top of page

HVAC RFI Response Statistics: Separate Completed Durations from Open Ages

Oct 26
4 min read
AI-generated illustrative scene: blank document stacks with coloured page flags on an engineering desk. Not product or project evidence.

AI-generated illustrative scene, not a product screenshot or a completed-project photograph. The worked calculation below uses explicit example inputs.


An RFI dashboard can show a reassuring response time while the questions blocking procurement remain unanswered. The problem is often the population behind the statistic. A median calculated only from answered RFIs describes that answered subset. It does not establish how long every issued RFI will take, and it does not describe the current age of questions that still have no qualifying response.


Use separate measures for completed response duration and open age. This article demonstrates a small hypothetical HVAC register, not Nexora customer records or a contractor performance assessment. It does not define contractual response obligations, decide liability for delay or replace a programme review. The reporting rules must be agreed before comparing teams or reporting periods.


1. Define the event and the clock


Decide what counts as a qualifying response. An automated acknowledgement, a request for missing information and a technically usable answer may be different events. Keep first acknowledgement, substantive response and implemented closure as separate fields when the workflow needs them. Do not silently replace one date with another to improve a metric. A response can arrive before its instruction has been checked and incorporated.


For this example, duration means elapsed calendar days from formal issue to the first recorded qualifying response. All timestamps use one consistent time zone and the same time of day, so the example durations are whole days. There are no stopped-clock periods. The snapshot is 26 October 2026. Real reporting should state timestamp precision, partial-day treatment and any agreed pauses rather than mixing calendar days with working days.


2. Calculate the answered subset transparently


Assume ten RFIs have been issued in the defined register population. Six have qualifying responses by the snapshot, with durations of 2, 3, 4, 5, 6 and 20 calendar days. Four remain unanswered. The answered subset is already sorted here; in a real export, sort a copy or use a correctly scoped statistical function without altering the underlying event records.


Answered count = 6; issued population = 10.


Completed response mean = (2 + 3 + 4 + 5 + 6 + 20) / 6 = 6.67 days.


Completed response median = (4 + 5) / 2 = 4.50 days.


Answered proportion at this snapshot = 6 / 10 = 60%.


The median is less affected by the 20-day observation than the mean, but neither statistic is a verdict on the process. Display the sample size and, for a small set like this, the individual durations or range. A four-and-a-half-day median does not mean every answer arrived within that time. The six observations also provide no basis for a promised response time on the next RFI.


Six completed response durations have median 4.5 days. Four still-open RFIs have median current age 21.5 days; their final response durations are unknown.

Original illustrative quantity/statistics diagram. Assumptions and scope are stated in the worked example.


3. Keep unanswered RFIs out of completed-duration arithmetic


Assume the four unanswered RFIs have open ages of 12, 18, 25 and 31 calendar days at the same snapshot. Their final response durations are unknown. Each has remained unanswered for at least its recorded age, and each final duration will be determined only if the qualifying event occurs. In statistics this resembles right-censored duration data; the NIST reference explains that general distinction using reliability examples, not RFI contracts.


Open count = 4.


Median current open age = (18 + 25) / 2 = 21.50 days.


Oldest currently open RFI = 31 days.


Open-age range = 12 to 31 days.


Do not insert four zeros into the response-duration column. That would imply four immediate responses that did not occur. Do not insert the four open ages as if those were final durations either: a current age is not a completed event time. Both shortcuts change the meaning of the dataset. Leave the completion field genuinely uncompleted and calculate open age in a separate, explicitly labelled field.


4. Make spreadsheet behaviour visible


For the six confirmed numeric durations placed in cells H2:H7, =MEDIAN(H2:H7) gives 4.5. Microsoft documents that referenced empty cells are ignored by MEDIAN, while numeric zeros are included. Consequently, a formula that substitutes zero for every unanswered item can silently distort the result. An empty-looking text string and a true numeric zero should not be used interchangeably as status codes.


Build a clear eligibility flag for the response statistic: valid issue timestamp, valid qualifying-response timestamp, response no earlier than issue, and membership in the reporting population. Review negative intervals, duplicate identifiers, withdrawn RFIs and reopened questions separately. If no eligible completed records exist, report the metric as not available with a count of zero completed responses; do not display a zero-day median.


5. Use the pair of measures to choose action


The practical summary is six completed responses with a 4.5-day median and four unanswered items with a 21.5-day median current age. Those numbers describe different populations and should remain side by side. Follow the older open items back to their actual required-by dates, responsible parties and affected activities before deciding what deserves escalation. Age alone does not prove a critical-path delay.


For week-to-week comparisons, keep the event definition and population rule consistent. An issue-date cohort, a response-date cohort and a current open snapshot answer different questions. Preserve each dated export so a reviewer can reproduce the figures. The outcome should be a small, honest decision aid: what has received a qualifying response, what is still waiting, and which unresolved decision needs attention next.



Continue the calculation review




Sources and calculation basis


Primary references checked on 26 October 2026. The worked examples and diagrams are original illustrative calculations, not manufacturer selections or project approvals. Apply the stated limitations and the requirements governing the actual project.




Comments

Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page