You follow up without nagging by making each contact predictable before it lands and different from the one before. Nagging is a pattern, not a tone: the same ask, at intervals nobody agreed to, carrying nothing new and cheap to set down one more time.
That’s an unsatisfying answer when what you wanted was a sentence to paste into a message. So here’s the practical consequence. Almost everything written on this question works on the wording of one email. The wording is real, and it’s the smaller half. Two people can send the identical message on the same Thursday, and one of them is following up while the other is nagging, on the strength of what happened in the six weeks around it.
- The judgment lands on a sequence, and it lands on a person rather than a thread.
- Your sense of whether you’re imposing is a steady bias rather than evidence, and it points the wrong way.
- Saying when you’ll come back is worth more than any amount of rewriting the message you come back with.
- Name what a chase is holding up. When you can’t, you’re sending it to settle your own nerves.
Three neighboring questions are settled elsewhere, and this piece assumes them. How long to leave a silence before the first follow-up is in how long to wait before following up. What the message itself should say is in the follow-up email that gets answered. Whether it reads as warm is its own question, and warmth works strangely in text: the polite reminder email has that. This piece is the space around all three. That’s where nagging happens.
What separates persistence from nagging
Ask somebody to describe a nag and you get a description of a person. That’s the clue. The verdict builds up across a run of messages, and three things build it.
The ask stops changing. A second message that’s the first message in warmer words has already been read and already been set down. Send it again and you mainly show that your messages can be set down. So change the content rather than the temperature: a new fact, a smaller question, a different person, a consequence that’s arrived since. The ladder of what a chase can change, and the order to climb it in, is worked out in managing client document requests without the chasing, for the case where you’re waiting on an errand rather than a decision.
The arrival is unannounced. A message that lands on a day the reader was expecting nothing is an interruption, whatever it says. The same message on a day you named two weeks ago is an appointment. Hardly anybody does this, and it’s the cheapest thing here.
Deferring is free. A request with no consequence attached can sit forever at no cost, and by the third one the reader has learned that. People are most reluctant about this part, because attaching a consequence feels like a threat. It’s usually a small, true, undramatic fact. The slot goes. The work moves a week. You file the extension, or you call instead.
There’s a fourth, and it hides from inside any one thread. Nagging is counted per person, one client at a time. You have the engagement letter outstanding with Dana. You have the missing bank statement too, and the question of whether the March work is going ahead. Three threads, three proportionate follow-ups this week, each one defensible on its own. Dana got three chases from you in five days. From where she sits, that’s a week of you.
The judgment you are making badly
Before you decide you’re following up too much, take the measurement. People are systematically wrong about this, and the error runs in one direction.
Vanessa Bohns of Cornell reviewed a decade of that work in (Mis)Understanding Our Influence over Others, published in Current Directions in Psychological Science in 2016. Across the studies she summarizes, participants made requests of more than 14,000 strangers: borrow your phone, sponsor me for a race, fill in this questionnaire. Before each one they predicted how many people would agree. The error is large and it is consistent. It holds across different samples and different requests, and Bohns puts the average size of it plainly: “on average, requesters underestimate compliance by 48%”.
The reason matters more than the number. Bohns attributes it to “requesters’ failure to appreciate the awkwardness of saying ‘no’ to a request”. You stand inside your own head, where the cost of asking is vivid. They stand inside theirs, where the cost of refusing is vivid too, and mostly larger than you imagine.
Now the complication, because this is where the finding stops flattering you. The underestimation is a face-to-face effect. In the same review, participants who made the identical request by email “overestimated compliance”. Asking in writing is weaker than asking in person by a margin most people miss, and writing an email from scratch goes into that properly.
Put the two together and you get something useful about following up. What you’re wrong about is mainly whether your first message ever landed. In the research those are errors in opposite directions. In practice they push the same way, because both of them argue against sending the second message. Over-worrying about the chase while over-trusting the original is the exact combination that leaves work sitting.
One more result belongs here, because it covers the case people find hardest. Daniel Newark, Francis Flynn and Vanessa Bohns tested what a refusal does to the next ask, in Once bitten, twice shy, published in Social Psychological and Personality Science in 2014. Participants approached strangers with a request, then made a second request of the same people whatever happened first. They predicted that a refusal would drop compliance on the second ask by 16 percent. Compliance actually rose by 10 percent. The authors read it this way: refusing somebody twice is harder than refusing them once. The person doing the asking, meanwhile, treats the first no as a fact about the other person’s character.
