Four endpoint conventions exist
An interval can include both endpoints, include the start only, include the end only, or exclude both. Each boundary changes the count only when that date is an eligible working day. NumUtility implements two of these combinations: the start is always considered, while a checkbox includes or excludes the end. It does not provide independent include-start and include-end switches.
Endpoint rules are separate from weekday eligibility. Including Saturday does not turn it into a Monday-Friday day; excluding a Tuesday removes a day that otherwise qualified. Establish the interval boundaries first, then apply the weekday test, then make any verified holiday adjustment.
Monday through Friday
Input: Monday September 7 to Friday September 11, 2026. Rule: include both endpoints. Process: all five dates qualify. Result: 5 working days. Interpretation: this is the tool with Include ending date on.
Input: the same range. Rule: include the start and exclude the end. Process: Monday through Thursday qualify and Friday is outside the interval. Result: 4 working days. Interpretation: turning off the checkbox does not exclude Monday. A system that excludes both boundaries would instead count Tuesday through Thursday, or 3.
Weekend boundaries
Input: Saturday September 12 to Tuesday September 15, end included. Rule: consider both endpoints but count Monday-Friday only. Process: Saturday and Sunday contribute zero; Monday and Tuesday contribute two. Result: 2 working days. Interpretation: inclusion and eligibility are distinct tests.
Input: Friday September 11 to Sunday September 13, end excluded. Rule: include Friday, omit Sunday, and exclude weekend dates in between. Process: only Friday qualifies. Result: 1 working day. Interpretation: switching end inclusion has no numerical effect when the end is already an ineligible Sunday.
Same-day intervals
Input: Wednesday September 9 to the same Wednesday, Include end on. Rule: the single date is both start and end and is included. Process: Wednesday qualifies. Result: 1 working day. Interpretation: an inclusive list can contain one date even though elapsed duration is zero.
Input: the same Wednesday with Include end off. Rule: the interval begins at the start but stops before the identical end. Process: no date remains. Result: 0 working days. Interpretation: same-day ranges expose endpoint semantics more clearly than long examples and should be part of every implementation test.
Words do not supply one universal rule
Terms such as “within,” “after,” “from,” “through,” and “by” can be defined differently by an organization, contract, statute, policy, or court. This guide does not interpret them. A mathematical +1 or −1 cannot determine which boundary the governing text intended. Obtain the responsible rule first and translate it into explicit include/exclude settings.
Also ask what happens when a boundary is a holiday, whether the next working day extends a deadline, and which locality supplies the calendar. Those are separate policy questions. The calculator's checkbox controls only the end date and does not resolve holiday or extension rules.
Reverse order and verification
For reversed inputs, the working-day function counts between the earlier and later dates and then negates the result. The end checkbox is applied during that normalized count, which may not match an intuitive description based on visual field labels. Enter dates chronologically when discussing endpoint wording; use negative output only as directional information.
Verify by drawing a row of dates, marking included boundaries with brackets and excluded boundaries with parentheses, then crossing out weekends and verified holidays. Common errors are adding one regardless of weekday, assuming both endpoints change together, and subtracting an excluded weekend endpoint. The focused working-days guide covers weekly counting and manual holiday adjustment; this page is limited to boundaries.
Interval notation makes the rule visible
Borrow mathematical interval notation for clarity: `[start, end]` includes both boundaries, `[start, end)` includes only the start, `(start, end]` includes only the end, and `(start, end)` excludes both. NumUtility implements `[start, end]` when the checkbox is on and `[start, end)` when it is off, before weekday eligibility is applied.
This notation does not decide whether a date is a working day. A Saturday can be inside the interval and still contribute zero. Put brackets around boundaries first, then cross out ineligible dates. Avoid writing merely “exclusive count,” because readers cannot tell which of two boundaries was excluded.
Endpoint effects can be zero, one, or two
Switching from both-inclusive to both-exclusive can reduce a count by two when both endpoints are eligible weekdays, by one when only one endpoint qualifies, or by zero when both are weekends or excluded holidays. Therefore the difference between conventions is not always a fixed two. Each boundary must be tested separately.
Example: Saturday through the following Sunday contains five Monday-Friday dates whether both weekend endpoints are included or excluded. Monday through Friday changes from five to three when both eligible boundaries are excluded. Monday through Sunday changes from five to four when Monday alone is removed. These comparisons isolate boundary effects without changing interior dates.
Communicating a chosen convention
Write a result as “five Monday-Friday dates, start and end included, before holiday adjustment” rather than simply “five business days.” The longer label makes assumptions reviewable. If the source rule excludes the start but includes the end, NumUtility cannot represent it directly with its current checkbox; adjust transparently or use a tool with separate controls.
When manually adjusting, add or remove an endpoint only if it is eligible under the working calendar. Preserve the unadjusted calculator output and adjustment note. This audit trail matters because two analysts can use identical dates and obtain different valid arithmetic under different stated conventions.
Consider a longer example to separate all layers. Input: Friday September 4 through Tuesday September 8, 2026, with Monday assumed to be a verified applicable holiday only for the illustration. Under both-inclusive boundaries, Friday and Tuesday are eligible weekdays, Saturday and Sunday are weekend dates, and Monday is removed by the manual holiday adjustment, giving two. Under start-inclusive/end-exclusive, Tuesday is omitted, leaving one. Under start-exclusive/end-inclusive, Friday is omitted, also leaving one. Under both-exclusive, neither boundary remains and the adjusted count is zero. The example shows why endpoint arithmetic must be performed on eligible dates rather than on the raw calendar-day count. NumUtility can directly reproduce the first two baseline conventions before the holiday subtraction, but not the latter two because it cannot exclude the start. If a formal instruction uses a different convention, do not manipulate the labeled input dates merely to force the desired number; document the unsupported boundary and calculate it transparently outside the tool.
For automated tests, create a matrix whose rows place the start on a weekday or weekend and whose columns place the end on a weekday or weekend. Run both supported checkbox states and same-day cases, then repeat in reverse order. The expected difference after toggling Include ending date is one only when the normalized ending boundary is an eligible weekday; otherwise it is zero. Add holiday adjustments outside that core matrix so a failed endpoint test is not confused with a calendar-data problem. In user interfaces, label the control with the exact boundary it affects and keep the result's convention visible after calculation. If results are exported, include a boolean such as `includeEnd` rather than relying on a surrounding sentence. Clear metadata prevents a later spreadsheet user from adding one again. When translating an unsupported include-end/exclude-start request, calculate the supported baseline first, remove the eligible start, and state the manual adjustment.