How to Build a Project-Risk Register, One Row at a Time

A project-risk register is one row for each risk, and every row names four things: the line it touches, the party bearing it, the document that places it there, and what it would do to the cover. A fall of 21.06 per cent in revenue removes the cover completely at Tapti Crossing Infrastructure Private Limited, so the rows that move revenue sit at the top of its register.

What is a project-risk register for, and what is it not for?

Other people will act on the register, so the purpose has to be settled before the first row goes down. A project-risk register is a record of who bears what, sized wherever sizing is possible, kept so that two decisions can be made: what to negotiate, and what to fund. The two decisions are the whole purpose. Everything on the sheet either serves one of them or does not belong there.

The register is not a score, not a ranking of how worried anybody in the room is, and not a document that gets written once and consulted never. The three refusals do more work than they look like they do. A score invites a reader to compare two rows that were never measured on the same instrument. A ranking of worry records the temperament of whoever built the sheet. And a register nobody returns to is worse than no register at all. The existence of the file persuades everybody that the thinking has been done.

Think about a household that has just taken a home loan on one salary. If somebody in that household writes a list headed things that could go wrong, and the list says job, health, interest rate, nothing has been recorded. Now suppose instead they write: if the salary stops, the loan instalment of a known amount has to come from somewhere, and the somewhere is four months of savings. The second version is a register row. The row names the line the risk touches, names who absorbs it, and sizes it against a figure the household actually has. The finance version is the same discipline applied to a company with one asset.

Tapti Crossing Infrastructure Private Limited is that company here. The company is a single-asset toll road holding, formed to build and operate one crossing, with no other business and no recourse beyond the project itself. Having no recourse beyond the project is why a register matters more here than it does for an ordinary corporate borrower. Set it beside Harivansh Packaging Limited, a packaging maker with many customers and many lines: when one thing goes wrong at Harivansh Packaging, the rest of the business is still there. When one thing goes wrong at Tapti Crossing Infrastructure, there is no rest of the business. Every risk lands on the same small stack of cash.

The stack of cash is short enough to write out. Annual revenue is Rs 310 crore, operating cost is Rs 62 crore, so earnings before interest, tax, depreciation and amortisation (EBITDA) is Rs 248 crore. Debt serviceThe cash a borrower must hand over in a period, being the interest that has accrued plus the principal the schedule requires to be repaid in the same period. for the modelled year is Rs 182.70 crore, being interest of Rs 119.70 crore on borrowings of Rs 1,260 crore at the project's own contracted rate of 9.5 per cent, plus scheduled principal of Rs 63 crore. The cover ratioThe number of times one period's operating cash covers that same period's debt service. Here it is Rs 248 crore divided by Rs 182.70 crore. It is computed for one year and says nothing about any other. is therefore Rs 248 crore over Rs 182.70 crore, or 1.36 times, and the residualWhat is left of a period's operating cash once the lenders have been paid everything that period requires. It is the sponsors' cash, and it is the first thing any bad outcome consumes. is Rs 65.30 crore.

What is the procedure, in order?

Eight steps, run in this sequence, on any project with one asset and one stream of cash. The order is doing work. Steps two, three and four decide which rows exist at all. Run the sizing before those three are done, and a beautifully computed figure ends up attached to a row nobody can act on.

1. State the purpose of the register, before a single row goes downA record of who bears what, sized where sizing is possible, used to decide what to negotiate and what to fund. Write that sentence at the top of the sheet.
   Checking: could a reader tell from the sheet what decision it is meant to serve?
2. List by the line each risk touches, never by categoryFour lines only: revenue, operating cost, the capital already spent, and the right itself. Every row is filed under one of them.
   Checking: is every row under exactly one line, and is any row under none?
3. Name the party who bears the rowThe sponsors, the project lenders, the operator, the builder, the granting authority, or the project itself. One entry per row.
   Checking: has any row been left with no party at all in the column?
4. Name the document that puts it there, and write unallocated where nothing doesThe concession, the construction contract, the operating contract, the finance documents, or the word unallocated.
   Checking: is the word unallocated present anywhere on the sheet, and if not, is that really true?
5. Size the row in the project's own rupeesWhat it does to EBITDA, what it does to the cover, and what it does to the residual. Three figures, not an adjective.
   Checking: does the size cell contain two states and the movement between them?
6. Where a row cannot be sized, write down why, in the size column itselfOne of three reasons: the figure is not in the record, the event has no natural size, or the size depends on a period nobody has.
   Checking: is any size cell blank rather than carrying a reason?
7. Order the register by what each row does to the coverNot by how serious the row sounds when read aloud. State on the sheet which measure the ordering used.
   Checking: does a dull row with a large effect sit above a dramatic row with a small one?
8. Write, against every row, what would change itAn event that would send somebody back to the row. Without it the register is a document rather than a process.
   Checking: does every row name an event, and is any of those events merely a date?

