What it means
Many business ideas look good on paper and fail in practice. A proof of concept, often shortened to PoC, tackles this by testing the riskiest part of the idea early.
If a new payment system needs to connect to an old accounting package, the PoC might test only that connection. A PoC is narrower than a prototype and a pilot.
A prototype shows what the product might look like or how it might feel, and a pilot is a limited real-world launch with real customers. The PoC comes first and answers the question of whether it is technically or commercially possible.
Good PoCs are set up with clear goals, a fixed budget, a fixed time and agreed success criteria. For example, the team may agree that the test will be a success if the system processes 1,000 invoices with an error rate below 1%.
Without agreed criteria, results become a matter of opinion. For finance teams, a PoC is a way to limit risk.
Spending $30,000 to learn that an idea does not work is far better than spending $3,000,000 and learning the same lesson later. The cost of the PoC is usually expensed rather than treated as an asset, since the future benefits are not yet certain.
PoCs are common in technology, product development, manufacturing and process improvement. Software vendors often offer PoCs to prospective customers, and finance leaders should agree who pays, what data is used and how success is judged.
Care is also needed with real customer data, which may bring privacy obligations. One trap is that a successful PoC does not guarantee a successful rollout.
The test often runs in ideal conditions with expert staff, and scaling it to the whole business adds costs, training and complexity. The PoC result should feed a full business case, not replace one.
In practice
Real-world examples.
Example
A bank wants to use a machine learning tool to flag unusual payments. It runs a six-week PoC on one month of historic transactions to see how many true problems the tool catches, before it asks the board for a full budget. The PoC costs $45,000, a small fraction of the planned $1,800,000 programme.
Example
A food manufacturer wants to use a cheaper packaging material. It makes 500 packs on one production line and tests shelf life for 60 days. The results show that the material works, and a pilot with three retailers follows.
Example
A software company selling to a large insurer agrees a two-month PoC using a small set of the insurer's data. Success is defined as cutting claims processing time by 20% in the test group, and the contract is signed once that is achieved. Both sides agreed the target in writing before the test began.
Case study
Seen in the real world.
Meridian Freight is an illustrative, fictional logistics company that considered investing $2,500,000 in automated sorting equipment. The operations director was confident, but the finance director worried that the system might not cope with the company's irregular parcel shapes.
The company ran a PoC costing $60,000 with a vendor's demonstration unit, using a sample of 5,000 real parcels. Accuracy was only 91% against the agreed target of 98%, mostly because of odd-shaped items. Meridian paused the investment and asked the vendor to improve the system, and the illustrative lesson was that a small test had avoided a very expensive mistake.
The vendor returned three months later with a revised design, and a second test of 5,000 parcels reached 98.4% accuracy. Meridian then approved a full purchase with a staged payment plan linked to performance. The illustrative decision was easier to defend because it rested on data from its own parcels.
Watch out
Common mistakes.
- Starting a PoC without clear success criteria, so the outcome becomes a matter of opinion.
- Treating a successful PoC as proof that the full rollout will succeed, when scale brings new costs and risks.
- Letting a PoC grow in scope and cost until it becomes a project in disguise.
Questions
People also ask.
How is a proof of concept different from a prototype?
A proof of concept tests whether an idea is feasible, while a prototype shows how a product might look and work.
How long should a proof of concept last?
Usually weeks to a few months, long enough to test the key assumption but short enough to limit cost. Setting an end date in advance stops the test drifting into an open-ended project.
Who should pay for a proof of concept?
This is negotiated, as a vendor may offer it free or at a reduced price, while the customer may pay to ensure commitment. Whoever pays, the scope, data and success measures should be agreed in writing before work starts.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%Related