Story 1 of Checkout 2026, our new original series on Tanzanian ecommerce

The First Wave (2014-2019)

“Based on our review of the path to success, we have made a difficult decision to cease our operations in Tanzania,” African e-commerce pioneer Jumia announced on November 27, 2019.
“While Tanzania has strong potential and we’re proud of the growth we’ve collectively seen stemming from Jumia’s adoption, we have to focus our resources on our other markets.”

This exit came as part of a wider retreat from smaller markets (including Rwanda and Cameroon) where Jumia struggled to make unit economics work. According to a 2019 analysis by Olumide Olusanya, the company’s total cost to attract and serve a user (Sh141,000) was five times higher than the actual money that a user brought back to the business.

Olusanya (who then operated Nigeria’s biggest online supermarket) also observed that Jumia had “absolutely NO MARKET POWER” over the people selling on its platform. While third-party merchants moved billions of shillings (aka gross merchandise value—GMV) through the site, Jumia was only catching 2% of the money. These commissions were often not enough to pay for the retailer’s massive logistics and office costs.

So leaving Tanzania was part of “optimization efforts” that aimed to create a sustainable, profitable, and scalable business model across the continent.

Jumia cited the need to focus resources on larger markets. But what the company didn’t say publicly was that Tanzania presented nearly every challenge that made online delivery a tough industry.

In 2017, 99% of all payments were conducted in cash, and there were very few restaurants that had digitized their menus or vendors listed on services like Google Maps. On top of that, startups faced high customer acquisition costs (CAC). You had to spend handsomely on marketing to win trust and cause behavioral change in a cash-driven ‘let me see it first’ society.

But thanks to mobile money wars and the nation’s first wave of fintech founders, we made some progress four years after Jumia’s exit.

By 2023, approximately 13% of Tanzanian adults (approx. 4.4 million people) could be termed as monthly active users of digital merchant payment solutions, which e-commerce platforms in their truest definition rely on.

That 4.4 Mn population, in our opinion, gives insight into the maximum number of Tanzanians who can buy or sell anything online. In hindsight, it also implies that in the five years Jumia operated in Tanzania (2014-2019), order frequency was just too low to justify logistics infrastructure costs.

Six years post-Jumia, local operators are trying again with different assumptions about what might work.

JumlaJumla

We visited David Mallya (CEO) and Alex Makule (CTO) at their SkyCity Mall office in Dar es Salaam to talk about MsosiApp, their food and grocery delivery app that launched in March 2025.

Both questions and answers, from our January 16 exclusive interview, have been edited for clarity and brevity.

For investment or partnership inquiries contact David at +255 766 139 709.

David: JumlaJumla is the company name and the bigger vision, which is to build an end-to-end trading platform for East Africa. But we realized we were trying to build too much at once.
We had code for groceries, electronics, clothing, everything. The scope kept growing and we couldn’t ship anything. So we took the code we’d already written for food and grocery delivery and launched that as MsosiApp. It’s our first product in the market.
The idea is that MsosiApp proves we can acquire customers and move things reliably. Once that’s working, we expand to the other parts of the JumlaJumla vision like fulfillment centers, credit products, and B2B trading. But right now, ten months after launch, we’re completely focused on making the delivery business work.

How’s it going so far?

David: We have onboarded hundreds of vendors over time, with around 200 currently active.
We have a fleet of 100 electric bikes with about 50 active riders serving several thousand registered users across Dar.
We’re processing between 1,500 and 2,000 orders per month right now.

Wait, per month? Not per day?

David: Per month. That’s about 50 to 65 orders per day currently.
Our near-term target is to consistently reach 500 to 1,000 orders per day as we expand our fleet, vendor base, and customer retention. That gap between where we are and where we need to be is significant, and it’s been harder to close than we projected.

Why is it harder than you expected?

David: Because the math you do in business plans is completely different from reality. You model consistent order volume and it looks achievable.
Then you’re in the market and you realize getting people to actually order regularly is incredibly difficult.
Most people here don’t have the habit of using delivery apps. They walk to the restaurant or call a vendor directly. We have to prove our way is genuinely better through repeated reliable service, and building that trust takes time.

What’s the actual addressable market here?

Alex: As you mentioned earlier, by 2023, about 4.4 million Tanzanian adults were using digital merchant payments regularly. That’s the real number for e-commerce. You can’t just look at smartphone penetration and assume those are customers.
People have phones but that doesn’t mean they can afford to order in-app three times a week. Purchasing power is the true constraint.
David: And adoption is concentrated. The pandemic helped create some behavior change. Before COVID almost nobody was ordering delivery regularly in Tanzania. Now there’s adoption but it’s mostly in specific neighborhoods with higher incomes. We’re not trying to serve everyone. We’re trying to serve the people who have both purchasing power and digital payment habits.

