Before totaling deal amounts, choose the precision and rounding point, define the tie rule and direction, and match the unit finance records.
Key takeaways
- Label the precision shown in your forecast.
- Record where rounding happens in the calculation.
- Test exact ties against your chosen method.
- Trace a sample from raw deal amounts to the total.
Every revenue leader has seen it. The week before the forecast call, someone adds up the deal amounts in the pipeline, the total lands a few thousand dollars away from the number the team quoted last month, and the next twenty minutes go to figuring out why. Sometimes the answer is a deal that slipped or a discount nobody logged. Each deal was rounded a little differently, at a different point in the process, and the small differences added up.
The examples below use made-up amounts to show the mechanics. The rules apply to whatever your CRM, spreadsheet, or BI tool produces.
Why rounding matters more in a forecast than in a report
A closed-won report that is off by a few dollars is a minor annoyance. A forecast is different. It is a commitment about the future, and people make decisions based on it: hiring, spending, and what the board hears. When the total moves for reasons nobody can explain, trust in the whole process takes a hit, even if the underlying pipeline is fine.
Rounding causes trouble in a forecast for three reasons.
First, the same deal can appear with different amounts in different places. A rep may enter a value with cents, the finance system may store whole dollars, and the dashboard may show thousands. Each step rounds on its own, so the figure you total is not always the figure the rep quoted.
Many deals each rounded down by a fraction of a dollar give a total that is noticeably low. If your rounding has a direction, it will show up in the sum.
A default setting in a spreadsheet or a query function decides how a half-unit is handled, and nobody remembers choosing it. The first step in this process is to make those choices explicit.
Rule 1: Decide the precision, then round to it on purpose
The first rule is to choose the number of digits you will keep before you sum anything, and then round to that number of digits at one defined point in your process.
Precision is a choice about what your forecast is meant to show. A forecast in whole dollars is easier to read on a slide, and a forecast in cents is more useful when you reconcile against invoices. Either can be right. What matters is that everyone uses the same one.
The DAX ROUND function in Microsoft's documentation takes a second argument, num_digits, that sets the position of the rounding. The documentation explains that if num_digits is less than 0, the number is rounded to the left of the decimal point. That is useful when you want totals in larger units rather than exact amounts. The same page includes examples that round 2.15 up, to one decimal place and round 21.5 to one decimal place to the left of the decimal point. The point of those examples is not the specific numbers. It is that the precision is written into the formula, so anyone reading it knows what the result means.
Here is how that plays out. Suppose a rep enters a deal with cents in it, and another deal with a different fractional amount. If the forecast is in whole dollars, you can round each deal and then add the results, or add the raw values and round the total. With many deals, they can drift apart, and the drift depends entirely on where you round.
So pick one approach and write it down. Two common choices are:
- Round each deal, then sum. The total equals the sum of the displayed deal amounts, so anyone checking a single deal against the total gets the same figure. This is usually the better choice for a forecast call, because the room can trace the number.
- Sum the raw values, then round the total. This is more accurate in aggregate, because small roundings do not accumulate. It is harder to explain, because the visible deal amounts may not add up exactly to the total shown.
Either approach can be defensible. The mistake is switching between them from one quarter to the next, or between one report and another, without saying so.
If the forecast is in whole dollars, say so in the header. If it is in thousands, say that too.
Illustrative example: rounding to a larger unit
Suppose two illustrative deal amounts are 40 and 60, and the chosen unit is 100. Rounding each amount to the nearest 100 gives 0 and 100, totaling 100. Adding the raw amounts first also gives 100. The approaches match in this example; that does not establish that they will match for another set of amounts. Record which calculation your forecast uses.
Rule 2: Pick one tie rule and one direction, and write both down
The second rule concerns the hard cases: amounts that land exactly halfway between two possible rounded values. A deal whose amount ends in exactly one half of the rounding unit has to go to either the lower or the higher neighbor. They also show up in totals when you round many values.
There are two separate questions to settle. The first is how to handle ties, meaning values exactly halfway. The second is whether you round toward the nearest value or always in one direction.
For ties, Microsoft's documentation for DAX states that ties are broken by rounding half away from zero, also known as commercial rounding. If your tool uses a different tie rule, such as rounding halves to the nearest even digit, your totals can differ from a hand calculation in ways that are hard to spot. Know which rule your tool uses, and check it against a small test case.
For direction, the same documentation notes that to always round up, away from zero, use the ROUNDUP function. That is a different operation from ordinary rounding. If your policy is to show the most conservative or most optimistic number, that choice belongs in writing, not hidden in a formula.
Here is a concrete case to test. Take several deals whose amounts each sit exactly on a tie. Both methods are valid roundings, but the totals can differ by a meaningful amount across a whole quarter. A tie rule that looks like a technicality can change the number you present.
The practical step is simple. Add a short note to your forecast methodology page:
- The tie rule used (for example, commercial rounding, meaning half away from zero).
- The direction used (nearest, always up, or always down).
- The point in the process where rounding happens (at entry, at export, or at total).
Then test the rule once with a handful of values whose correct rounded results you already know. A short check now prevents a confusing discussion in front of the leadership team later.
