RTO benchmarks: what refusal rates really look like in Indian D2C

The headline bands: what RTO runs in practice

Working benchmarks across Indian D2C in 2026: COD orders: 15–35 percent RTO is the broad industry band, with unmanaged COD flows at the top and disciplined ones (verification + NDR workflows) at the bottom; festive-quarter COD refusals spike well above baseline. Prepaid orders: 1–4 percent — the gap between the two numbers is the entire argument for prepaid-shift incentives. Blended: a brand at 60 percent COD share and average discipline typically lands 12–20 percent blended RTO; the well-run food and wellness brands we work with hold blended single digits. By category: fashion and footwear run the worst (size doubt + impulse), food and grocery staples the best (need-based, low regret), supplements and personal care in between. If your numbers sit above these bands, the problem is process, not "India".

Region and pincode: where refusals cluster

RTO is violently non-uniform. Patterns that repeat across client data: Tier 2/3 towns run higher refusal than metros (longer delivery promises, higher COD share, weaker address quality — three causes stacked); specific pincode clusters within the same city can differ by 2–3× (address ambiguity, courier last-mile quality, buyer-profile mix); North and East corridors often out-refuse South and West for identical SKUs, driven mostly by COD-share and delivery-day differences rather than anything mystical; and long-promise lanes (5+ days) refuse dramatically more than 2–3-day lanes — impulse cools at roughly a percentage point a day. The operational conclusion: manage RTO at pincode level, not national level — gate, verify or prepaid-only the worst clusters, and route lanes to the courier that performs there (the lane audit is the tool).

RTO is not a logistics metric. It is the truth about how much your buyer trusted you at the doorstep.

What a refusal actually costs: the per-order autopsy

Count all of it: forward shipping (₹40–90 typical), return shipping (similar, often padded), COD handling fee lost, packaging written off, product risk (crushed, expired shelf-life share for food, or full write-off), warehouse handling both ways, payment-cycle capital locked for weeks — and the ad spend that bought the order, wasted in full. Realistic all-in for a food/wellness brand: ₹150–350 per refused order before counting the acquisition spend. At 2,000 monthly COD orders and 25 percent RTO, that is ₹7.5–17.5 lakh a year burning quietly in a line most P&Ls don't break out — while the dashboard ROAS looks fine, exactly the illusion dissected in the real-ROAS piece.

The levers, ranked by measured impact

1. WhatsApp order confirmation on COD (order intent re-verified within minutes): the single biggest cheap win — meaningful refusal reduction for the cost of a utility message.
2. Prepaid-shift incentives (visible discount or gift for paying now, funded by your per-refusal cost): every 10 points of COD share converted to prepaid takes roughly 2–3 points off blended RTO.
3. NDR discipline: call-before-delivery, same-day address-fix workflows, three genuine attempts — contract terms that matter more than rates.
4. Delivery-promise honesty: under-promise by a day; a kept 4-day promise refuses less than a missed 3-day one.
5. Pincode gating: COD caps or prepaid-only on your proven-worst clusters.
6. Address intelligence: checkout-level address validation and completeness prompts — boring, effective.

Running RTO as a weekly metric

Put four numbers on the Monday sheet next to CAC and ROAS: blended RTO percentage (trend, not snapshot), COD share of orders, refusal rate by top-10 pincode clusters, and per-courier RTO on shared lanes. Set the targets by category band — a food brand should be engineering toward single-digit blended; a fashion brand fighting toward the low 20s on COD. And close the loop into marketing: net-of-RTO revenue is the only revenue your value-based bidding should learn from, or the algorithm optimises for buyers who refuse. The brands that treat RTO as a weekly operating metric rather than a quarterly lament are the ones whose "real ROAS" and dashboard ROAS actually converge.