A homeware store ships around 900 parcels a month. On a Tuesday morning the owner opens a notification from a public complaint site: a customer, order placed nine days ago, says the parcel never arrived and nobody has replied to two emails.
She checks. The tracking shows a pickup scan on day one and nothing since. The two emails are in the shared inbox, unread, below forty newer messages. The parcel has been silently stuck for eight days, and in that time the store has done nothing at all — not because anyone decided to do nothing, but because nothing in the system was watching.
So she does what most people do. She apologizes, refunds the order, and replies to the public complaint. The parcel problem is now closed.
The customer problem is not, and it's worth being honest about what actually happened: she didn't lose that customer when the parcel got stuck. Parcels get stuck at every store, at every carrier, in every country. She lost that customer over the eight days in which nobody said anything, and then over the discovery that the only way to get a response was to complain somewhere public.
That gap — between resolving the parcel and keeping the customer — is what this guide is about. It's a playbook for handling delivery complaints: what each kind of failure actually needs, who should speak first, what a remedy should cost, what to do when the complaint goes public, and the handful of numbers that tell you whether any of it is working.
This guide sits next to two others and deliberately doesn't repeat them. Reducing WISMO tickets is about the contacts you can deflect with better information — the ones where nothing is actually wrong. Lost and damaged parcels is about recovering the money from the carrier. This guide is about the third thing: the conversation with the customer when something genuinely has gone wrong, which no notification can deflect and no claim can repair.
Two Outcomes, One of Which Nobody Manages
Every delivery failure produces two separate outcomes, on two different clocks.
The first is the parcel outcome. The parcel arrives late, or it's found, or it's replaced, or it's refunded. This outcome is concrete, it has an owner, it eventually closes, and most stores manage it reasonably well because it can't be ignored — an unresolved parcel keeps generating work until someone deals with it.
The second is the customer outcome. At some point during the failure, the customer forms a view about whether ordering from you again is worth the risk. This outcome is invisible, has no owner, never closes, and generates no work at all. Nothing in your inbox tells you it happened.
Here's the uncomfortable part: the second outcome is worth far more than the first, and it is decided almost entirely by things the first outcome doesn't measure. Two customers can have identical parcel outcomes — both got a full refund on day nine — and completely opposite customer outcomes, depending on whether they were told on day three or had to find out themselves on day eight.
Most stores have a delivery-problem process that's really a parcel process. Refund or replace, file the claim, close the ticket. Ask the same store how many customers who had a delivery failure last quarter ordered again, and the answer is usually a shrug. The data exists — order history joined to delivery outcome — but nobody has ever run it, because nothing forces you to.
Fix it: Treat "the parcel is resolved" and "the customer is retained" as two different jobs with two different definitions of done. The first ends when the money or the goods are settled. The second ends when that customer places another order — and if you never look, you'll never know how often that happens.
Who Speaks First Decides Almost Everything
Of all the variables in a delivery failure, one predicts the customer outcome better than the rest, and it happens to be the one entirely under your control: who started the conversation.
A customer told on Wednesday that Thursday's parcel is running late is dealing with a store that's paying attention. The same customer, discovering on Saturday that the parcel never came and writing in to ask, is dealing with a store that wasn't. The parcel is identical. The relationship is not.
The research supports this in two directions. A Gartner survey of more than 6,000 customers found only 13% could recall ever receiving any kind of proactive customer service — meaning the bar here is genuinely low, and clearing it is unusual enough to be noticed. On the other side, a Genesys study puts the reduction in complaints from proactive communication at around 40%.
Our own corpus numbers point the same way: proactive shipment notifications reliably cut WISMO contacts by 40% to 80%, and a delivery exception detected by you rather than reported by the customer converts your worst ticket into your best message.
There's a blunter way to put it. A complaint is the invoice for your silence. You are going to pay for the failure either way — in support time, in remedies, in reputation. Speaking first is simply the cheapest payment method available, and its price goes up every day you wait.
