The question is rarely about the technology
This question almost never originates inside the engineering team. It arrives from a board paper, a vendor demonstration, or a competitor press release. And when a conversation starts at the platform, it tends to end with a proof of concept that works technically and that nobody uses.
The more useful frame is this: a distributed ledger is not a faster or cheaper database. It is an expensive way to remove a referee from a process that currently needs one. Everything it gives you — a record that independent parties can rely on and that none of them can quietly rewrite — is paid for in throughput, operating cost, governance effort and legal work.
So the question is not whether you could use a blockchain here. You almost always could, which is why that question tells you nothing. The question is whether there is a referee you cannot remove any simpler way, and whether removing them is worth what it will cost you to do so.
Five conditions. All of them, or do not start.
This is the test we run before any conversation about networks or platforms. Each condition is a yes or a no. There is no "mostly".
One: more than one independent organisation writes to the same record. Not departments. Separate legal entities, with separate incentives and separate lawyers. Two business units inside the same group are not independent parties, however much they distrust each other.
Two: those organisations need to act on the same state, and today they reconcile instead. Look for the evidence of reconciliation — spreadsheets moving by email, a monthly meeting to agree balances, a master copy held by whoever shouts loudest.
Three: no participant will accept another participant, or a neutral vendor, holding that master copy. This is the condition that fails most often. If everyone is content for the largest buyer in the chain to operate the system, then operate the system. It will be cheaper, faster to launch, and it will work.
Four: disagreement about what happened has a cost you can name with a figure from your own books — a delayed payment, a re-inspection, a contractual penalty, an insurance claim, a written-off consignment. If you cannot name the cost, you are treating an irritation rather than a problem.
Five: the same parties will still be in the network in three years. A network with one member is not a network. It is a database with unusually poor ergonomics.
The sharpest single version of the whole test is one question: name the second organisation that will operate a node, and name the individual inside it who will sign off the budget line. If you cannot produce both names, what you have is not a blockchain project. It is an internal systems project that will be delivered faster with ordinary tools.
What a shared ledger does not fix
Immutability is a property of the record, not of reality. A wrong weight captured at the loading dock stays wrong forever, now protected by a cryptographic guarantee. Trust actually lives at the edge: the scanner, the inspector handset, the ERP integration, the person pressing approve. Most of the value and most of the risk in these projects sit there — and yet most of the budget goes to the chain.
A shared ledger also does not give you privacy by default. Sharing means visibility unless you deliberately engineer otherwise, and prices, volumes and counterparty identities are competitive information nobody will expose to rivals. The standard treatment is to keep data in the systems that own it and publish only a verification hash and a timestamp — which means a large part of the promised transparency gets designed back out again.
It does not remove governance either. It relocates it. Who joins, who is removed, what happens at a protocol upgrade, who pays for infrastructure, what happens when a member loses their keys or leaves and demands their data back — these end up in a written consortium agreement. That is legal and commercial negotiation, not engineering. Deferring it until after the build multiplies its cost.
And it will not shorten your internal process. If approval passes three desks today, it will pass three desks afterwards. Removing one of those desks is an organisational decision you can take this week with no new technology at all.
The middle path most teams skip: anchoring
A large share of the requests that reach us come down to one requirement: proving that a document or record has not been altered since it was issued. That is a perfectly legitimate requirement, and it does not need your system moved onto a chain or a consortium assembled.
The practical approach: keep your existing database, compute a cryptographic hash for each record or batch of records, publish that hash periodically to a public network or through a timestamping service, and hand the verification method to whoever needs assurance. Anyone can then check for themselves that the certificate, invoice or audit log in their possession matches what existed on that date — without asking you for anything, and without having to trust you.
Cost is modest, integration is shallow, and there is no membership agreement and no shared infrastructure to operate. When the requirement is third-party verification rather than shared writing between parties, this is usually the correct answer, and we say so even when it means a far smaller project than the client expected.
Where the bar genuinely does get cleared
The first pattern is verification of documents or credentials issued by one organisation and checked by many parties with no prior relationship to it. The point is that the verifier is a stranger and should not have to open an account or sign an agreement simply to confirm that a piece of paper in their hand is real.
The second is multi-party traceability where each participant holds only a fragment of the record and nobody sees the whole chain — producer, carrier, cold store, inspection body, importer. The value is not transparency as a slogan; it is a shared timeline that nobody can tidy up retrospectively once something goes wrong. Even here, the project frequently collapses on a simple question: is there a dominant party who could just mandate their own system? If there is, the centralised system is cheaper and it will succeed.