That’s strangers and small favors, so take it as direction rather than as a rate you can apply to clients and engagement letters. The direction is clear enough. An explicit no leaves the door open, and so does a silence.
The cost that the reply rate hides
Reminders still cost something, and one field experiment measured both sides of the ledger properly.
Mette Trier Damgaard and Christina Gravert worked with the Danish charity DanChurchAid in the summers of 2013 and 2015, and published the result as The hidden costs of nudging in the Journal of Public Economics in 2018. In the first experiment, some people received a donation request on its own. Others received the same request, then an unannounced reminder a week later.
The reminder worked. Giving rose from 0.35 percent to 0.53 percent, an increase of about two thirds. In absolute terms that’s modest, and it did not survive every specification the authors ran. Unsubscriptions rose too, from 2.1 percent to 3.7 percent, and that difference was the firm one. One extra message, sent once, pushed a meaningful share of the list to leave permanently.
That’s a charity mailing list, and Danish donors reading a summer appeal. What transfers is the shape rather than the rates. A reminder produces a visible gain and an invisible loss, and only one of them shows up in your reply rate. The people who answer your third chase are in front of you. The ones who quietly reclassify you stay out of sight. That asymmetry, and what it means for how a reminder should be timed, is taken further in email reminders that arrive while there is still time.
Two details are worth carrying out of that study. The reminder that cost them was unannounced. And around 70 percent of the unsubscriptions happened on the day the email arrived, so the damage was a reaction rather than a considered decision. Irritation is fast, and it moves before your reasons arrive.
The follow-up they already know is coming
The same pair ran a second experiment, and its result is the one worth building a habit on.
Subscribers were used to hearing from the charity about once a month. One group got the standard solicitation. Another got the identical solicitation plus one sentence, telling them that over the coming three months they would receive only one email. Everything else stayed the same: the ask, the wording of the appeal, the cause.
Unsubscriptions fell from 0.49 percent to 0.30 percent, a reduction of 39 percent. Giving was unaffected.
Read what that isolates. The wording stayed where it was. At the point the opt-outs were counted, the frequency stood where it was as well. The one thing that changed was what the reader expected to happen next, and that alone removed close to two fifths of them. Annoyance runs partly on a forecast of messages to come, and the forecast is something you can edit directly.
For a small practice that turns into one habit, and it belongs on the first message rather than the follow-up. When you ask a client for something, say when you’ll come back if you haven’t heard.
If I haven’t heard by the 19th I’ll check back once, and then leave it with you.
That sentence does three things at once. It tells the reader what the follow-up is, so when it lands it reads as you keeping your word rather than losing patience with them. It caps the sequence, so answering slowly stops meaning an indefinite drip. And it settles the question for you in advance. On the day, you’re a person keeping your word, because you already told them you would.
Then be back on the 19th. A date you name and let pass teaches the reader that your dates are decorative. Every message you send that person afterward is worth a little less.
Three chases in a week is one message
The per-person arithmetic from earlier has an obvious fix, and almost nobody applies it, because email is organized by thread and threads hide the count.
Before you send any chase, look at what else is outstanding with the same person. If anything is, the chases are one message.
Dana, three small things sitting with you, and none of them urgent today:
The engagement letter, which I need signed before we can start. The bank statement for March, whenever it surfaces. Whether the March work is going ahead, which decides how I plan the next two weeks.
If the answer to the third one is not yet, say so and I’ll stop asking about the other two until it is.
That’s one interruption instead of three. It also tells the reader more than three separate messages do, because it shows how the items relate, and the relation is often what unblocks them. Dana may be sitting on the engagement letter precisely because she’s still deciding about March. While the two live in different threads, that stays invisible to both of you.
There are two limits on this. Keep a genuinely urgent item on its own, because bundling dilutes urgency. And batch within one person’s own work: if the bank statement is really the bookkeeper’s and the March decision is really the owner’s, one message to both is one message that neither of them owns. The rule is one message per person per occasion, one per human being rather than per client organization.
So the constraint is visibility rather than writing skill. Can you see, before you write, everything one person currently owes you? Almost nobody can, because the record of what you asked for lives in a sent folder ordered by date. Where that debt should actually live, and how to keep the list without keeping it by hand, is tracking what you are owed.
Asking for the date instead of the thing
The other structural move changes what you’re asking for. Then your next contact has something of theirs to refer to.
