves the risk. The expensive mistake now is a well-built product nobody needed, or a feature that ships without a way to tell if it worked.
Most MVPs skip the hard part: who the user is, what they do today, what would change their behavior, and the one number that proves it.
Quick prototypes also get thrown away because they were never built to last: no tests, no security, no owner.
Define the metric first, test it where you can, ship small releases often, and treat security and privacy as part of done. Those habits come from twelve years at Microsoft.
Signs you need a Fast MVP
The idea is stuck in slides
Leadership agrees it matters, but nobody has seen it work or agreed how to measure it.
Several ideas, one budget
You need to know which feature, app, or model is worth building first.
Spreadsheets run the process
A shared workbook and a pile of emails hold a process that should be an app or dashboard.
You need proof for a decision
A board, a lender, or a big customer wants to see it before you commit.
Fast, without the throwaway
Code lives in your repository from the first commit. Every change goes through automated tests and a CI/CD pipeline, so the MVP can grow into the product instead of being rewritten.
We build on what you already run: a web app, a data app on Databricks Apps, or an internal tool in your cloud. Data stays in your account.
Sign-in uses your identity provider. Secrets live in a vault, not in the code. Logs and basic monitoring are on from launch.
AI coding tools write a first draft. A senior engineer reviews, tests, and owns every line.
Analysis first, then a build you can keep
Research
Users, market, competitors, and your own data. We look at the problem from several angles before choosing one. You see working software every week, not status reports.
Define success
Agree the one metric the MVP must move, and how we will measure it.
Build
Short cycles with a senior engineer using AI engineering tools. A live demo every week.
Launch and measure
Real users on real data, measured against the success metric.
Plan what's next
Keep, change, or stop, with a strategy and execution plan for the next phase.
Deliverables
Product analysis: who the user is, the problem worth solving, and where the idea is strong or weak.
Research: user, market, and competitor findings, plus what your own data shows.
A success metric: the one number that decides keep, change, or stop, with a way to measure it.
A working MVP (feature, app, model, or dashboard), deployed in your cloud, with source code, tests, and a CI/CD pipeline in your repository.
Strategy and execution plan: a ranked roadmap, team and effort estimates, and a handover runbook so your team can extend it.
If your data is on Databricks, the MVP can live there too
Databricks Apps runs web and data apps inside your workspace, with Unity Catalog permissions applied to every query. No separate servers, and no copy of your data leaving the platform. The pipeline under the app is Lakeflow jobs, Delta tables, and tested deployments with Asset Bundles.
What a first version can look like
Real estate
A tenant portal for maintenance requests tied to your property data.
Hospitality
A staff app that shows today's arrivals, VIPs, and special requests.
Fintech
A merchant dashboard showing settlements and disputes in near real time.
Logistics
A dock scheduling tool that cuts phone and email back-and-forth.
Four to eight weeks, matched to the metric
The engagement is 4–8 weeks. The three phases below sit inside that window. Simpler scopes reach first users sooner.
Research and define
Product analysis, research, and the success metric agreed. First clickable screens.
Build with weekly demos
The MVP on your data. A live demo every week. This phase stays inside the 4–8 week engagement.
Launch, measure, and plan
Measure real use against the metric, hand over, and deliver the strategy and execution plan.
Access control, secrets handling, logging, and monitoring ship in the first version, not the second. Security is part of how we build, not an add-on. A deeper security assessment is available through a trusted local partner.
References
DORA research program, on delivery practices such as testing and CI/CD: https://dora.dev/research/
Databricks Apps documentation: https://docs.databricks.com/aws/en/dev-tools/databricks-apps/
What is Unity Catalog? https://docs.databricks.com/aws/en/data-governance/unity-catalog/
McKinsey, The economic potential of generative AI (2023): https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
Kohavi, Tang, and Xu, Trustworthy Online Controlled Experiments (Cambridge University Press, 2020): https://www.cambridge.org/core/books/trustworthy-online-controlled-experiments/D97B26382EB0EB2DC2019A7A7B518F59
Marty Cagan, INSPIRED, 2nd edition (SVPG): https://www.svpg.com/inspired-how-to-create-products-customers-love/
Bring one idea. Leave knowing if it's worth building.
Start with a free discovery session. We look at the idea, name the metric that matters, and tell you honestly whether it fits in four to eight weeks.