Every return-on-investment figure published in this category is unverifiable before purchase, including ours. Savings of sixty-five to eighty-five thousand dollars per vessel annually, payback in four to six months, seventy percent less administrative time, seventy-eight percent fewer deficiencies — these numbers circulate widely, they come from suppliers rather than from audit, and no buyer can test any of them in advance because the counterfactual does not exist. A fleet that installs software and records fewer detentions the following year cannot know whether the software caused it, whether trading patterns changed, or whether the previous year was simply unlucky. That is not an accusation against any particular vendor; it is a structural property of the claim. Which leaves a buyer needing a business case and unable to build one from anything a vendor supplies. The way out is to stop trying to verify somebody else's numbers and build the model from the only evidence that reflects your fleet: your own last three years of detentions, deficiencies, off-hire, deferred work and yard change orders. This page is about how to construct that, and it will not give you a payback figure — because anybody who gives you one has not seen your data. Start a free trial of Marine Inspection and test the assumptions against your own operation.
A claim you will see in this category
Fleets achieve annual savings of sixty-five to eighty-five thousand dollars per vessel, with payback in four to six months.
Can you verify it before buying?
No
Can you verify it after buying?
Not reliably — no counterfactual exists
Was it independently audited?
Almost never in this sector
Does it describe your fleet?
Unknown — it describes somebody else's
Apply those four tests to any figure in a proposal, including any figure we publish. What survives is a marketing claim rather than a forecast, and a business case built on it will not withstand a finance director who asks where the number came from.
Two Kinds of Benefit, and Only One Is Modellable
The distinction that makes a business case defensible is between benefits you assert and benefits you can price from your own history. Book a Marine Inspection demo and build the second column rather than quoting the first.
Efficiency benefits
Real, and almost impossible to defend in a paper
Administrative time reduction, frequently quoted in the region of seventy percent
Faster regulatory reporting, reported at seventy to eighty percent
Reduced inspection duration and faster close-out
Better capital planning through lifecycle analytics
The problem is not that these are false. It is that saved time only becomes money if the headcount changes or the freed hours are demonstrably redeployed, and in most fleets neither happens visibly. A finance director will discount them heavily and will be right to.
Avoided events
Fewer, harder, and priceable from your own records
A detention, with the full cascade of delay, port cost and commercial consequence
An off-hire day attributable to a technical failure
A yard change order arising from scope missed in the docking specification
A fine following an expired certificate found at inspection
An expedited part, an attending service engineer, a positioning voyage
Each of these has a cost you have already paid at least once, recorded somewhere in your own accounts, on a date you can find. That makes them the only benefits worth putting in a board paper, because every number in the model comes from your own past rather than from a supplier's.
Build the Model From Four Columns
The structure below produces a defensible figure in an afternoon, and its most useful property is that every input is yours. Sign up for Marine Inspection and fill the third and fourth columns honestly.
Event type
Your cost, last time
Your frequency
Plausibly addressable
Detention
From your own last one — delay, port charges, overtime, freight, charter consequence
Count across three years, divided by three
The share where the root cause was a record or a missed item rather than a physical failure
Technical off-hire
Daily rate multiplied by days, plus the repair
Days per vessel per year from your own off-hire log
The share where a part was not aboard or a defect had been noticed earlier
Yard change orders
Variation cost on the last docking against the original specification
Per docking cycle
The share traceable to scope known aboard and absent from the specification
Deficiency close-out
Cost of rectifying findings under time pressure versus in planned work
Findings per inspection from your own record
The share that recurred after a previous inspection had already raised them
Expedited procurement
Premium paid over planned purchase price, plus freight
Frequency from purchasing records
The share caused by stock position being wrong rather than by genuine emergency
Multiply columns two, three and four for each row and add them. That total is the honest annual value at risk, and it is defensible because a finance director can trace every input to your own accounts. Compare it against a three-year all-in per-vessel software cost, and you have a business case that does not depend on believing anybody.