David Nickerson and Todd Rogers tested the strongest version of this during the 2008 United States presidential election and reported it in Do You Have a Voting Plan?, in Psychological Science in 2010. With 287,228 people in the experiment, callers ran one of three scripts. The standard one reminded people about the election and their duty to vote. The second added a question asking whether they intended to vote. The third added three more: “what time they would vote, where they would be coming from, and what they would be doing beforehand”.
The standard call left turnout where it was. Asking people to predict their own behavior left it there as well. The three planning questions raised turnout by 4.1 percentage points among the people actually reached. In households with a single eligible voter, where anybody else who might have prompted the same planning was absent, the gain was 9.1 points.
That’s voting, and a phone script. What transfers is the distinction the experiment isolates, because it is precise. Reminding somebody left turnout flat. Getting them to say they would do it left it flat too. Getting them to work out when and how is what moved it.
So the ask that ends a chase is usually the plan rather than the thing itself.
Not asking you to dig it out today. Is this a this-week thing or a next-week thing?
An answer to that is cheap to give, which is why you tend to get one. It also changes the next message completely. Without a date, the chase you send reports how long they have taken. With one, your message says: you thought next week, and it’s now the week after, is it still coming? That’s a reference to something they said. People answer their own commitments far more readily than they answer your requests.
Ask for a date twice and hear silence twice, and you’ve learned something a fourth reminder would never have told you. The problem sits somewhere other than memory. That’s a different situation with a different repair, in chasing a client who will not reply.
The chases you should not be sending
Part of the worry about nagging is well founded: some messages are better held. Three kinds.
The one that is about your nerves. The test is simple and slightly uncomfortable. Name what changes if they answer today rather than Friday. If nothing does, the message is for you, and it buys a few minutes of relief followed by the same waiting. Write it, then leave it in drafts until the day it would be true.
The one that spends credibility you will want later. If everything you send carries the same urgency, the reader learns to discount all of it, and the day something genuinely can’t wait you’ll have no way to say so. Chasing has a budget, and it’s drawn against one account for a whole relationship. Spend it on the item you could have absorbed yourself, and you arrive at the deadline with nothing left to spend.
The one that should not be an email. Sometimes what you need is a decision the person can’t make alone, or an answer that’s awkward to write down. Another message stands in for the right move there. A short call settles in ninety seconds what four emails leave open, and offering one reads as help. Where the outstanding thing is a meeting time rather than a decision, the whole exchange has a different shape: see scheduling without the email chain.
Saying that it landed
The habit that most reliably keeps a persistent person sounding reasonable is about what you do when something finally arrives.
Confirm it, by name, the same day. Name the item, so the reader can see you opened it: got the March statement, that’s everything I needed for the quarter. It takes eight seconds and it does two jobs. It closes the loop for the other person, who can now tell that their effort landed and that the question is done. And it makes your next chase credible, because you’ve shown that your requests track real states of the world rather than a script that runs whatever anybody does.
The silent recipient makes a bad chaser, for mechanical reasons rather than manners. Where acknowledgment is missing, your messages and what anybody does about them look unconnected from the other side. That’s precisely the impression a nag creates. The person who confirms receipt is visibly keeping count, and somebody keeping count can be trusted to stop.
The same logic runs the other way, and that’s the part almost nobody does. When a thread resolves on its own, say so and take it off the table.
Ignore my question about the March scope, we covered it on the call. Nothing outstanding at your end now.
Threads you walked away from stay open in the other person’s mind, and every one of them adds to the sense that dealing with you means an accumulating pile.
What to do if you already are one
Take the worst case. Several unanswered chases, on the same subject, over some weeks. And a growing suspicion that your name in an inbox now produces a small sinking feeling. There’s a way out, and it’s one message rather than a better-worded fifth attempt.
Send one message that closes the loop from your side.
I’ll stop crowding your inbox about this. I’m putting the March work down as on hold and letting the slot go. Should it come back around later in the year, drop me a line and we can start again from here. No reply needed to this one.
Three things are true about that message. It costs the reader nothing, which is why it gets answered when nothing else has. It hands them the graceful version of the no they may have been struggling to write, which is one of the more common reasons for a long silence. And it resets your standing, because you’ve shown that your chasing has an end.
As often as not, what comes back is the decision you asked for six weeks ago, with an apology attached. Either way the question is settled and you can put the thread down.
Keep the wording level as you go. A sequence that gets sharper each time is a negotiation with somebody who is absent from the negotiation, and you’re the only person paying for it.
What Point counts while you wait
All of this is easy to agree with and awkward to run, for a mechanical reason. What you’re owed sits in a sent folder sorted by the day you wrote, so the pattern you’re judged on is the one thing that folder hides. You learn that Dana got three chases in five days when Dana mentions it.