The third is settlement and reconciliation between entities in an ongoing commercial relationship who currently spend real staff hours every month agreeing balances and entitlements. One important caveat applies: any path touching payments, financial assets, funding instruments, asset tokenisation or fundraising is subject to independent legal and regulatory review before design rather than after it, and we provide no investment or financial advice of any kind. If your case sits in that territory, the legal track starts on day one, in parallel with design.
Permissioned or public — and the second-year cost
A permissioned network among known parties gives you member identity, control over privacy, predictable performance, and no exposure to volatile fees. In exchange you operate infrastructure, and nodes must genuinely be run by more than one organisation — otherwise you have built a database with extra steps and called it decentralised.
A public network solves the membership problem for you and offers neutrality that strangers will believe, along with mature tooling. In exchange: variable fees, public visibility of whatever you publish, full responsibility for key management, and dependence on a network whose governance and upgrade schedule are not yours.
The decision is rarely settled on year-one facts. It is settled on year two: a member joins, a member leaves, keys are lost and need rotation, an upgrade arrives that nobody wants to fund, and someone has to answer the phone at two in the morning when a node stops. Ask any vendor to describe precisely what happens when a member exits the network and demands their data extracted and their records removed. The quality of that answer tells you more about the vendor than any architecture diagram in the deck.
A thirty-day path to a real decision
Before any development budget is committed, five steps you can complete with internal resources and take to a board.
First, write the counterparty list — named organisations, each with the named individual inside it who will champion the project. A list containing "suppliers" and "regulators" in the plural is not a list. It is a wish.
Second, assemble the dispute log for the last twelve months. Every instance where the process stalled because parties disagreed about what happened, and exactly what each one cost you in time and money.
Third, price the status quo. Reconciliation hours, days of delayed collection, penalties, re-inspections. This is your own operational data and needs no market report to support it.
Fourth, design the boring alternative properly and price that too — a shared database run by a neutral operator under a clear contract, API integration between the systems that already exist, or the anchoring approach described above. The ledger has to beat the strongest version of that alternative, not a weak one built to lose.
Fifth, write proof-of-concept acceptance criteria before the build rather than after — which real commercial transaction runs through it, which parties actually use it, and what measurable difference it makes against today. In parallel, put the regulatory and legal questions to qualified counsel early.
You will end with one of two outcomes. Either five yeses, a named second organisation and a named sponsor inside it — in which case you have a real project with a defined scope. Or a budget you did not spend and a documented decision you can defend, which is a successful outcome by any reasonable measure.
Key takeaways
- Run the five-condition test before any platform conversation — a single no is a stop, not a caveat.
- Name the second organisation that will run a node and the person there who signs its budget. No names means it is an internal systems project.
- If the requirement is proving records were not altered, anchor hashes from your existing system instead of migrating onto a ledger.
- Price the boring alternative seriously — shared database, neutral operator or API integration — and make the ledger beat it.
- Write proof-of-concept acceptance criteria before development, and put regulatory questions to qualified counsel before design begins.
Frequently asked questions
We are being told we need blockchain for our supply chain. When is that right?
It is right when each party holds only a fragment of the record and no single party can compel the others onto its own system. If a large buyer or a central authority can mandate a platform for everyone, the centralised platform is cheaper, launches sooner and achieves the same operational outcome.
Can we start with a proof of concept and decide later?
Yes, provided acceptance criteria are written before the build: which real transaction runs through it, who uses it, and what measurable difference it makes. A proof of concept that starts without criteria ends as a successful technical demonstration that leads to no decision, which is the most common fate of these projects.
Do we need to issue a token?
The large majority of enterprise cases do not, and adding one introduces regulatory complexity with no operational return. Any work touching financial assets, funding instruments or fundraising is subject to independent legal and regulatory review before design. We give no investment advice and make no financial promises.
What if our partners will not join?
Then you do not have a network. There are three realistic options: shrink the scope to the two parties genuinely committed and build from there, use anchoring which requires nobody to join anything, or shelve the project until the incentive matures. Proceeding in the hope that partners join later is the single most common cause of failure here.
Sources
- Blockchain Technology Overview (NISTIR 8202) — National Institute of Standards and Technology (NIST) · https://nvlpubs.nist.gov/nistpubs/ir/2018/NIST.IR.8202.pdf