Skip to content
Robnu
Field NotesField Notes4 min read

SLA watchdog: how we catch a missed manifest before the marketplace does

The SLA watchdog is one of the highest-impact features Robnu ships, and one of the simplest. This post covers the headroom formula, the escalation thresholds, why we picked 30% and 10%, and the underrated part: it reframes the wait-one-more-shipment decision so you make it deliberately.

Hiren Bhuva
Co-founder, Onviqa Inc. · Robnu
TL;DR
  • SLA-headroom is the percentage of the closure window remaining. The watchdog's job is to make this number visible at 30% and 10%.
  • Defaults are conservative — most sellers find them noisy at first, then realise they were saving money.
  • The point isn't to nag you. It's to prevent the accidental late closure.

The SLA watchdog is one of the highest-impact features Robnu ships. It’s also the simplest. Here’s the mental model.

SLA-headroom

For every open batch, the watchdog computes:

headroom = (closure_window_end − now) / (closure_window_end − closure_window_start)

Headroom is a percentage. 100% = full window left. 0% = window closed (breach). Negative = past the deadline.

You see this number on every batch in the Robnu dashboard. The watchdog escalates when it crosses thresholds.

Diagram of the SLA headroom formula. Headroom equals closure window end minus now, divided by closure window end minus closure window start, as a percentage. 100 percent means the full window is left, 0 percent means the window has closed, and negative means past the deadline. The bar runs from a safe zone through a warning zone to a breach.
Figure 1 — SLA headroom is the percentage of the closure window still remaining (illustrative).

The thresholds

Defaults:

  • 30% headroom — heads-up. Mild ping in the dashboard, no email. “This batch is in the back half of its window. If you have more shipments to add, do it now.”
  • 10% headroom — act now. Email + dashboard banner. “This batch will breach if you don’t close in the next few minutes.”
  • 0% headroom — breached. Slack notification + dashboard error. “Closure now will land outside the window. Either close immediately and accept the deduction, or escalate to support.”

The thresholds are tunable in settings. Most sellers leave them at defaults.

Diagram of the watchdog escalation thresholds. At 30 percent headroom a mild heads-up pings the dashboard with no email. At 10 percent headroom an act-now email and banner fire, giving roughly 30 minutes. At 0 percent headroom a breached notification and dashboard error fire. Thresholds are tunable but most sellers keep the defaults.
Figure 2 — Three escalating thresholds: heads-up at 30%, act-now at 10%, breached at 0% (illustrative).

Why 30% / 10% specifically

We picked these numbers from the data of early users. The 30% threshold catches the “I forgot the time” pattern — most sellers who get a 30% ping close within an hour, well before the actual deadline.

The 10% threshold catches the “I’m in the slip queue and I lost track” pattern — sellers in the middle of operations who didn’t realise the clock was running. 10% gives roughly 30 minutes to either close cleanly or make a deliberate decision to delay.

Below 10%, you’re in scramble territory. The watchdog still helps — it tells you to stop trying to add shipments and just close — but you’re now in the cost-of-rushed-decisions zone.

What the watchdog does not do

  • It doesn’t auto-close. Even at 0% headroom, you decide what happens.
  • It doesn’t override your closure policy. If you’re on delayed mode and the delay time hasn’t hit, the watchdog escalates but doesn’t preempt.
  • It doesn’t pretend a breach is recoverable. If the deadline passed, the deduction is coming. The watchdog tells you that bluntly.

What it does do that’s underrated

The watchdog’s biggest lift isn’t catching the obvious miss. It’s normalising the act of closing earlier than your gut says you should.

Most sellers’ gut tells them to close at the very end of the window. “I might get more shipments ready; let me wait.” The watchdog reframes the question: “you have 18% headroom; is the next shipment 18% likely to be ready in time?”

Diagram of how the watchdog reframes the wait-one-more-shipment decision. The gut says close at the very end of the window in case one more shipment is ready. The watchdog reframes it as a probability question: with 18 percent headroom, is the next shipment 18 percent likely to be ready in time? It surfaces the trade-off so the seller decides deliberately.
Figure 3 — The underrated lift: the watchdog reframes the wait-one-more-shipment decision (illustrative).

Usually the answer is no. Most batches that “might get one more shipment” don’t, and the cost of waiting (rushed closure, error rate up, midnight stress) exceeds the cost of waiting one more shipment for tomorrow’s batch.

Read more

Tags:slaprocessengineering

Frequently asked questions

  • For every open batch, the watchdog computes headroom as the closure window end minus now, divided by the closure window end minus the closure window start. It is a percentage: 100 percent means the full window is left, 0 percent means the window has closed (a breach), and a negative value means you are past the deadline. You see this number on every batch in the Robnu dashboard.

Start Robnu free

See where you're losing rupees on Ajio

Robnu walks every Ajio order from open through manifest, flags every silent deduction, and watches every SLA. Free during early access. No caps. No card. No trial timer.

  • Ajio order processing — every stage covered
  • Free for ≤ 25 orders/day — forever
  • 11-stage flow, document pipeline, SLA watchdog

Sources & further reading

  1. Meesho supplier panel — dispatch SLA terms
    Meesho Supplier HubAccessed July 2026
Hiren Bhuva
Co-founder, Onviqa Inc. · Robnu

Hiren has spent over a decade shipping commerce software for Indian sellers and runs Onviqa Inc., the parent company behind Robnu. He writes about marketplace ops, deduction defense, and the boring infrastructure that decides whether a small Indian brand keeps its money.

Related reading

All posts
build e7713058ee9ee67dffe938623a3f859dcb157b2a · 2026-07-24T12:14:00+05:30