Skip to main content
← All articles
Tooling · November 2026

In-House NDR Handling or Hand It to a 3PL?

11 minutes read
Handled in-houseHandled by your 3PLNDR code lands in the panelYour team reads the real reasonYou speak to the buyer yourselfYou raise the re-attemptYou keep the outcome dataNDR code lands in their panelTheir rules pick an actionA template speaks to the buyerRe-attempt raised for youYou get a summary, not the dataFive steps either way.The question is who owns step three — and who keeps step five.
Summary

NDR handling breaks into five steps and both routes run all five. What changes is who holds the customer conversation, how fast the first contact happens, and whether the outcome data stays yours. Those three things decide it — not price per NDR.

Every brand running COD at volume reaches the same fork. Failed delivery attempts are arriving faster than anyone can work through them, and there are two obvious answers: put people on it, or let the courier or aggregator handle it with their own panel and their own automation. The decision usually gets made on cost per NDR, which is the least informative number available. The useful question is narrower — which of the five steps in an NDR do you actually need to own.

The five steps, whoever runs them

An NDR is not one task. It is a short chain, and each link can sit in a different place.

One: the code arrives. The courier’s field system reports a failed attempt with a reason code — customer not available, premises locked, refused, address incomplete, phone unreachable, out of delivery area. This step is identical in both models. It is data arriving.

Two: the code gets interpreted. This is where the two models genuinely diverge, and it is the step most merchants do not realise is a judgement call. NDR reason codes are entered by field staff under time pressure, and the code is a category, not an explanation. “Customer not available” covers a buyer who was at work, a buyer who was never called, a buyer who has changed their mind and is avoiding the call, and an attempt that did not physically happen. Deciding which of those you are looking at requires context — the order, the product, the buyer’s history, the lane, that courier’s habits in that pincode.

Three: somebody contacts the buyer. A call, a WhatsApp message, an SMS, or several in sequence. The aim is a confirmed delivery slot and a confirmed intention to pay.

Four: the re-attempt is instructed. The outcome of step three is pushed back to the courier as an instruction — re-attempt tomorrow, re-attempt with this alternate number, deliver to this corrected address, hold at the hub for pickup, or RTO now and stop spending money on it.

Five: the outcome is recorded. What was tried, what the buyer said, whether the re-attempt succeeded. In aggregate this is the only asset the whole exercise produces: it is what tells you which courier reports codes honestly, which lanes fail structurally, and which buyers should never see COD again.

Both models do all five. Run the comparison step by step and the trade-off stops being abstract.

What in-house actually buys you

The honest case for in-house is not that your team is better at making phone calls. It is that steps two, three and five are where the value sits, and in-house is the only arrangement where all three are yours.

Interpretation improves with context only your team has. Your agent knows that this SKU generates second thoughts, that this buyer has refused twice before, that this particular courier’s codes in this particular city tend to be optimistic. None of that is available to someone working a queue across hundreds of merchants. The result is that in-house handling can distinguish a logistics failure from a buyer who has changed their mind — and those two situations need opposite responses. One needs a re-attempt. The other needs a conversation about what went wrong, or a clean cancellation before you spend more on freight.

The conversation can go off script. A buyer who hesitates because they are unsure about size can be reassured. A buyer who wants a different delivery window can be accommodated. A buyer who wants to switch to prepaid can be sent a link on the spot, which turns a probable RTO into a safe delivery. Templated flows cannot do any of this, because the decision tree that covers it is not a decision tree.

The data stays yours. This is the most underrated argument. If your team records what was actually wrong, by courier and by lane, within a few months you hold something your couriers will not give you: an independent read on their reporting quality. That read is what gives you leverage in a rate conversation and what makes your allocation rules better than guesswork. Hand step five to a vendor and you get a dashboard of their performance, not a dataset about your business.

And it compounds into other decisions. NDR outcome data feeds your risk scoring, your repeat-offender list, your pincode policy and your prepaid-incentive targeting. It is the same dataset, used four times.

What in-house actually costs

The cost is not the salary. It is the shape of the work.

NDR volume is spiky and the spikes are the expensive part. It clusters around festive peaks, courier network disruption, weather events and your own sale periods — precisely when your team is already stretched. Staffing for the average leaves you drowning at the peak, and staffing for the peak means paying for idle capacity for most of the year.

There is a hard daily deadline. Couriers accept re-attempt instructions within a cut-off window. Miss it and the shipment moves another step towards RTO regardless of how good your conversation was. This makes NDR handling genuinely time-critical in a way most support work is not — an NDR worked well but worked late is worth very little.

It needs management, not just headcount. Scripts, escalation rules, a disposition taxonomy everyone uses the same way, QA on what agents are recording. Without that discipline you get step three without step five, which is the worst of both worlds: you pay for the labour and still have no data.

And it needs panel access per courier. If you ship through three couriers directly, your team is working three interfaces with three cut-offs and three code vocabularies. This is the operational drag that pushes brands towards an aggregator in the first place.

What the 3PL route actually buys you

Aggregators and couriers that offer NDR management are solving a real problem, and the case for them is stronger than in-house advocates usually admit.

Coverage, not quality. Automated flows contact every NDR, including the ones that arrive at 9pm on a Sunday and the three hundred that arrive on the first day of a sale. A mediocre contact on every NDR frequently beats an excellent contact on sixty per cent of them, because the shipments nobody reached are pure loss.

The integration is already built. The re-attempt instruction goes back to the courier through a connection that exists, inside the cut-off, without a human remembering to do it. For multi-courier operations this alone can be worth the fee.