The fourth column is the honest one, and it is where most business cases fall apart
Nobody should claim software prevents every detention or every off-hire day. The question is narrower and answerable: of the events you have already paid for, how many had a root cause that a better record would plausibly have caught — a certificate nobody was tracking, a defect noticed and not logged, a part the stock figure said was aboard, a finding from the last inspection that was never closed. Go through three years of events and mark each one. That fraction is your case.
Use the Base Rate Before You Use an Estimate
One useful external number exists, it is published annually by regulators rather than by vendors, and it lets you sanity-check your own exposure. Schedule a walkthrough and place your fleet against it.
The published rate
The Paris MoU recorded a detention rate of around four percent across more than sixteen thousand inspections in 2025, with deficiencies found in roughly sixty percent of inspections and detentions arising from an accumulation of findings rather than from single catastrophic faults.
How to apply it
Count your own inspections per year across the fleet. Multiply by the published rate to get an expected detention count at the industry average. Then compare that with your actual history — if you sit above the average, your addressable exposure is larger than the base rate suggests; if below, the honest model is smaller and you should say so.
Why it matters
A model built only on your own three years is vulnerable to a quiet period making the case look weak, or a bad year making it look implausibly strong. Anchoring against a published external rate gives the paper a second reference point and shows the finance function that you looked for one.
Fleet Size Changes Which Lever Matters
The same software produces value through different mechanisms at different scales, and a model that ignores this either overstates the case for a small fleet or understates it for a large one. Start a free trial and model the lever that actually applies to you.
One to five vessels
Single-event exposure dominates
A per-vessel software cost is small relative to a single detention or a week of technical off-hire. The case is essentially an insurance argument: you are not buying efficiency, you are reducing the probability of one expensive event that would dwarf the annual spend. Efficiency savings barely feature because there is no headcount to reduce.
Model on: probability times cost of the worst single event
Six to twenty vessels
Both levers operate
Event exposure is now recurring rather than hypothetical — at this size something happens every year — and a shore team exists whose time is genuinely consumed by reconciliation. The strongest case combines avoided events with a specific, named administrative saving rather than a percentage.
Model on: annualised event cost plus one named role's recovered capacity
Twenty vessels and above
Standardisation and comparison dominate
The distinctive value is no longer per-vessel; it is the ability to compare vessels, spot the outlier before it becomes an event, and apply one standard across a fleet operated by different crews and sometimes different managers. Volume licensing also changes the cost side — larger fleets are reported to achieve meaningful per-vessel reductions.
Model on: outlier detection value plus volume pricing effect
Notice what this does to the vendor ROI claim. A single figure quoted as applying to fleets generally cannot be right at all three scales, because the mechanism producing the value is different in each. When a proposal quotes one payback period without asking your fleet size, it is describing a model that was not built for you.
Which of the five event types cost you most in the last three years?
That answer determines the shape of your business case, and it is in your own records rather than in anybody's proposal.
Where Business Cases Go Wrong
Six recurring errors, each of which produces a number that does not survive the first serious question. Book a walkthrough and pressure-test your own draft against these before it goes upstairs.
01
Quoting a vendor's percentage
A supplier figure imported unexamined invites exactly one question — where did that come from — and the answer undermines the whole paper. If you must reference an external figure, cite it as an industry claim and build the case without depending on it.
02
Counting saved time as saved money
Hours recovered are only money if headcount falls or the hours are visibly redeployed. Claiming a cash saving from time that stays inside the same salaries will be discounted to nothing, and it damages the credibility of the harder numbers alongside it.
03
Ignoring the internal cost
Somebody inside your organisation will own the implementation. Their time is a real cost that appears on no invoice, and leaving it out of the model is the single most common reason a project that met its business case still feels like it did not.
04
Expressing return as a percentage
A large percentage on a small base is a small number wearing a big hat. Four hundred percent on a modest per-vessel spend is a figure a board will correctly regard as immaterial. State the absolute amounts and let the ratio be an observation rather than the headline.
