EngineeringSeries · Part 1 of 1

Why we built GrowMagik? One platform for India's neighbourhood merchants

Why a gold purchase savings scheme and a trip fund turned out to be the same thing, and what that meant for how we built it

Walk into a jewellery shop in any small town in Kerala and ask about the monthly scheme. The person at the counter will open a long register, or hand you a small passbook. You save towards jewellery by paying a fixed amount every month, say ₹2,000, or choosing a flexible scheme where you contribute an amount that suits your budget. At the end of the scheme, the shop adds a little extra, and you take home a chain for your niece’s wedding or a pair of bangles before Onam.

These schemes have run for decades on trust and good handwriting. And they work. But sit with the owner for ten minutes and ask a few simple questions. How many customers missed this month's payment? Whose plan finishes next week? Of the people whose plan ended last year, how many came back for another one? Most of the time the answer is, "Give me one day, I'll check and tell you."

That one day is where GrowMagik started.

This is the first of many posts on how we at QuickBee built GrowMagik. Over the next six months we will write about how the system is put together, how we keep every shop's data separate, the server side, the web screens and the mobile apps, how we use AI to write and check our code, and how we keep the whole thing running without a big monthly bill. We are starting with the why, because most of the later decisions go back to it.

What we kept seeing at the counter

The jeweller is only one example. Travel agents run something very similar: families put money aside every month for a trip, often noted down in a book at the agency. Other neighbourhood businesses have their own versions of the same habit.

Most of these shops already have some software. That surprised us at first. A typical store might have:

  • billing software on the counter computer,
  • a spreadsheet for the savings scheme, kept by one person in the back office,
  • the owner's own WhatsApp, with a few broadcast lists for festival offers,
  • a receipt book for the collection agent who goes house to house on his scooter,
  • a points card from some loyalty app they signed up for two years ago.

None of these talk to each other. So the same customer lives in five places. In billing she is "Mary Anne", in the scheme register she is "M A Joseph", and in the owner's phone she is saved as "Mary Aunty Thrissur". When she misses a payment, nobody connects it to the necklace she bought last month, or to the fact that she has stopped replying to offers.

Meanwhile the big jewellery chains and travel companies have their own apps, loyalty programmes and marketing teams. The neighbourhood shop is going after the same customer with a register and a phone. It still wins a lot of the time, because people trust the shop they know. But it gets a little harder every year.

A gold purchase savings scheme and a trip fund are similar

We started with jewellers, because the gold scheme was where the pain was clearest. The obvious next step was to build a very good jewellery app.

Then we looked properly at travel agencies. A trip fund works exactly like a gold scheme. The customer pays a fixed amount every month towards something they want later. The business keeps count, sends a reminder when a payment is due, maybe adds a bonus at the end, and the customer spends it at the counter. One saves in grams of gold, the other in rupees towards a holiday. The steps are the same.

So we stopped asking "what does a jeweller need?" and started asking "what does every one of these shops do with a customer?" The list was shorter than we expected:

  1. Catch the customer's eye with an offer.
  2. Help them save a little every month.
  3. Arrange a top-up loan from a regulated lender when the savings fall short.
  4. Get them to come in on a planned day, not just walk in.
  5. Close the sale with one clear bill.
  6. Thank them, and invite them back for the next plan.

A jeweller and a travel agent use different words for each of these. They use different units too, and a few rules really are different: the gold rate changes every day, and a trip has a departure date. But the jobs are the same. That is why GrowMagik is one platform, and not a jewellery app with a travel app tied on later.

In practice it means we build each job once. Whatever changes from trade to trade, we try to keep as settings and words, not as separate code. A jeweller's customer sees "gold scheme" and a travel agent's customer sees "trip fund", but behind the screen it is the same plan, kept by the same rules. Some things did need their own code, like the daily gold rate. We just made sure those stay the exception.

Four decisions that came out of this

Once we saw it as one platform for many kinds of shops, a few other decisions more or less made themselves. Each one gets its own posts later in the series.

The shop's name, not ours. Customers trust their jeweller. They have never heard of us, and they don't need to. So the app on the customer's phone carries the shop's own name and logo, and every WhatsApp message goes out in the shop's name. GrowMagik stays in the background.

One customer, one record. Every plan, payment, visit, bill and reward point sits on one profile, across every branch of the shop. Mary Aunty is one person again.

Every shop's data kept apart. Many shops share one platform, but no shop can ever see another shop's customers. Each shop's data lives in its own separate section of the database. We'll explain how, and why we chose that over the other options, in the coming posts.

Cheap to run from the first day. A neighbourhood shop will not pay what a big company pays for software, so the platform has to cost very little to run for each shop. We started with everything running as one single program, and we split parts out only when the traffic gives us a reason. That one choice shaped a lot of our code.

Coming up next

On Thursday we draw the whole of GrowMagik on one page: the layers, and the building blocks every screen and app stands on. After that come the cost rules we set before writing any code, why all our code lives in one place, and how one set of code serves many kinds of shops.

A new post goes up every Tuesday and Thursday. If you have built software for small businesses in India, we would like to hear what you got wrong. We got plenty wrong ourselves, and we'll be honest about it as we go.

Discuss this post on LinkedIn →

Series

How we built GrowMagik

Part 1 of 1

  1. 1
This is the first partThis is the latest part

Follow along

New posts appear here first. Follow QuickBee to see the rest of what we are building.