When customers apply for a visa on Atlys, they hand us something irreplaceable: their passport. We take that responsibility seriously. But for a long time, we had almost no control over what happened to it once it left their hands.
That is because picking up a passport is not like picking up anything else. It is not an ecommerce return. It cannot be rescheduled casually. It has a travel date attached to it, which means it has a deadline that is non-negotiable. The customer has already booked flights. They have taken leave. The embassy window is fixed. If the passport is picked up late, or lost in transit, or sitting at a hub nobody is tracking, the downstream consequences are real: a missed visa appointment, a cancelled trip, a customer who trusted you completely and has no recourse.
That asymmetry, between how logistics companies think about reliability and what a passport actually requires, is the whole problem. This is a postmortem on how we solved it.
India's logistics problem is worse than you think
Before we talk about what we built, it is worth understanding the environment we were building in.
India's logistics sector is valued at over $300 billion and growing.¹ Those are big numbers. What those numbers hide is how dramatically under-penetrated organized logistics infrastructure still is. Outside the major metros, the network thins fast: local fleet operators, brokers, intermediaries, and informal networks that are unreachable by API, untrackable in real time, and completely non-standardized across geographies.
For years, logistics cost as a percentage of GDP in India was cited at 13–14%. A 2025 government study revised that to roughly 8%,² closer to the US at 8.8% and Germany at 8.7%.³ But national averages flatten a reality that is anything but flat. The efficiency gains are concentrated in top metros and among large organized shippers. Move outside those corridors and the old infrastructure shows: your shipment was delayed because the local hub did not scan it, the delivery driver called from an unregistered number, and nobody knows where the package is because it has not been touched since it left the origin hub three days ago.
India has over 19,000 PIN codes.⁴ Organized couriers cover most of them on paper. In practice, coverage means different things across geographies. In Bengaluru, a top-tier courier delivers next day. In a tier 3 city in UP, the same provider might mean "someone will come eventually." The network is real, but its reliability degrades fast as you move away from metro centers.
In the US, the top five logistics carriers cover over 95% of ZIP codes with next-day or two-day SLAs and real-time scan events at every touchpoint. The infrastructure was built for reliability and accountability at scale. India is getting there, but the gap is still significant, and it shows up most painfully when the document you are moving cannot tolerate variance.
For an ecommerce return, delays are annoying. For a passport, they are unacceptable.
The logistics terms you need to know
When logistics companies talk about moving packages, they think in two flows: forward logistics and reverse logistics.
Forward logistics is what the entire industry is built around: your warehouse ships a package, it moves through hubs, and it lands at the customer's door. Every part of India's courier infrastructure, hub placement, routing, fleet allocation, SLA commitments, account manager attention, is oriented around this direction. It is the core product.
Reverse logistics is the opposite: picking something up from a customer and bringing it back. Think ecommerce returns. In theory, it should be symmetric. In practice, it is treated as a secondary by every major logistics provider in India. Fleet allocation is thinner. ETAs are softer. The hub network was never designed around inbound flows from customers, and the moment volume or reliability matters, that asymmetry becomes painful.
At Atlys, we are fundamentally a reverse logistics company. Our core operation is: go to the customer, collect their passport, route it to the embassy or VFS center, get it stamped, and deliver it back. The final delivery back to the customer is the easier half. The pickup is where almost all of our failures lived.
How we started: the obvious answer and why it failed
Like every startup doing physical logistics in India, we started with the incumbents: well-known courier companies with trusted brands, real infrastructure, and broad geographic coverage. The obvious choice.
The problems surfaced within weeks.
The AWB cancellation problem. AWB stands for Air Waybill: the unique tracking number assigned to each shipment when a pickup is scheduled. In reverse logistics, AWB cancellations are far more common than you expect. A pickup agent would show up and decide the package was not ready (even when it was), or would simply not show up, and the AWB would be quietly cancelled in the system. The customer would wait. We would find out hours later when our ops team chased manually. By then, embassy appointment windows had sometimes already shifted.
The node office problem. India's major logistics providers often operate on franchise-style models. The national brand looks centralized. But the entity that actually picks up your package is a local franchisee running a node office, often off spreadsheets, WhatsApp, and local knowledge. There is no central system. Status updates do not flow in real time from the node office to the national dashboard. When something went wrong, our ops team was calling individual node offices, explaining the situation from scratch each time, and hoping someone there had both the knowledge and the authority to fix it.
At our scale at the time, we were not large enough to get a dedicated central account manager at any of these providers. That meant no SLA enforcement, no escalation path, no ability to push for systemic fixes. Every failure was a one-off, handled manually, by someone on our side tracking down someone on their side through a phone call.