05
Modelling only the first year
Implementation costs land once and benefits accrue continuously, so a one-year view systematically understates the case while a five-year view flatters it. Three years is the defensible horizon and it should match the horizon used on the cost side.
06
Claiming certainty about prevention
Software does not prevent a detention. Better records make some causes visible earlier, and some of those would have been acted on. Saying so explicitly, with a stated addressable fraction, is more persuasive than a confident claim — because the person reading it knows the confident version is not true.
The Cost Side, Briefly
Kept short deliberately, because the cost half of this equation is covered properly in our pricing benchmark and quote normalisation guides. What matters here is using the same horizon on both sides. Start a free trial and get one comparable figure without a proposal cycle.
Table 1: What Goes on the Cost Side of the Model
| Line | Basis | Common error | What to ask for |
|---|---|---|---|
| Licence | Per vessel, per user or per module depending on the model quoted | Comparing across different units without converting | The figure restated as cost per vessel per year |
| Implementation | One-time, frequently a range rather than a fixed fee | Using the bottom of a quoted range in the model | A fixed-fee ceiling for go-live across the fleet |
| Data migration | Scales with the condition of your existing records | Assuming your data is better than it is | A defined position on what happens if it is worse than expected |
| Annual uplift | Applied at each renewal across the horizon | Modelling three years at the year-one rate | A contractual cap, and the capped figure in the model |
| Retraining | Recurring, because crews rotate several times a year | Treating training as a one-time implementation line | The per-rotation cost, or confirmation it is included |
| Internal ownership | Whoever runs the project, for however long it runs | Omitting it because it does not generate an invoice | An honest estimate of days, costed at a real rate |
| Fleet scale effect | Larger fleets are reported to achieve per-vessel reductions through volume terms | Applying a small-fleet benchmark to a large fleet or the reverse | A quote for your actual vessel count rather than a published band |
| Horizon | Three years, matching the benefit side | Costing three years and benefiting one, or the reverse | Both sides stated over the same period, explicitly |
Questions to Put to the Vendor
These are designed to test whether a supplier's claims can survive contact with your model, and a good vendor will welcome most of them. Schedule a demo and ask them before a proposal is written rather than after.
Table 2: Testing an ROI Claim
| Area | The Question | A Real Answer | What Should Worry You |
|---|---|---|---|
| Provenance | Where does that savings figure come from? | A named source, a sample size, and an admission that it is customer-reported | A figure with no origin, repeated confidently |
| Fleet match | Which fleet size and vessel type produced that number? | A specific profile you can compare against your own | One figure quoted as applying to fleets generally |
| Mechanism | Through what mechanism does the saving arise? | A traceable chain from a behaviour change to a cost line | An outcome asserted with no mechanism between |
| Counterfactual | How do you know the software caused it? | Candour that attribution is difficult, and what evidence exists anyway | Certainty, which is the least credible possible answer |
| Reference | Can I speak to a customer of my size about what actually changed? | A contactable reference with a comparable fleet | Logos, case study PDFs, and no reachable person |
| Failure cases | Tell me about an implementation that did not deliver | A real example, with what went wrong and what they changed | A claim that every deployment succeeded |
| Adoption evidence | What proportion of a reference customer's crew use it weekly? | A number the vendor volunteers, confirmed by the reference | Licence counts offered instead of usage figures |
| Time to value | When would the first measurable benefit appear, and which one? | A named early indicator you could check within a quarter | A payback period with no leading indicator behind it |
| Exit | If the model does not hold, what does leaving cost? | Export in an open format at no charge, stated contractually | Vagueness, which raises the real cost of being wrong |
2026 MODELLING REALITY
Every efficiency and savings figure referenced on this page is supplier or industry reported rather than independently audited, including figures published by us. They are cited to illustrate what circulates in the market and how to treat it, not as inputs you should adopt. Attribution is genuinely hard, not merely inconvenient. Fleets change trading patterns, crews, managers and vessel composition constantly, so isolating the effect of one system change is difficult even after the fact — which is why a model built on your own historical event costs is more defensible than one built on projected savings. Detention statistics vary by region and year. The Paris MoU figures referenced reflect one memorandum's published annual reporting; other regions differ, and your own exposure depends on trading pattern, flag, vessel age and company performance history. Cost benchmarks belong in a separate exercise. Per-vessel pricing bands and quote normalisation are covered in our dedicated pricing guides; this page assumes you have already produced one comparable cost figure to model against.