Two of those steps are subtractions rather than additions. Step four adds a word for the rows nobody has allocated, and step six adds a reason where a figure is missing. Both exist so that the sheet can say what it does not know. A coloured grid can never do that.

How are risks listed so the list is complete rather than long?

Step two. Most registers are built by category, and a category list is somebody else's taxonomy borrowed wholesale. Construction risk, market risk, regulatory risk, environmental risk, counterparty risk: the words are respectable, the list is long, and not one of those rows can be sized. A category is not an event that touches a number. Worse, a category list has no natural end. Somebody can always propose a fifteenth category, and nobody in the room can say whether the list is now complete.

Listing by line fixes both problems at once. Cash reaches Tapti Crossing Infrastructure Private Limited through exactly one revenue line, leaves through exactly one operating cost line, sits in the capital already spent on the asset, and depends entirely on the right to operate and collect. A risk that touches none of those four cannot reach the project at all. A line list is therefore exhaustive by construction rather than by somebody's judgement.

The test applies to a category people love: reputational risk. The question is which of the four lines it reaches. If it changes what may be charged, or how many vehicles use the crossing, it is a revenue row and gets sized as one. If it forces additional expenditure, it is an operating cost row. If it puts the right in question, it is a right row. If it reaches none of the four, it is not a project risk at all, and the row has said something useful by failing the test.

How is the party bearing each risk recorded?

Step three. Against every row at Tapti Crossing Infrastructure Private Limited one party is written: the sponsors, the project lenders, the operator, the builder, the granting authority, or the project itself. Most rows on a real sheet end up with the last of those, and that last entry is the one that repays attention.

Writing the project in the bearing partyThe party whose own money moves when the event on the row happens. It is a question about cash, not about whose conduct caused the event. column looks like a shrug. Naming the project is the opposite of a shrug. The project is a vehicle with two sets of money in it and a strict order in which they are consumed. Rs 248 crore of EBITDA arrives, Rs 182.70 crore goes out as debt service, and Rs 65.30 crore is left. Anything that reduces EBITDA eats the residual first, all the way down, and only when the residual is gone does the shortfall reach the debt service the project lenders were promised.

A risk borne by the project is therefore a risk borne by the sponsors first and by the project lenders second, and writing project rather than a named party is exactly what makes that order visible on the sheet. A row that says the sponsors bear it has said something narrower and slightly wrong: the sponsors bear the first Rs 65.30 crore of it in the modelled year, and after that somebody else is in the loss.

How is the contract that puts it there recorded, or the fact that nothing does?

Step four. The fourth column names the document that places the risk where the third column says it sits. At Tapti Crossing Infrastructure Private Limited that is the concession for most rows, the construction contract for some, the operating contract for others, and the finance documents where a requirement to hold cash back is recorded. Each document is named as the place a requirement lives; how any of them allocates a risk in law is covered separately.

The interesting cell is the one where none of them applies. Write the word unallocatedWritten in the contract column when no document places the risk on any party. It does not mean nobody bears it. It means nobody has agreed in advance who does. and leave it in plain view.

Every risk is borne by somebody whether or not a document says so, and the rows nobody wrote down are precisely the rows that surprise people. Unallocated is therefore the most valuable word on the sheet. Consider the household again. Nobody signs an agreement about who pays if the roof leaks, and the roof still leaks, and the money still comes from somewhere. The absence of a document does not create an absence of consequence. The absence of a document creates an absence of warning.