The practical version of this is a detection rule, not a communication rule. You cannot speak first about something you don't know about. That means:
- Every shipment gets a promised date recorded at label time, not just a carrier tracking number. Without an expected date, "late" is undefined and nothing can fire.
- Any shipment with no new scan for a defined window — 24 to 48 hours past the expected delivery date is the common threshold — gets flagged the same day.
- Any exception event the carrier reports (address problem, failed attempt, return to sender) gets flagged immediately, because these never resolve themselves.
Fix it: Measure the share of your delivery problems where you contacted the customer first. Most stores who check this for the first time find it's under 20%. It is the single most improvable number in this entire guide.
The Trap: Proactive Done Badly Costs More Than Silence
This is where most advice on the subject stops, and where it becomes actively misleading. "Be proactive" is not the whole rule, because a proactive message that raises a question it doesn't answer doesn't deflect a contact — it creates one.
Gartner's research on this is unusually specific and worth taking seriously: around two-thirds of customers contact customer service after receiving proactive outreach, and those contacts arrive through the expensive human channels. The lift that proactive service gives to how customers rate a brand erodes when the outreach leaves obvious questions hanging. Roughly 10% of consumers get in touch purely to check the message isn't a scam — a number that rises to around a quarter in a business-to-business context.
Read that as an operating constraint rather than a reason to stay quiet. It means a delay message has a pass/fail test, and "we're sorry, your order is delayed" fails it. That message tells the customer something they can already see on the tracking page, gives them nothing to do, and reliably produces a reply asking the only question that mattered.
Four things belong in every proactive message about a problem:
- What happened, in plain language, without carrier jargon. "Your parcel is still at the Ankara transfer center" beats "an exception has been recorded on your shipment."
- The new date — or, if you genuinely don't have one, a decision deadline: "if it hasn't moved by Thursday, we'll send a replacement without you needing to ask."
- What you're doing. Not "we've escalated it," which means nothing. "We've opened an investigation with the carrier this morning" means something, provided it's true.
- What they need to do — and if the answer is nothing, say nothing explicitly. "You don't need to do anything" is a complete sentence and it prevents a reply.
Add a fifth thing that isn't content but format: provable sender identity. Send from your store's name, on a channel the customer already associates with you, with no shortened links and no request for information you should already have. The scam-check contacts in Gartner's data are a direct cost of getting this wrong, and in a market where WhatsApp reaches 88.6% of individuals and arrives from a verified business profile, the identity problem is largely solvable by channel choice.
Fix it: Before any proactive message goes live, read it as the customer and ask "what would I reply to this?" If you can think of a reply, the message is incomplete. Rewrite it until the honest answer is "nothing."
Four Failures, Four Different Answers
Treating every delivery problem with one template is the second most common mistake after silence. The four situations below look similar in a support queue and need genuinely different responses.
Late but moving: the customer needs a date, not an apology
The parcel is scanning normally, it's simply going to arrive after the date you promised. This is the most common failure and the easiest to handle badly.
The mistake is apologizing without re-dating. An apology that repeats the original promise — "sorry for the delay, it should arrive soon" — leaves the customer in exactly the state that made them anxious in the first place. What they need is a replacement of the broken promise: a new date, stated as confidently as the first one was.
If you can't produce a reliable new date, say so and name the day you'll come back to them. Vagueness is tolerable; open-endedness is not. And take the estimated delivery date problem seriously upstream too, because a promise that was never realistic manufactures these conversations by the hundred.
Stalled with no scan: commit to a decision, not to information
The parcel hasn't moved in days and the carrier has nothing to tell you. This is the hardest case, because the thing the customer wants — information — genuinely doesn't exist.
The move here is to stop trying to supply information and supply a commitment instead. "There's been no movement since Monday and the carrier has opened an investigation. If there's no update by Thursday, I'll ship you a replacement that day — you won't need to chase me." That message resolves the customer's anxiety without resolving the parcel, which is precisely the trick, because the parcel may not resolve for another week.
