ManageTime

Article

Which email to answer first: oldest, newest, or neither

Scan newest first, then reply oldest first. One pass down the new mail shows what has already resolved itself and what genuinely cannot wait. After that, work up from the bottom so nothing ages quietly. No study has raced the two orders, so treat that ranking as practice rather than evidence. This article covers what the research on reply order actually found, why a late reply costs the sender more than it costs you, and how to pick a first email without re-sorting the inbox every morning.

Should you answer emails oldest first or newest first?

Monday, nine in the morning, forty unread. The top of the list is the loudest, so you start there and work down, which means the message from Thursday afternoon gets read for a fourth time and answered for the first on Wednesday. Most people never decide this order. It falls out of whichever way the client happens to sort, and the client sorts newest first because that is what a mail client has always done.

Split the job in two and both halves get easier. Scan first, in whatever order the inbox offers, reading sender and subject only, and pull out the few messages that have to move today. A good share of new mail closes itself in that pass: the question somebody else already answered, the meeting cancelled two replies later, the thread where you were only ever in copy. Then stop scanning and start replying from the oldest thing still standing.

Nothing in the research raced oldest against newest, and this site will not dress a preference as a result. What Heejung Byun and David Kirsch did measure, in Academy of Management Discoveries in 2021, is what people already do. Working through one organization's email corpus across five years, they found reply priority running on four heuristics. Order of arrival is one. The other three are what the subject line implies without opening the body, how much effort a reply would take, and what the sender appears to expect about timing. Arrival order is on that list. The oldest-first habit is not a trick somebody sold you; it is one of the rules your inbox was already running.

Pure arrival order has one failure, and it is the whole reason for the scan. A queue worked strictly from the bottom has no way of learning that something landed an hour ago with a deadline at noon. The scan buys that knowledge cheaply, because it reads headers rather than messages. In practice that pass tells you more about the day than the same minutes spent answering whatever arrived most recently.

Both halves cost something, and the costs are worth naming before you commit to the arrangement. The scan tempts you to start answering inside it, which is the failure that turns five minutes of triage into an hour. The reply pass is slower than cherry-picking, because the oldest mail is usually the mail you were avoiding, and you meet it while you are fresh rather than at half past four. The email pillar covers the sessions all of this sits inside. On the days your scan keeps returning more than a handful of urgent things, reach for the Eisenhower Matrix, which sorts work into four quadrants by urgency and importance.

Does the order you reply in actually matter?

Mostly it matters to other people. Whichever order you work, roughly the same messages get answered by the end of the session, and the minutes you save by optimising the sequence are not where the gains are.

Byun and Kirsch's second study is about the other end. They looked at what happens when somebody replies out of step with the timing their organization has settled into. That settled pattern gets a name in the paper: organizational timing norms, the local expectations about when a reply is due that nobody wrote down and everybody absorbed. Members whose replies fell out of sync showed up as a bottleneck in the organization's communication, and the authors report that the pattern predicted employee turnover. Their abstract does not say whose turnover, so read it as evidence that out-of-sync replying tracks something real rather than as a warning about your own job.

Turn that around and the sequence gets a purpose it did not have. The question stops being which order is optimal for your morning and becomes whether anything in the queue has sat long enough that somebody downstream has stopped being able to work. Oldest-first surfaces those by construction. Newest-first, run on its own, is the arrangement that reliably buries them, because a message only gets further from the top.

Timing norms also explain why general advice about response times never quite fits your desk. A norm is local. One team treats same-day as generous and another treats an hour as late, and neither published the standard it is holding you to. You can read yours off the mail you receive instead of guessing: watch how long your own colleagues take on an ordinary request, and you have the figure your replies are being judged against. That number, not a rule from a productivity article, is the one worth matching.

Do colleagues think less of you for a slow reply?

This is the fear the whole sorting problem is built on, and it has less support than it feels like it has.

David Westerman and Justin Walden ran two experiments on how people form impressions of a managerial email exchange. They manipulated time cues, the timing information a message carries, alongside source cues about whether a human or an automated sender produced it. The two data collections sat three years apart, one before the pandemic and one well after it, with ChatGPT released in between. Results held across both. Neither cue moved the impressions participants formed. The authors go further in their own reading: attention-check data suggested time cues may no longer predict impression formation in workplace communication. Business and Professional Communication Quarterly published the study in 2025.