It absorbs the spikes. Volume elasticity is the thing you cannot buy with headcount. Festive season is the period when an outsourced flow most clearly outperforms a team that is also packing orders.

It is faster to first contact. Automation fires on the NDR event. A human queue fires when someone gets to it. Since the usable window before the next attempt is short, time-to-first-contact is one of the few levers that reliably moves re-attempt success, and automation wins it outright.

Where the handover quietly costs you

Step two gets collapsed into a rule. A rules engine maps a reason code to an action. It cannot tell a logistics failure from a change of mind, so it treats both as a re-attempt. You pay freight on re-attempts that were never going to land, and the buyer who wanted to cancel gets two more messages telling them a parcel is coming.

The conversation cannot convert. A templated message confirms availability. It does not rescue a hesitant buyer, does not switch anyone to prepaid, and does not surface the product issue behind a refusal. Every one of those is a recoverable order lost to the limits of the format.

You lose your independent read on courier quality. When the party reporting the NDR and the party resolving it are in the same commercial relationship, you have no outside check on code accuracy. This is not an accusation of bad faith — it is simply that you have given up the only measurement that would tell you either way.

The brand voice is not yours. A delivery call from an unknown number, asking a customer to confirm a payment, is a trust-sensitive moment. Generic flows get confused with fraud calls, and a buyer who thinks they are being scammed refuses the parcel.

Incentives do not fully align. An NDR counted as actioned is not the same as a delivery recovered. Any arrangement priced on activity rather than outcome will drift towards activity. Worth reading whatever is being reported to you with that in mind.

The split most brands end up with

The sensible configuration is not one or the other. It is to send the automated flow at everything immediately, and put people only where people change the outcome.

Automated first touch on every NDR, within minutes. No triage, no queue. Speed is the point, and an automated message costs almost nothing. Plenty of NDRs are resolved right here by a buyer who simply confirms a time.

Human second touch on a narrow, deliberately chosen slice. Order value above a threshold, where freight on a wasted re-attempt genuinely hurts. Repeat customers, where the relationship is worth the call. Codes that imply refusal rather than absence, because those are the ones where a rules engine does the wrong thing. First-time buyers on high-consideration SKUs. And anything on its second NDR, where the automated path has already demonstrated that it cannot resolve this one.

Your own disposition taxonomy on top of both. Whatever handles the contact, insist that outcomes come back in your categories, not the vendor’s. This is the point of negotiation most merchants skip, and it is the one that preserves step five. If the arrangement cannot give you per-shipment outcomes in a form you can join to your order data, you are buying coverage and giving up measurement — a trade worth making knowingly rather than by accident.

What to measure before you decide

Both routes get evaluated on “NDR resolution rate”, which is close to meaningless because every party defines it differently. Baseline these instead, on your own data, before you change anything.

Time from NDR raised to first buyer contact. The single most diagnostic number in this whole decision. If yours is measured in hours and there is a daily cut-off, automation will beat your team on arithmetic alone.

Re-attempt success rate by reason code. Not blended. If “customer not available” converts well and “refused” almost never does, you have just found the slice where re-attempting is burning money and a human conversation is worth having.

Contact coverage. Of NDRs raised, what share got any contact at all before the cut-off. In-house teams are usually worse here than they believe.

Freight spent on failed re-attempts. Count the re-attempts that failed and price them. This is the cost of treating every code as a re-attempt, and it is the number that tells you what better interpretation is worth.

NDR rate by courier, for the same lanes. If two couriers serve the same pincodes and one reports NDRs at a consistently different rate, that difference is information — about the network or about the reporting. Either way you want to know.

Run these for a few weeks and the decision usually makes itself. Brands with poor coverage and slow first contact should automate before they do anything else. Brands with good coverage, a high average order value and a real courier-quality problem should keep people on the slice that matters and hold onto their data.

When each one is the wrong choice

In-house is the wrong choice if your NDR volume is spiky and your team also packs orders, if you ship through several couriers and nobody owns the panels, if your time-to-first-contact is longer than the re-attempt window, or if you have no disposition taxonomy and no intention of enforcing one. Doing it in-house badly is more expensive than outsourcing it adequately, and it produces no data either.

Outsourcing is the wrong choice if your average order value is high enough that a wasted re-attempt costs more than a phone call, if your refusal-type codes dominate your mix, if you are in a category where buyers need reassurance rather than confirmation, or if you are actively trying to build courier-quality data to renegotiate rates. It is also wrong — in the sense of premature — if your NDR rate is being driven by something upstream. A high NDR rate caused by bad addresses or slow dispatch is not an NDR problem, and neither route will fix it; start with address quality and dispatch speed instead.

Where COD King fits, and where it does not

COD King works upstream of all of this. It verifies the order before it is created — phone verification, pincode and risk rules, partial advance collection — so that fewer shipments reach the NDR stage at all. The cheapest NDR is the one that never happens, and an unverified order is the most reliable way to generate one.

It is not an NDR tool, and it would be the wrong thing to buy if NDR handling is your actual bottleneck. If parcels are shipping to real buyers who then cannot be reached by the courier, that is a post-dispatch coordination problem and it needs either a team or a 3PL flow — a verification gate at checkout will not touch it. The two are complementary rather than alternatives: verification reduces the volume, NDR handling resolves what is left. If you are weighing them against each other, look at whether your NDRs are concentrated in orders that were suspect at checkout or spread evenly across good ones. That tells you which end of the pipe to work on first. The NDR management guide covers the operational detail of the downstream side.

This article describes mechanisms rather than benchmarks, because published NDR and re-attempt figures vary enormously by category, lane and courier mix. Measure your own before acting on anyone’s averages.