Rule 3: Round to the unit your money actually moves in
The third rule is about the unit itself. Deal amounts are not always in neat increments. If you round without thinking about the unit, you can create totals that do not match what finance will eventually book.
Start with the number of fractional digits. PostgreSQL's documentation for its money type says that the range shown in the table assumes there are two fractional digits. If your deal data uses a different number of decimal places, for a currency or an internal system with a different convention, the storage and the rounding need to agree. Otherwise the amounts you total may not be the amounts you can actually record.
Next, consider rounding to a multiple rather than to a number of digits. Sometimes the business rule is not "whole dollars" but "nearest block," where the block is whatever size your commission or commitment policy uses. Microsoft's guidance for this case is to use a different function: to round a number to a specific multiple, for example to round to the nearest multiple of 0.5, use the MROUND function. The same idea applies in any tool. If forecast commitments are tracked in blocks, round to those blocks explicitly rather than simulating them with digit counts.
Payment systems show this clearly. Stripe's documentation on supported currencies says that for invoices where the amount is fractional after prorations, coupons, or taxes, Stripe automatically rounds that amount to the nearest number evenly divisible by 100. That is one vendor's behavior for one kind of invoice, not a rule for your forecast. It is still a useful reminder: the system that records the money may round in a particular way, and your forecast total should either match that behavior or explain the difference.
For revenue leaders, the lesson is to ask two questions before you total anything. What unit does finance book in? Does any of our data, tools, or billing system round to a different unit along the way? If the answers differ, decide which one the forecast reflects, and document the conversion.
There is a common trap around currency. If a deal is booked in one currency and forecast in another, the conversion and the rounding happen in some order. Convert first and round after, or round first and convert after, and the totals can differ. Pick the order that matches your reporting policy, write it down, and apply it everywhere. Do not let the order depend on which dashboard someone opened. Use the currency amount checks when tracing that path.
Putting the three rules together
Taken together, the rules give you a short checklist to run before each forecast call.
- Confirm the precision. Is the forecast in whole dollars, cents, or thousands? Is that stated on the page?
- Confirm the point of rounding. Are deals rounded when entered, when exported, or only when the total is calculated?
- Confirm the tie rule and the direction. Is it commercial rounding, and is the total rounded to nearest or always up?
- Confirm the unit. Does finance book in the same unit? Does any system round to a block, such as the nearest hundred?
- Reconcile one sample. Pick three or four deals, trace each from the CRM to the total, and check that the arithmetic matches your written method.
The last step matters most. A single traced sample catches many of these before the call. For a broader comparison, use the dashboard reconciliation checks.
It also helps to keep the raw values. If you round at entry and discard the cents, you cannot later check whether a total difference came from a deal or from rounding. Keeping the unrounded value in a column next to the rounded one costs almost nothing and makes reconciliation much faster.
Common mistakes to avoid
Several mistakes come up again and again.
Decide whether the stored value or the shown value is the source of truth, and make the two agree.
Rounding twice. A deal rounded when it is exported from the CRM, and rounded again when it is loaded into a forecast tool, can drift from the original. Each rounding step should have a reason. If two systems both round, make sure they use the same precision and the same tie rule.
Mixing methods across periods. If last quarter's forecast summed rounded deals and this quarter's sums raw values, the change between quarters includes a rounding effect nobody intended. When you compare periods, use the same method for both, or restate the prior period on the new basis.
Assuming the tool's default is the right one. Defaults are chosen for general use, not for your forecast policy. Check what your tool does with ties and with negative values, and confirm the result on a test case.
Treating rounding as a reason to stop looking. A rounding difference explains a small gap, but it should not excuse a large one. A value rounded once to nearest differs from its raw amount by no more than half the rounding unit. Look for a missing deal, a duplicate, or a stale amount.
What a good forecast total looks like
A good forecast total is one that anyone on the team can reproduce. Given the same export, they should get the same number. Given a deal, they should be able to say how it was rounded and why. Given a change from last week, they should be able to say whether it came from the pipeline or from the method.
When those pieces are in place, the forecast call gets easier. Instead of spending the first part of the meeting arguing about arithmetic, the team can spend it on what actually changes the number: which deals are real, which are at risk, and what needs to happen to close the gap. Rounding will never be the most interesting topic in a revenue meeting. Handled well, it stops being a distraction.
A short starting plan
First, write a one-page rounding policy. Keep it short. Include the precision, the point of rounding, the tie rule, the direction, and the unit.
Second, check your tools against that policy. Look at how your CRM exports amounts, how your spreadsheet sums them, and how your dashboard displays totals. Test each with a few known values, including at least one exact tie.
Third, reconcile one quarter's total from the raw data. Compare the result with the number you reported. If they differ, find the cause before moving on.
Finally, review the policy each quarter. New currencies get added, billing rules change, and new tools get adopted. A short review keeps the method aligned with how money actually moves through your company.
Today, take one deal from your latest export and trace its raw amount through to the displayed forecast total.
Common questions
Should I round each deal or only the total?
Choose the method that matches your forecast policy and use it consistently. Rounding each deal makes the total traceable to displayed amounts; rounding only the total avoids accumulating individual rounding differences.
What should I check when totals disagree?
Trace the raw amounts through each rounding stage. Check for different precision, tie handling, repeated rounding, missing deals and duplicates before attributing the gap to rounding.