You mentioned electric bikes. How does that work?

David: We’re one of the early companies in Tanzania moving fully toward EV-powered last-mile delivery.
We have 100 electric bikes right now. Some we own outright, some are partner-financed. About 50 riders are active at any given time.

Why go fully electric?

David: Operating costs.
Fuel prices keep going up. But electric bikes are cheaper to run once you have the charging infrastructure in place. We’re also looking to install battery swapping stations, which will let us scale the fleet to 200-plus bikes without riders sitting around waiting to charge.

Walk us through what actually happens when someone places an order.

David: Customer opens the app, browses restaurants or shops, places an order. Then someone from our team calls the vendor to verify.
We literally call every single vendor and say, “Hello, you have an order. Can you prepare this?” It’s completely manual. At least, for now.

Why are you calling vendors manually? That can’t possibly scale.

David: Because vendors don’t reliably check their apps. They’re busy running their businesses. Or they’re out of stock and haven’t updated their menu. If we don’t call to verify, the order sits there and the customer gets angry and blames us. So we solve it with a phone call.
You’re right that it doesn’t scale, but right now we care more about getting the experience right than about scaling. If we grow and the experience breaks, customers never come back.

What happens when a vendor tells you they’re out of something?

Alex: Our system already knows what that customer usually orders based on their purchase history, so it automatically recommends alternatives at the same price point. But we still have to call the customer and ask if they want the substitute. It’s not fully automated yet.
David: This is the reality of operating here. Most vendors don’t have the systems or habits to operate digitally. A restaurant owner has their menu but they don’t know how to structure it for an online platform. They don’t take good product photos. They don’t write descriptions that work online. So we do all of that for them.
Our team goes to every restaurant, photographs everything, records prices, and builds the complete menu. We’ve done this for over hundreds of vendors now.

That’s incredibly labor intensive. Why not just let vendors upload their own stuff?

David: Because they can’t or won’t. It’s a skill and infrastructure gap. If you look at Alibaba, they have 8.5 million annual active sellers but those vendors already know how to operate online. They have e-commerce experience. In Tanzania, that capability doesn’t exist yet. So we’re building it from scratch, one vendor at a time.

Show us how you’re tracking all of this.

[Alex opens his laptop. A map of Dar es Salaam appears covered in colored dots of different sizes.]
Alex: This is what we call god mode. Every dot represents a customer. Gold dots are high spenders who don’t care about price. Green dots are regular users who are more price sensitive. The size of each dot shows their total spending.
[He zooms into Masaki where gold dots cluster thick across the screen.]
See Masaki here? Some of these customers order late at night and pay premium for convenience. They’re high lifetime value customers.
[Then to Kinondoni with fewer, mostly green dots.]
Kinondoni is different, and more price sensitive. They compare options carefully before ordering. We track all of this because every transaction teaches us something about customer behavior and preferences.

What are you actually doing with all this data beyond pretty maps?

Alex: We are building a data-driven intelligence layer that learns from delivery patterns, vendor performance, and customer behavior to improve decisions over time. Most delivery apps use basic recommendation algorithms.
There are four layers to our cognition system.
First is memory architecture. We have short-term memory handling real-time transactions, long-term memory storing every pattern we’ve observed, and what we call scar memory that tracks failures. When a vendor is consistently late or a delivery route always has problems, the system remembers that and adjusts future decisions.
Second is learning from consequences. When we recommend a substitute product and the customer accepts it, that decision gets weighted higher in future scenarios. When they reject it, the system learns from that. It’s reinforcement learning where every transaction improves the model.
Third is self-regulation. The system knows when not to act. We have three states: frozen, degraded, and active. When there’s data starvation or unsafe conditions, the system won’t make automated decisions. It requires human intervention.
Fourth is self-awareness. The system maintains confidence scores for every decision and explicitly refuses actions when it’s uncertain. It can explain its reasoning. We’re building systems that reason, remember, and regulate themselves, not just black box automation.

That’s a lot of sophisticated infrastructure for 50 to 65 orders per day. Why build all of this now instead of using simpler off-the-shelf solutions?

Alex: Because you can’t retrofit intelligence later. We’re building the foundation now so that when we scale to thousands of orders per day across multiple regions. The AI is already trained on millions of transactions and the architecture can handle it. Meanwhile, the marginal cost of processing more transactions on infrastructure built for scale is minimal.
David: And honestly, the technology isn’t our biggest challenge. The challenge is changing behavior and building trust in a market where most people have never used delivery apps regularly. The infrastructure will pay off when we scale, but getting to scale requires solving human problems, not technical problems.

What does this infrastructure cost to run?