Then keep the commitment, unprompted, on the day you named. A kept deadline in a bad situation does more for retention than a smooth delivery ever will.
Marked delivered, not received: never make the customer prove innocence
The tracking says delivered. The customer says it isn't. Somebody is wrong, and the temptation is to find out who.
Resist it, at least at the start. The instinctive response — asking the customer to check with neighbors, to look around the building, to confirm they were home — is heard as an accusation, because that's what it is. You can do the checking yourself: request the proof of delivery and the GPS or signature data from the carrier, look at the delivery time against the customer's message, check whether that address has a history of problems.
Set an internal value threshold below which you simply replace or refund without investigation, and treat that threshold as a cost of doing business rather than a defeat. Above it, investigate on your side, keep the customer updated while you do, and give them a deadline. The cases where a customer is genuinely trying it on are real, but they are rarer than the cases where a courier left a parcel with a neighbor and marked it delivered — and the process you design for the rare case is the process every honest customer has to walk through.
Damaged: speed first, evidence second
A damaged item is the one failure where the customer already has the bad outcome in their hands. The clock on their patience started before they wrote to you.
Two things need to happen and their order matters. First, resolve the customer — replacement or refund, decided quickly. Second, collect what you need for the carrier claim: photos of the item and the packaging, the box kept until the claim closes. Ask for the photos in the same message that tells them the replacement is already on its way, so the request reads as process rather than as a condition.
What must never happen is the claim becoming visible to the customer. Your dispute with the carrier can take weeks. Their resolution cannot wait for it, and any sentence beginning "once the carrier approves our claim" turns a recoverable problem into a lost customer. If damage is a recurring pattern rather than an occasional event, the fix isn't a better apology — it's packaging and carrier selection.
Fix it: Write one page with these four scenarios and the standard response to each, and put it where whoever answers messages can see it. Not templates to paste — decisions already made, so the person replying at 6pm on a Friday doesn't have to invent a policy.
Never Make the Carrier Your Answer
There is a sentence that appears in an enormous number of delivery-problem replies, and it costs more than anything else in this guide: "The parcel is with the carrier — you can contact them directly with your tracking number."
Every word of it may be true. It is still the wrong answer, for two reasons.
The first is contractual. The customer bought from you. They chose you, paid you, and have no relationship with the carrier at all — they didn't select it and can't hold it to anything. Handing them the carrier's number is asking them to chase a company they never did business with about a problem they didn't cause.
The second is that they weren't going to blame the carrier anyway. A YouGov survey commissioned by Seven Senders found that around 40% of European shoppers who had a poor delivery experience would buy from somewhere else next time — and they attributed the failure to the shop, not the courier. Published estimates of how many shoppers abandon a retailer after a bad delivery range widely, from roughly 40% to well over 80% depending on the survey and how the question is framed, but the attribution finding is consistent: the delivery is part of your product.
The important nuance is that this applies to what you say, not to what you do. Internally, the carrier is very much the answer. A store that can produce complaints per thousand delivered parcels, split by carrier and by region, has turned a customer-service problem into a procurement decision — and that's the only version of "it's the carrier's fault" that ever changes anything. That belongs in your carrier scorecard and your next rate conversation, not in your reply.
Fix it: Ban the carrier hand-off from your outbound messages entirely. If you need the customer to interact with the carrier — to accept a redelivery, for instance — arrange it and tell them what will happen, rather than delegating it.
The Remedy Ladder: What to Offer, in What Order
Two failure modes dominate here, and they're opposites. Some stores refuse everything and argue; more stores, especially small ones, reach straight for a refund because it ends the conversation fastest. Both are expensive.
Remedies form a ladder, and money is the top rung, not the first:
- Information. A real answer with a real date. Resolves more complaints than anything below it, and costs nothing.