Point is a full email client, so what you sent and the answer you’re still waiting on sit in one place. When the next move belongs to somebody else, you say so on the thread in ordinary words as you send. The wait turns into a dated task sitting with the rest of your workload, carrying one link back to where it came from. On the day itself, a conversation still sitting unanswered comes back up your feed with a prompt on it.
For this problem the useful part is the view. Anything you’ve handed off is tracked and lands on one list, so you spot three open items with the same client before you write rather than afterward. Loops that resolved on their own stop counting against you as well. Work that already looks finished gets marked as such with the reason showing, held for your agreement. When you’re waiting on an event rather than an answer, a standing request watches for it and tells you once.
You decide separately, for each type of work, how far Point goes alone. The dial runs from suggest-only at one end to fully handled at the other, and review is the setting it all begins on. A plain-language record of what Point did is kept in sequence. Undo covers what Point did, worked back from that record, with the exception you’d expect for mail that has already reached somebody else’s server. What else Point does with an inbox is set out on the benefits page.
Writing the chase normally stays with you, and that division is the right one. A conversation that stopped is a relationship still in motion, so the line that restarts it usually names something only you were there for.
Common questions
How many times can I follow up before it becomes nagging?
There’s no number, because the reader is measuring something other than the count. They’re measuring whether anything changes between messages, and whether the messages arrive when they expected them. Four announced follow-ups, each carrying a different ask, sit better than two unannounced identical ones. So use a discipline rather than a number. Send a follow-up whenever you can name what’s new in it. When you run out of new, you’ve reached the end of the sequence, and the next message closes it out.
Is it nagging if I have already asked three times and heard nothing?
You’re fine. Three unanswered messages tells you the ask is failing, which is a different problem and a fixable one. Something is off with the size of the request, the person you’re asking, or the channel. A fourth message in the same shape confirms that rather than fixes it. The productive move at that point is to make the ask smaller, ask for a date instead of the thing, or offer the exit.
Does following up actually get better results, or does it just annoy people?
Both, measurably, which is why the pattern matters more than the intent. In the field experiment that tracked both sides of it, a single unannounced reminder raised the action rate and the permanent opt-out rate at the same time. The gain and the cost are both real. What you control is the ratio. Three levers move it: announcing the return, changing the ask each time, and batching everything you need from one person into one message.
What if I have already been nagging somebody for weeks?
Close the sequence with one message rather than continue it with a better-written one. Say what you’re assuming in the absence of an answer, say that no reply is needed, and mean both. That message is cheap to answer, which is why it often gets the answer nothing else did. And it establishes the one thing your previous messages left open: that your chasing has an end.
Should I apologize for following up again?
No, and confidence has little to do with it. An apology adds a second thing to be dealt with before anybody can deal with the first. It also tells the reader you regard the request as skippable, which is an odd position to take on your own work. To lower the temperature, change the structure rather than the manners: shrink the ask, or name what happens in the absence of an answer. Which softeners help and which quietly undo the message is worked through in the polite reminder email.
Can I have software send the follow-ups so I stop worrying about it?
Automate the part that’s actually failing, which is remembering, and be careful with the rest. A tool that fires the same message on a schedule is a machine for producing exactly the pattern this piece is about: unchanging, unannounced, disconnected from anything real. What’s worth handing over is the return of the thread to your attention, with the whole context attached, at a time you chose when you still remembered what you asked for.
The short version
Nagging is something a sequence becomes, and it’s judged per person rather than per thread. That’s why three defensible chases in five days add up to one impression. Two measured errors push you the wrong way on it. You overestimate how much your asking imposes, and you overestimate how well the email you already sent landed. Both of those argue against the follow-up that’s probably needed. So follow up more often than feels comfortable, and change something every time you do.
The cheapest protection is one sentence on the original message saying when you’ll come back, because a follow-up somebody is expecting arrives as a kept appointment. After that, four habits carry it. Ask for a date instead of the thing. Put everything you need from one person into one message. Confirm out loud when something arrives. And close the loop yourself when the answer is clearly staying away. Persistence is fine on its own. What makes somebody a nag is persistence with no end, no news and no acknowledgment, and each of those three is yours to fix.
Which of these are worth chasing at all is a question about the work rather than about email. A signature you’re missing in an accounting practice behaves differently from money you’re missing, which gets its own treatment in following up on an unpaid invoice, and eight trades have their versions set out.