There is a second reason the word earns its place. An unallocated row is the shortlist for the next negotiation. When a sponsor asks what to press for in a discussion with the granting authority, or a lender asks what to require before drawing, the answer is not the largest row on the sheet. The answer is the row that is large and unallocated. A large row already placed on a counterparty by a document has been dealt with once already.

How is a row sized in the project's own figures?

Step five, and this is the step that separates a register from a list. SizingTurning a row into figures the project already has: what the event does to EBITDA, to the cover ratio, and to the residual, stated as two values and the movement between them. means answering three questions in the project's own rupees. What does this row do to EBITDA? What does it do to the cover? What does it do to the residual?

Here is a row that is not sized: a deferral of what may be charged would reduce revenue. Read it again. The row names a direction and no quantity, it cannot be compared with the row above it, and it gives a reader nothing to act on. Every register in the world contains sentences like that one.

Here is the same row sized, on the figures for Tapti Crossing Infrastructure Private Limited. Revenue falls 10 per cent, from Rs 310 crore to Rs 279 crore. Operating cost is unchanged at Rs 62 crore, so EBITDA falls from Rs 248 crore to Rs 217 crore. Debt service is unchanged at Rs 182.70 crore. The schedule does not care what the crossing earned. The cover therefore moves from 1.36 times to Rs 217 crore over Rs 182.70 crore, or 1.19 times, and the residual moves from Rs 65.30 crore to Rs 34.30 crore.

A sized row states two figures and the movement between them, and a described row states neither. Only one of the two survives being read by somebody who was not in the room. Notice what the sized version also does that the described version cannot: it invites the obvious next question. How far can the revenue fall before the cover is gone entirely?

The revenue at which the cover disappears can be worked out exactly. The cover reaches 1.00 times when EBITDA equals debt service of Rs 182.70 crore, and reaching it needs revenue of Rs 244.70 crore, a fall of Rs 65.30 crore. On revenue of Rs 310 crore that is 21.06 per cent. The record rounds this to 21.1 per cent, and the rounding is worth stating rather than hiding: at a fall of exactly 21.1 per cent the arithmetic gives EBITDA of Rs 182.59 crore and a cover of 0.9994 times, marginally under rather than exactly at 1.00 times. The teaching point holds either way, and a rounded figure should be marked as rounded.

What goes in a row that cannot be sized?

Step six, and this is where most registers quietly give up. A row cannot be sized. The temptation is to leave the cell empty and move on, and an empty cell is read by every later reader as unfinished work rather than as a finding.

Write the reason in the same column where the figure would have gone. At Tapti Crossing Infrastructure Private Limited three reasons cover every unsized row on the sheet, and each of them says something different about the state of the project's information.

The first is that the figure is not in the recordWritten where a figure would require an input the project's own documents and model do not contain. It reports a gap in the information rather than a gap in the analysis.. Use of the crossing falling below what the sponsors expected is a genuine revenue row, and it cannot be sized on these figures. The record carries no traffic forecast at all. Not a low one, not a high one, none. The honest cell reads not in the record, and the honest next action is to go and get the forecast.

The second is that the event has no natural size. Anything requiring the asset to be spent on again is a capital row, and there is no single figure for it. The same description covers a small repair and a major rebuild. The buffer standing against it can still be written down: the debt service reserve of Rs 91.35 crore, being two quarters of the Rs 182.70 crore debt service, or 5.08 per cent of the Rs 1,800 crore already spent on the asset.

The third is that the size depends on a period nobody has. Any event affecting the ability to operate costs the project the revenue it would have earned for as long as it lasts, and this record fixes no concession period, no debt tenor and no schedule beyond the modelled year. Multiply an unknown rate by an unknown period and the product is a number with no content.

A reason in a size column is information and a blank is not, and the difference is that a reason tells the next reader what to go and find.

How is the register ordered once the rows are written?

Step seven. Every row at Tapti Crossing Infrastructure Private Limited is now written, and the sheet needs an order. Whatever sits at the top gets read and whatever sits at the bottom gets skimmed. The order runs by what each row does to the cover, and the sheet states that this is the measure used.

The result on this project is uncomfortable and correct. Revenue rows go first. Not because revenue is more important in some general sense, but because the cover is 1.36 times and a revenue fall of 21.06 per cent removes all of it. Revenue is the line this particular structure is thinnest against. A fall of 20 per cent sounds survivable when said out loud, and it takes EBITDA to Rs 186 crore and the cover to 1.02 times, leaving a residual of Rs 3.30 crore against Rs 65.30 crore in the modelled year.