- Control. Let the customer change something: a new delivery date, a different address, redirect to a pickup point, hold for collection. Restoring agency to someone who feels powerless is disproportionately effective.
- Speed. Upgrade the delivery, or ship a replacement immediately rather than waiting for the original to turn up.
- Goods. A replacement, or the missing item, or a small addition to the reshipment.
- Future money. Store credit or a discount on the next order. Cheaper than a refund and it buys another chance — but it is a poor first response to real anger, because it reads as a sales offer at the wrong moment.
- Money back. Refund the shipping fee first — it's the part of the transaction that actually failed, and it's a fraction of the order value. Partial refunds on the goods next. A full refund last, because it also cancels the sale.
Two rules make this usable.
Set the thresholds in advance. Decide now what a person answering messages can authorize without asking: below this order value, refund the shipping without approval; below that value, replace without investigation; above it, escalate. The worst possible time to work out what an apology is allowed to cost is while a customer is angry and someone junior is improvising.
Compare the remedy to acquisition cost, not to order value. A remedy that looks expensive next to a small order usually looks cheap next to what you paid to acquire that customer in the first place. Run your own numbers — average order value, gross margin, customer acquisition cost, repeat rate — because the answer differs enormously between a store selling one high-margin item and a store selling frequent small baskets. What's consistent is the direction of the error: stores that get this wrong nearly always under-remedy relative to what a replacement customer would cost them.
Fix it: Write the authority table on one page: value band, what's authorized, who to ask above it. Review it once a quarter against what you actually paid out.
When the Complaint Goes Public
A complaint that goes public is a recovery that already failed twice — once when the delivery went wrong, and once when your private channel didn't work. It also needs a completely different response, because the audience has changed.
In Turkey this channel is unusually consequential. Şikayetvar's 2025 figures put 2,868,914 complaints on the platform for the year, with 533,117 resolved by the platform's own count — a resolution rate under 20%. E-commerce was the single most-complained-about sector at 365,395 complaints, with cargo and transport sixth at 141,851. Within e-commerce, 63% of complaints concerned cancellation, return and exchange processes and 34% concerned undelivered products. The platform gives brands a defined window — 72 hours — to respond through its solution center, and complaints are published without brand pre-approval.
The seasonal concentration is worth noting too: during the November 2025 campaign period alone, e-commerce complaints hit 28,864 and cargo complaints 12,108, with complaints about brands' own websites up around 20% year on year. December was the platform's heaviest month of the entire year at 267,049 complaints. If you sell into a November peak, your complaint volume peaks with it.
Elsewhere the same dynamic runs through Google reviews, Trustpilot and marketplace seller ratings. The mechanics differ; the rule doesn't:
- Write for the next reader, not the complainant. The person you're really addressing is a stranger deciding whether to buy from you. They cannot verify who was right, so they judge how you behave.
- Never argue the facts in public, even when the facts are on your side. A store that wins the argument loses the reader every time.
- Acknowledge, then move it private, then come back. Reply quickly and specifically, take the details off the public thread, resolve it there, and return to close the thread publicly once it's done. That final step is the one almost everyone skips, and it's the one the next reader sees.
- Treat these pages as part of your brand search results. They rank for your store's name. An unanswered complaint is a permanent liability sitting in the results for anyone who searches you before their first order; a resolved one is the opposite.
Fix it: Assign public complaint channels to a named person with a daily check, and hold to a response time inside the platform's own window. Speed matters more here than eloquence.
Don't Plan on the Service Recovery Paradox
You will find a lot of content arguing that a well-handled failure produces a more loyal customer than a flawless order — the "service recovery paradox." It's a real phenomenon with a real academic literature, and it is much weaker than the way it's usually quoted.