Frequently Asked Questions
Why can I not just use the vendor's ROI figure?
Because it is unverifiable in both directions. Before purchase you cannot test it, and after purchase you cannot isolate the software's contribution from everything else that changed — trading pattern, crew, vessel composition, market conditions, or simple variance in how many inspections happened to go badly. That is a structural property of the claim rather than a criticism of any particular supplier, and it applies equally to figures we publish. A finance function will discount an imported percentage heavily and will be right to. A model built from your own historical event costs survives the question of where the number came from, because the answer is your own accounts.
What should the model actually contain?
Four columns per event type. First, the event — detention, technical off-hire, yard change order, deficiency close-out under pressure, expedited procurement. Second, what it cost you last time, taken from your own records rather than from a published range. Third, how often it happens, counted across three years and annualised. Fourth, and most importantly, the fraction plausibly addressable — the share of those events whose root cause was a record failure rather than a physical one. Multiply the three numeric columns per row and add them. That total is your annual value at risk, and every input traces to something a finance director can check.
Should efficiency savings go in the business case at all?
Mention them, do not lean on them. Administrative time reduction is real and frequently substantial, but hours saved only become money if headcount falls or the recovered time is visibly redeployed to work that was previously not being done. In most fleets neither is demonstrable, so a claimed cash saving from time will be discounted to nothing — and worse, it undermines the credibility of the harder avoided-event numbers sitting alongside it. If you do include time, name the specific role, the specific task, and what that person will do instead.
How does fleet size change the argument?
It changes the mechanism producing the value, not just the scale. Below about five vessels the case is essentially an insurance argument — a single detention or a week of technical off-hire dwarfs the annual software cost, and there is no headcount to make efficiency claims about. Between roughly six and twenty, events become recurring rather than hypothetical and a shore team exists whose time is genuinely consumed, so both levers operate. Above twenty, the distinctive value shifts to standardisation and outlier detection across vessels operated by different crews. A vendor quoting one payback period without asking your fleet size has not modelled your situation.
Is there any external number I can use?
Regulatory detention statistics are the most useful, because they are published by authorities rather than suppliers. The Paris MoU reported a detention rate of around four percent across more than sixteen thousand inspections in 2025, with deficiencies found in roughly sixty percent of inspections. Multiply your own annual inspection count by the published rate for an expected detention figure at the industry average, then compare it with your actual history. Sitting above the average means your addressable exposure is larger than the base rate suggests; sitting below means the honest model is smaller, and saying so improves rather than weakens the paper.
What makes a business case fail even when the numbers were right?
Usually the omission of internal cost. Somebody inside the organisation owns the implementation — configuring, migrating data, chasing crews, answering questions — and their time appears on no invoice, so it is routinely left out of the model. The project then meets its stated business case while feeling to everyone involved as though it did not, because a real cost was incurred that nobody budgeted. The second most common failure is claiming certainty about prevention. Stating an addressable fraction and acknowledging the rest is more persuasive than confidence, because the reader already knows the confident version is not true.
Your cost
times
Your frequency
times
Your addressable share
A Number Nobody Can Argue With, Because It Is Yours
No imported percentages, no payback period borrowed from another fleet, no efficiency saving that evaporates under questioning. Five event types, three years of your own records, an honest fraction in the fourth column, and a three-year all-in cost figure on the other side of the page. It takes an afternoon and it is the only version that survives a finance director asking where the number came from.