Operating cost rows come second, and they are sized by exactly the same arithmetic. The EBITDA line does not distinguish between money that failed to arrive and money that had to be paid out. Rs 31 crore of additional operating cost gives EBITDA of Rs 217 crore, a cover of 1.19 times and a residual of Rs 34.30 crore. Rs 31 crore is 10 per cent of Rs 310 crore, so an added cost of that size lands exactly where a 10 per cent revenue fall lands. Capital rows come third, sized only against the reserve. The right itself comes last by size and first by seriousness, and that sentence goes on the sheet in those words.

Ordering by effect on the cover puts a dull row with a large effect above a dramatic row with a small one, and a reader who wants the other ordering has to say which measure they are ordering by. Naming the measure on the sheet is what lets a reader disagree with the order. The ordering is not a claim about what matters most in the world. The ordering is a claim about what this structure, with this cover, is most exposed to, and a reader is free to reorder by a different measure as long as they name it.

How is the register kept alive rather than filed?

Step eight. Against every row at Tapti Crossing Infrastructure Private Limited, write the event that would send somebody back to it. A revision notified by the granting authority. A notice raising an operating standard. The first full year of actual counts. An inspection that names work to be done. Each of those is a triggerAn event written against a row that requires the row to be looked at again. It is an occurrence, not a calendar date, so nobody has to remember to check., and each is an occurrence rather than a date in somebody's diary.

The distinction is not fussiness. A review date depends on a person remembering, and people leave, change roles and get busy. An event arrives whether or not anybody was expecting it, and if the register says which row that event belongs to, the row gets reopened by the event itself.

A row with no trigger can only be revisited by somebody who chooses to revisit it, so a trigger written against a row is what turns a register from a document into a process. The missing trigger is also the honest answer to why so many registers are stale. Stale registers were never wrong. They simply had no mechanism for becoming right again.

What does the finished sheet look like for this crossing?

Run all eight steps once, on Tapti Crossing Infrastructure Private Limited, filling in every column the record supports and refusing to fill in the ones it does not. Five rows come out of it, and each figure below is recomputed from the project's own numbers rather than carried across from anywhere.

Read the sheet as rows and it is a list. Read it as three counts and it becomes an account of what this project actually knows about itself. Two rows of five are sized. Three of five are not, and each carries a different reason, so no two of them are missing for the same cause. One of five is unallocated, and it is a revenue row. On a structure this exposed to revenue that is the single most useful sentence on the sheet.

The three counts describe the state of the project's information better than any individual row does, and that is why they belong at the foot of the sheet rather than in somebody's head. A register where five rows of five are sized is a project with a complete model. The register for this crossing is not that, and the sheet says so in a form a reader can check.

One row deserves a closing note. A careless build would invent something there. The row on the right itself is last on the sheet by effect on the cover and first by seriousness, and both halves of that sentence are written on the sheet in those words. Sizing that row would require a period, and this record fixes no concession period, no debt tenor and no schedule beyond the modelled year. A register that quietly assumed six months of interruption would have produced a figure, and the figure would have been fiction dressed as arithmetic.

Why is a likelihood column the thing to leave out?

Most readers have seen a register built the other way, as a grid of likelihood against impact with every row coloured, presented as a risk assessment. The likelihood column was not forgotten. Leaving it out is a decision, and the case for that decision rests on what a second reader can check.

Ask what can be checked on such a grid. The likelihood column was produced by judgement nobody wrote down. The impact column was produced by adjectives. The colour was produced by multiplying the first by the second. Not one of those three can be reproduced by a second person working from the same documents. A sheet like that cannot be audited, corrected or argued with.

A coloured grid carries authority it has not earned, and the cost of that authority is paid by the rows it colours green. On this project a 20 per cent revenue fall takes the cover to 1.02 times and the residual to Rs 3.30 crore, and if somebody judged it unlikely, that row is green and gets read last. The sized version of the same row cannot be hidden that way. A cover of 1.02 times is on the sheet, and any reader can see how close to 1.00 times that is.