What the research actually says is worth knowing before you build a strategy on it. The meta-analysis by de Matos, Henrique and Rossi (2007) found a significant positive effect on satisfaction but no consistent effect on the things that pay you — repurchase intention, word of mouth, corporate image. Michel and Meuter (2008), working with large real-world datasets, concluded the effect is rare and small enough to have limited managerial relevance. Michel and Coughlan (2009) narrowed the conditions further: it tends to appear only against a mediocre baseline of service, only when the failure is seen as minor, and only when the customer believes the cause was outside the firm's control.
So the honest position is this: recovery is damage control, not a growth lever. Done well it stops you losing a customer you were about to lose. It does not reliably produce a better customer than a delivery that simply worked, and any plan that depends on it — deliberately under-promising to over-recover, or treating failures as loyalty opportunities — is building on a finding the evidence doesn't support.
That's not a reason to invest less. Stopping the loss is worth plenty on its own, and it's a return you can actually measure. It's a reason to keep the priority order straight: fix the delivery first, recover second. A store with a 4% failure rate and excellent recovery is losing to a store with a 1% failure rate and mediocre recovery, every time.
Five Numbers That Tell You Whether Any of This Is Working
Recovery is unusually easy to feel good about and unusually hard to verify, because the successful outcome is invisible — a customer who quietly comes back. Five numbers make it visible. None of them require new tooling beyond what you should already have.
1. Proactive share. Of all delivery problems in the month, the percentage where you contacted the customer before they contacted you. This is the headline number of this entire guide. Baseline it once and you'll almost certainly find it low; it's also the fastest one to move.
2. Time from exception to first contact. Measured in hours from the moment the tracking data went wrong, not from when the ticket was created. This is the number that actually drives the first one.
3. Second-contact rate. How often a customer had to write again after your first message. This is the direct test of the Gartner failure mode above — a high second-contact rate means your messages are triggering questions rather than answering them, and it's usually fixable by editing four sentences.
4. Repeat-purchase rate by delivery outcome. Split last quarter's customers three ways — delivered on time, failed and recovered, failed and not recovered — and compare how many ordered again. This is the number that turns delivery from a cost line into a retention line, and it's the one that will get you budget for everything else in this guide.
5. Complaints per 1,000 delivered parcels, by carrier. Not total complaints, which just tracks your growth. Normalized and split by carrier, and by region if you can, this converts a customer-service metric into a procurement metric and belongs alongside your other shipping KPIs.
Fix it: Pick numbers 1 and 4 to start. The first tells you whether your process changed; the second tells you whether it mattered. The other three are diagnostics for when one of those two won't move.
What Not to Do
The restraint list, because most of the damage in this area comes from over-correcting.
- Don't apologize without a new date. An apology that restates the broken promise is worse than silence — it confirms you know and aren't fixing it.
- Don't forward the carrier's tracking text as your reply. The customer has already read it. Repeating it tells them you looked at the same screen they did and added nothing.
- Don't lead with a refund. It's the most expensive rung on the ladder and it ends the relationship along with the conversation.
- Don't automate the apology. Automate the detection — the flag that says this shipment has stalled. A templated apology sent by a rule, with no human deciding it fits, is how a store ends up apologizing for a parcel that was delivered an hour earlier.
- Don't argue in public. Ever. Move it private and win it there.
- Don't say "we've escalated it" unless something actually happened. Customers have heard it too many times for it to carry information, and it's checkable.
- Don't paper over a carrier problem with discount codes indefinitely. If one carrier or one route generates most of your complaints, the remedy budget is treating a symptom. Fix it through carrier performance data and a second carrier, not through goodwill.
- Don't build your recovery process around the customers who abuse it. Design for the honest majority and price the abuse in. A process built to catch the rare bad actor punishes everyone else on the way through.
Where Shipink Fits
Almost everything above depends on one capability: knowing that a shipment is in trouble before the customer tells you. That's an integration problem before it's a customer-service problem, and it's what Shipink is built to do.
Connect your sales channels and your carriers once, and every order lands in one place with its promised date, its live tracking status and its cost attached. From there:
- Exception and delay detection across every carrier at once, so a stalled parcel surfaces on a screen you're already looking at instead of arriving as a complaint nine days later.