The fundamental mismatch. None of this was built for time-critical documents. A passport is not an Amazon return. When a customer's visa is approved, there is a travel date attached. That date is fixed. The flights are booked, the hotel is paid, the leave is approved. If the passport arrives a day late because a node office sat on it, it is not an inconvenience. It is a missed flight. The ETA on a passport pickup is not a guideline. It is a hard commitment.
The SLAs offered by traditional couriers are expressed as "best effort" or "3-5 business days." In most industries, that is acceptable. For a document with a travel date attached, it is not.
Looking at the data
After about a year of fighting this problem manually, we pulled our pickup data and looked at it properly.
55% of our passport pickups were concentrated in tier 1 cities: Mumbai, Delhi, Bengaluru, Hyderabad, Chennai, and Pune. These are cities where Atlys has AFCs. An AFC (Atlys Fulfillment Center) is a physical Atlys-staffed location where customers can walk in, get help with their application, and submit documents directly. In tier 1 cities, we had some operational presence and muscle.
The other 45% was distributed across tier 2 and tier 3 cities: Jaipur, Ranchi, Surat, Lucknow, Coimbatore, Bhubaneswar, Nagpur, Vadodara, and hundreds of smaller cities beyond.
These two clusters needed completely different solutions. We had to solve them separately.
Solving tier 1: the hyper-local experiment
For the 55% in tier 1 cities, we explored two paths before finding what actually worked.
Path 1: Build our own logistics network. We briefly explored owning the problem entirely. Hire drivers, buy or lease vehicles, run the operation ourselves. This sounds straightforward until you start thinking through shift planning, vehicle maintenance, driver background verification, insurance, GPS tracking infrastructure, weekend coverage, and what happens when a driver calls in sick on the exact day a customer has an embassy appointment. You are now building a logistics company inside a visa company. The unit economics collapses immediately and the operational complexity would have consumed the core product team.
Path 2: Hyper-local logistics providers. This is where things got genuinely interesting.
Hyper-local logistics is a specific category of on-demand delivery built for intracity movement. Unlike hub-and-spoke couriers that route packages across India through regional hubs, hyper-local providers operate within a single city using real-time driver dispatch. The model is closer to Uber than to a courier company: a driver is assigned dynamically, picks up from point A, delivers to point B, and the median time the package is in their hands is under an hour.
This changes the problem in important ways.
With a traditional courier, your passport enters a network. It passes through a hub. It may sit in a staging area overnight. Scans happen when someone remembers to scan. You find out where it is when someone updates the system. Median time in transit: 24 to 72 hours, with variance that is hard to predict.
With a hyper-local provider, the driver is assigned in real time. You can see them on a map. Pickup-to-delivery in tier 1 cities runs 45 to 90 minutes. Tracking is complete, live, and GPS-pinned. Rescheduling is frictionless: a few taps in an app instead of a call to a node office. And because the driver is in transit for under an hour, there is essentially no dwell time where the document is sitting untracked somewhere.
From a pure operations standpoint, this was clearly the better product. But we hit a wall we did not see coming.
Customers did not trust it.
We could show them live GPS tracking. We could show them the driver was 2 km away. We could explain that the passport was insured. We wrote it in multiple ways across the app and SMS. None of it moved the needle. The mental model of handing your passport over to someone who arrives on a motorcycle, with a soft delivery bag, and no visible institutional affiliation, felt wrong. It felt unsafe. The very properties that made hyper-local fast, lean dispatch, no physical infrastructure, app-driven coordination, were the exact properties that made customers anxious.
Trust-based escalations on passport pickup remained stubbornly high. Customers were calling our support team mid-transit asking if we were sure this was the right approach. Some refused to hand over the passport when the driver arrived.
The safety box
The insight that broke the deadlock was this: the problem was not security. It was the perception of security.
A passport is treated like jewellery. When people hand it to someone, they look for signals that it will be handled with care: a secure container, a receipt, a logo, something that communicates "this is official and can be tracked." A delivery driver with a nylon courier bag triggers no such signal regardless of what you tell them in advance.
So we built a metal safety box.
The box is sturdy, branded, and lockable with a unique numeric code. Before the driver arrives, the customer receives their specific code. They place the passport inside, lock the box, and hand the locked box to the driver. The driver cannot open it. At the destination, our AFC staff open it using a corresponding code on their end. Inside each box we placed a GPS tracker, so the customer can pull up the Find My app and see the physical location of the box in real time.