Two limits before that goes anywhere near your inbox. Participants judged a simulated exchange they had nothing riding on, which is a weaker test than a colleague waiting on an answer they need to finish something. And no measured effect on impressions is not the same as no cost. Byun and Kirsch's bottleneck is a cost. It lands on your work rather than on what anyone thinks of you.

If speed is the wrong axis to sort on, the useful one is what the message actually is. Andre Lanctot and Linda Duxbury built and validated two scales for exactly that, in the Journal of Computer-Mediated Communication in 2022. One is a six-item measure of whether you see an email as important. The other is an eight-item measure of whether you see it as important and urgent together. Both predicted work-role overload, the sense of having more assigned than your day can hold, even once the usual email-demand measures were taken into account. Their starting point is worth repeating, because it is the reason so much inbox advice is guesswork: nobody had written down what knowledge workers use when they triage.

The separation is the part that transfers to a Monday. Important and urgent are two different questions, and an inbox sorted by date answers neither of them. Ask both in the scan, off the subject line, before you have opened anything and committed yourself to a reply. Would you rather make that call the evening before than at nine in the morning? The Ivy Lee Method, which has you name tomorrow's six tasks tonight and work them in order, moves the decision to a quieter hour.

How does your working style change the answer?

Which half of this you get wrong depends on how you already work. Readers closest to this site's Architect style build the scan into a routine and then keep refining it, turning four minutes of triage into a system that needs maintaining. Sprinters skip the scan and reply hard from the top. Their problems collect quietly at the bottom of the queue. Improvisers answer whatever is on screen and lose the queue as a concept, which is survivable at twelve messages a day and not at eighty. Visionary-leaning readers run the scan well, then answer the interesting thread instead of the old one.

Role sets the outer edge. On a support desk or a dispatch queue the order is the job, a tool already assigns it, and none of this is yours to choose. If your mail runs to a dozen messages a day, the whole question is too small to build a system around. And when the same three names generate most of your queue, the fix is a conversation with them about how they flag things, not a sorting rule you apply afterwards.

A week of the split will settle the argument faster than another article about it.

So run it for a week and keep one note. Every time the scan turns up something at the top that genuinely could not have waited until you reached it from the bottom, write down what it was and who sent it. Most people finish the week with two or three entries and a clear pattern in the sender column. The working-style self-check points at the methods most likely to survive contact with your own week.

Frequently asked questions

Should you answer emails oldest first or newest first?

Scan newest first to find what has resolved itself or cannot wait, then reply starting from the oldest message still standing. No study on this page raced the two orders against each other, so treat that ranking as practice rather than a finding. What the research does show is that replying out of step with what your organization expects makes you a bottleneck for the work behind you. A queue worked from the bottom is the one that surfaces those messages first.

Does the order you reply to emails in actually matter?

It matters mostly to the people waiting on you rather than to your own day. A study of one organization's email across five years found that members replying out of sync with local timing expectations formed a bottleneck in the organization's communication, and reports that the pattern predicted employee turnover. Your own throughput barely moves with the order you pick. What sits behind a two-week-old unanswered message does.

Do colleagues think less of you for a slow reply?

Less than the worry suggests: two experiments run three years apart found time cues did not change how participants judged a simulated email exchange. The authors report that attention-check data pointed the same way, suggesting time cues may no longer predict how impressions form in workplace communication. That is a judgment of a simulated exchange by people with nothing riding on it, not a promise about a colleague who needs an answer. A slow reply still costs the work waiting behind it.

What do you do with old emails you never answered?

Answer the few that still have work waiting on them, and tell the rest plainly that you are not going to. A message old enough to be embarrassing has usually been resolved another way already, and a short note saying so closes it faster than a full reply would. For the handful that are still live, name the delay first and then answer. Then set a floor, such as nothing older than two weeks staying unanswered, so the same pile does not rebuild behind you.

Sources