- Automatic status notifications on email, SMS and — on Pro and Enterprise plans — WhatsApp from your own branded business account, which solves the sender-identity problem that generates scam-check contacts.
- A branded tracking page the customer lands on instead of the carrier's, so the answer to "where is it" is on your domain, in your language, with your support channel visible.
- Automation rules that flag or route shipments meeting your own conditions — past the promised date, no scan for 48 hours, exception event received.
- Per-carrier reporting that lets you produce complaints and failures per thousand parcels by carrier, which is the number that makes your next rate conversation a different meeting.
- One order view for whoever answers messages: channel, carrier, promised date, live status and history in one screen, rather than three tabs and a guess.
None of that writes the message for you. It just means the message gets written on day two instead of day nine, which is the entire difference this guide is about.
Why the Time to Write This Down Is Before You Need It
There's no version of this work that goes well under pressure, and there are two reasons the pressure is arriving.
The first is structural. Delivery expectations in Turkey have compressed sharply — Ministry of Trade figures for 2025 put average e-commerce delivery time at 42.2 hours, with 53% of orders delivered within 24 to 48 hours. When two days is normal, five days isn't "a bit late," it's more than double the baseline the customer is measuring you against. The tolerance window shrank; the complaints didn't get more sensitive.
The second is seasonal. Complaint volume concentrates hard in the campaign months — November and December were the heaviest of 2025 by a wide margin — and 11.11 is now under twelve weeks away, with Efsane Cuma on 27 November 2026 behind it. Whatever your recovery process is on 11 November is whatever it was in October, because nobody designs a remedy ladder during a peak. This is the same logic as the T-30 change freeze: the last useful moment to change how you handle failures is about a month before your volume does.
The good news is that this is the cheapest peak-season preparation available. Detection rules, a one-page scenario sheet and an authority table cost nothing but an afternoon, and unlike capacity, they work at any volume.
The Delivery Complaint Checklist
Detection — can you find out first?
- Every shipment has a promised delivery date recorded at label time
- Shipments with no scan for 24–48 hours past the promised date are flagged automatically
- Carrier exception events surface the same day they're received
- Someone owns the flagged list daily, by name
Response — do you know what to say?
- A one-page sheet covers all four scenarios: late, stalled, delivered-not-received, damaged
- Every proactive message carries all four elements — what happened, new date or decision deadline, what you're doing, what they need to do
- Messages send from a channel and sender the customer recognizes, with no shortened links
- No outbound message hands the customer off to the carrier
- Damage and loss claims are invisible to the customer and never gate their resolution
Remedies — is it decided in advance?
- An authority table sets what can be given at each order-value band, without approval
- Remedies are compared against customer acquisition cost, not order value
- Shipping fee is refunded before goods value
- A replace-without-investigation threshold exists for low-value delivered-not-received cases
Public and measurement
- Public complaint channels are checked daily by a named person, inside the platform's response window
- Public threads are closed publicly after being resolved privately
- Proactive share and time-to-first-contact are measured monthly
- Repeat-purchase rate is split by delivery outcome at least quarterly
- Complaints per 1,000 parcels are tracked by carrier and reviewed before every rate negotiation
Stop Letting the Second Outcome Take Care of Itself
Delivery failures aren't going away. At any real volume a percentage of your parcels will be late, stalled, damaged or lost, no matter how good your carriers are — and that percentage is largely out of your hands.
What is in your hands is which of the two outcomes you manage. The parcel outcome resolves itself eventually, because it has to. The customer outcome resolves itself too, and if you're not in the conversation it resolves against you, silently, with no ticket and no notification and nothing in your inbox to tell you it happened.
Speaking first is the whole game, and speaking first requires seeing first.
Try Shipink free and connect your channels and carriers in one place — or talk to us about how your delivery exceptions surface today, and how much earlier they could.