It sounds like theater, and in a narrow sense it is: the passport is not meaningfully more physically secure inside a locked metal box than it would be in a well-managed delivery bag. But the experience of handing a locked, weighted, branded metal box to a driver is categorically different from handing an envelope. The physical weight communicates value. The lock communicates control. The GPS tracker communicates transparency. The combination signals: this has been designed by people who take this seriously.
Trust-based escalations on tier 1 pickups dropped to effectively zero after we rolled this out.
The 45% problem: tier 2 and tier 3
Tier 1 was solved. But 45% of our pickups were coming from everywhere else, and hyper-local providers simply do not operate in most of these cities. The model requires enough driver supply and enough demand density to make real-time dispatch work. In a city of 800,000 people doing 15 Atlys pickups a month, that market does not exist.
What this meant was going back to traditional couriers for this cohort, with all the problems we had already documented.
We tried six different logistics providers over roughly eighteen months. The pattern was consistent: initial weeks would work fine, then slippages would start. A passport would sit at a node office for two days waiting for a connecting vehicle. A driver would arrive on day three of a two-day SLA. We would find out when the customer called support, not from any system alert, because there were no system alerts.
Opening our own offices in these cities was never viable. The unit economics do not make sense. Tier 2 volumes per city, even as we grew, were not large enough to justify the fixed cost of rent, staff, and local operations management. And running 40 small Atlys offices across India while running the core product would have required a management layer we were not built for.
The solution was sub-AFCs.
We built a network of trusted and verified local partners in tier 2 and tier 3 cities. These are established local businesses with a physical presence: travel agencies, financial services offices, businesses that customers in those cities already know and have some baseline trust in. We call them sub-AFCs. They are not Atlys employees, but they operate under our SOPs use our systems, and handle passports under our SLAs with accountability enforced through our partner agreements.
The pickup flow in a tier 2 or tier 3 city now works like this:
A hyper-local driver goes to the customer's home, collects the safety box, and delivers it to the nearest sub-AFC in that city. The transit time is still under an hour.
The sub-AFC logs receipt, stores the passport under our protocol, and provides the customer with a confirmation.
We run scheduled, vetted consolidated pickups from sub-AFC locations to our primary processing centers on a fixed cadence.
The document spends minimum time in transit. The customer gets real tracking at the last-mile stage. The sub-AFC provides a physical accountability checkpoint in a city where we otherwise had no presence. And the scheduled consolidation from sub-AFC to processing center runs on a predictable, manageable cadence that we can actually enforce.
Where we are today
We now cover 95% of our passport pickups within our committed turnaround time. The remaining 5% are cities with pickup volume too low to support any sub-AFC partnership, where we still rely on traditional logistics providers with the manual oversight that requires.
The full arc of this problem, from "use BlueDart" to where we are today - took the better part of three years. That timeline is not because the team was slow. It is because the failures were systematic, the environment was hostile to clean solutions, and each approach revealed a new constraint that the previous approach had hidden.
What we learned:
Reverse logistics is not a feature of India's courier infrastructure. It is an edge case. If your core operation depends on it, you cannot fully outsource the accountability.
Perception of security is a product problem, not a communication problem. You cannot explain your way to customer trust when the physical experience contradicts it. The solution was a physical product.
Hyper-local logistics is dramatically underestimated for high-stakes documents. The properties that make it feel "too casual" - speed, leanness, app-driven dispatch - are exactly what you want when a document cannot be late and cannot be lost.
Building your own logistics network is a company inside a company. It is tempting when you are frustrated with third-party providers. It is almost never the right answer unless logistics is your product.
Partner networks beat national couriers in tier 2 and tier 3 cities for reliability, because accountability is personal and local rather than systemic and diffuse.
The unsexy version of this story is that we spent years calling node offices, chasing AWBs manually, and tracking packages on behalf of customers who had no visibility into what was happening. The cleaner version is that we built a pickup infrastructure that is purpose-designed for a document that cannot be late, cannot be lost, and belongs to someone who is trusting you completely.
Both versions are true. The clean version only exists because we did the unsexy one first.
Sources
India logistics market valued at $244–$384 billion in 2025, depending on scope and definition. (IMARC Group, 2025; Mordor Intelligence, 2025)
DPIIT and NCAER, Assessment of Logistics Cost in India, September 2025. India's logistics cost assessed at 7.97% of GDP for FY 2023–24. The previously cited 13–14% was based on Armstrong & Associates (2016) and other external estimates using different methodologies.
CSCMP and Kearney, 36th Annual State of Logistics Report, June 2025: US logistics cost at 8.8% of GDP for 2024. Germany at approximately 8.7% per Armstrong & Associates (2025).
India Post. Approximately 19,100 active PIN codes across 28 states and 8 union territories.