Alex: About 600,000 shillings per month. That can handle about 80,000 people using the app at the exact same time.
The biggest cost, though, would be API calls if we hadn’t pivoted to open source options like OpenStreetMap. Previously, every time we calculated a delivery route through Google Maps, we would pay in dollars for the to-and-fro ride. Even a few dozen orders per day routing through mapping services adds up quickly. That’s why server configuration and optimization matters so much.

Walk me through your business model. How do you actually make money?

David: We take 10 to 12 percent of each order, which includes commission and delivery fees.
Average order value is about 25,000 shillings, so we make roughly 2,500 to 3,000 shillings per transaction. At our target of 500 orders per day, that’s around 270 to 320 million shillings in annual revenue. But we’re not at 500 orders per day yet. We’re working toward 100+.

That’s a big gap to close

David: It is. We’re in that zone where we've invested heavily in the fleet, the technology, the team, but we’re not at sustainable unit economics yet. Getting from 65 orders per day to 500 requires growing our active vendor base, improving retention, expanding the rider fleet, and building consistent customer habits. Each of those takes time and capital.
Alex: This is why we are currently fundraising to meet sustain our growth.

David: Continued growth in customer retention. Right now we’re targeting above 55 percent retention rate.
If we can keep customers ordering week after week, the order volume follows naturally. The signal we care about most is repeat behavior. Someone orders once, comes back the next day, comes back again. That tells us we’re delivering value.

You don’t accept cash at all. Doesn’t that limit your market even more in a country where cash is still dominant?

Alex: It does limit us, but cash created too many operational problems. Wrong change at delivery, security issues with riders carrying money, or reconciliation nightmares at the end of every day. We went mobile money and card-only and it simplified the entire operation.
The customers who can use mobile money are also the customers who order regularly anyway. We’re optimizing for quality customers, not maximum reach.

Your three-year target is 30 billion shillings in GMV. Walk me through how you get there from where you are now.

David: Multi-city expansion and growing our vendor base significantly. As I said earlier, right now we have 200 active vendors.
We’re targeting 5,000 vendors within two years. That requires systematic vendor acquisition and digitization across multiple cities. We need to launch Arusha, probably in the next six months, then Mwanza and one other city after that.
The math only works with hundreds of orders per day in multiple locations. Dar alone can’t get us to 30 billion.

What needs to be true to make Arusha successful?

David: We’d start with 30 to 40 riders and scale based on actual demand. The mistake would be launching too big and burning cash on infrastructure that isn’t being utilized.
We need to digitize vendors there, build relationships, prove the service works, then expand the fleet as orders grow. It’s the same strategy as Dar but we have to execute it carefully and not assume what worked here automatically works there.

Looking at your LinkedIn profiles, you list all these other products. JumlaJumla Delivery, Credit, Fulfillment centers, Collection Points. How much of that is real versus aspirational?

David: MsosiApp is real and in market. The others are part of the larger JumlaJumla vision but they’re not built yet. The idea is that food delivery is the wedge that acquires customers and proves we can move things reliably. Once that’s working sustainably, we expand to fulfillment centers for storage and processing, credit products so vendors get quick settlements and customers can buy now pay later, collection points so you don’t have to be home for delivery.
Alex: From a technical standpoint, the infrastructure we’re building works for any commerce problem, not just food delivery. We’re building the data and logistics layer that can support all of those products eventually.

That sounds ambitious for a company that isn’t consistently hitting 400 orders per day yet.

David: [laughs] It is ambitious. Look, we might be wrong about all of this. Other companies have tried and failed in this market.
We have a different approach and we think we understand the constraints better, but there’s no guarantee. We just have to execute one step at a time and find out. Right now that means making MsosiApp work sustainably before we think about anything else.

If you were starting over today, what would you do differently?

David: Start much smaller. We tried to cover all of Dar from day one.
I would launch with maybe ten restaurants in one specific neighborhood, prove the unit economics there with actual customers making purchasing decisions, then expand methodically to adjacent neighborhoods.
The gap between what looks achievable in business plan projections and what actually happens in the market is enormous. You learn that through execution, not planning.

What keeps you up at night right now?

David: Making sure we don’t grow faster than we can handle. If we acquire customers but the experience breaks, they never come back. In this market you get one chance, maybe two if you’re lucky.
So we’re obsessed with retention rate over growth rate. Sustainable growth where customers keep ordering week after week is what actually builds a business. Everything else is vanity metrics.
Alex: For me it’s making sure the technology scales smoothly as we grow. Right now everything works well at our current volume. But when we hit 500 orders per day across multiple cities, will the architecture still perform? Have we built the right systems? That’s what I think about constantly.

Exclusive: JumlaJumla Targets TZS 1.5 Billion in Revenue