Most of what gets called "performance marketing" for Bangladeshi e-commerce is just Facebook ads with a Boost button. I want to write about what the actual work looks like — the parts that are different in Bangladesh than in markets the big international playbooks were written for, and the parts where the playbooks still apply.
This isn't a comprehensive guide. It's notes from running paid acquisition for 57 e-commerce clients at Ngital over the past several years, focused on the things that surprised me, the things that took longest to learn, and the things I'd want a new client to understand before we started working together.
The cash-on-delivery problem changes everything
The single biggest factor that makes Bangladeshi e-commerce performance marketing different from the international playbook is cash on delivery. Depending on category, CoD is 60-90% of orders for most Bangladeshi e-commerce operations.
CoD changes the unit economics of paid acquisition in ways that aren't obvious until you've absorbed several painful months of it.
The order placement event isn't the conversion. The platform you're optimizing on — Meta, Google, TikTok — sees the order placed and counts it as a Purchase. Your ad campaigns optimize toward more orders placed. But in CoD reality, 20-40% of those orders never actually get delivered. Returns at the doorstep. Wrong addresses. Customers who don't pick up the phone when the delivery agent calls. Customers who change their mind. Customers who ordered while drunk on a Friday night and won't take the package when it arrives Tuesday morning.
If your tracking sends "Purchase" the moment the order is placed, you're feeding the ad platforms inflated conversion data. The platforms then optimize toward audiences that place orders — not toward audiences that actually buy things. Over months, this drifts your campaigns toward worse and worse audiences from a real revenue perspective.
The fix is straightforward in concept and operationally annoying in execution: send the conversion event when the order is actually delivered and paid for, not when it's placed. This means integrating your delivery confirmation system with your conversion tracking and sending events to Meta CAPI, Google Ads, and other platforms only after successful delivery.
The technical implementation I covered in detail in Conversion API Setup Across All Major Platforms. The operational implementation is harder because it requires coordination between your warehouse, your delivery partner (typically Pathao, Sundarban, RedX, eCourier, or whoever you use), and your tracking infrastructure.
Brands that fix this typically see 20-30% improvement in real campaign performance over 60-90 days, even with no other changes. The lift comes from the platforms finally optimizing toward audiences that actually convert in revenue terms.
The channel mix that actually works
The default channel mix for most Bangladeshi e-commerce brands looks roughly like this when they come to us: 85% Facebook, 10% Boost-from-Page, 5% everything else combined. That mix made sense in 2017. It doesn't make sense now.
Here's the channel mix I'd argue for in 2026, calibrated for a Bangladeshi e-commerce brand spending BDT 5-30 lakh monthly on paid acquisition:
Meta (Facebook + Instagram): 50-65% of spend. Still the largest single channel for most categories. The audience is still here, the targeting still works for most consumer categories, and the platform's optimization remains strong when fed clean conversion data.
TikTok: 15-25% of spend. Has grown from "experimental channel" to "non-negotiable channel" for most consumer categories over the past two years. Particularly strong for fashion, beauty, FMCG, and food. Weaker for high-consideration purchases.
Google Ads: 10-20% of spend. Underused by most Bangladeshi e-commerce brands. Google Shopping in particular is dramatically underexploited — the inventory is cheaper than equivalent Meta inventory for high-intent shopping queries, and the audience converts at materially better rates because they're actively searching rather than passively browsing.
YouTube: 5-10% of spend, primarily for awareness layers. Works well for brand-building combined with direct response. Often dismissed by Bangladeshi performance marketers because the immediate-conversion attribution looks weaker; the actual impact on Meta/TikTok performance for the same brand is real but harder to measure.
Other (Pinterest, Snap, programmatic display): 0-5% combined. Useful for specific verticals; not core channels for most Bangladeshi e-commerce.
The percentages aren't sacred. A fashion brand will skew higher on TikTok and Instagram. A B2B e-commerce operation will skew higher on Google. A premium electronics brand will use more YouTube. The mix should follow the audience and category, not a generic template.
What I want to push back on specifically is the default 85% Meta allocation. That's not strategic. That's habit. Brands sticking with it have generally optimized themselves into a corner where Meta CPMs keep rising, audiences keep narrowing, and performance keeps degrading because every other brand in their category is bidding against them for the same shrinking pool of attention.
Channel diversification protects against this. The brands at Ngital that diversified channel mix aggressively over the past 18 months have generally outperformed the ones that stayed Meta-heavy.
What we've stopped doing
Some practices that were standard 2-3 years ago that I'd now actively argue against:
Stopped: Boost Post as a meaningful budget category. Boost is a tool for occasionally amplifying organic posts that are already performing. It's not a campaign strategy. Brands allocating substantial budgets to Boost are spending money for the platform's convenience rather than for results. Move that budget into proper campaigns with proper optimization objectives.
Stopped: Daily budget tinkering by managers without clear reasons. A manager logging into Ads Manager every morning and adjusting budgets by 15-20% based on yesterday's results does more harm than good in most cases. The platforms need time to learn; constant fiddling resets the learning. Set budgets weekly, evaluate weekly, adjust based on weekly patterns not daily ones.
Stopped: Lookalike audiences from tiny seed audiences. Lookalikes built from seed audiences under 1,000 customers don't have enough signal to produce useful matches. Brands creating lookalikes from 200-person seed lists are essentially generating random targeting. Either build proper seed audiences (10,000+ customers with good data) or skip lookalikes for that segment.
Stopped: Carousel ads as the default creative format for everything. Carousels are excellent for category browsing and discovery. They're typically worse than single-image or video ads for direct-response performance objectives. The default-to-carousel habit costs accounts CPA performance regularly.
Stopped: Treating Bangla copy as optional. For B2C e-commerce targeting Bangladeshi consumers, Bangla copy typically outperforms English copy by 20-40% on click-through and conversion rates. The brands still running English-only creative are leaving substantial performance on the table for no good reason.
Stopped: Single-pixel installation as adequate tracking. I've covered this extensively in the technical posts, but it's worth repeating in this context. A Meta Pixel firing browser-side only is no longer adequate tracking in 2026. Server-side Conversion API is now standard, not advanced.
What's actually working in 2026
The opposite list — practices that are working unusually well for Bangladeshi e-commerce right now:
Working: WhatsApp click-to-message campaigns for high-consideration categories. For products where customers want to ask questions before buying — furniture, electronics, services with customization — campaigns that drive to WhatsApp conversations rather than direct web checkout produce dramatically higher conversion-to-sale rates. The web traffic looks bad on paper (low session duration, no e-commerce conversion event). The actual sales close in WhatsApp threads that aren't visible to web analytics.
This requires integrating WhatsApp Business with your CRM and your conversion tracking so you can attribute sales back to the ad campaigns that originated the conversation. Most brands don't do this work, then conclude WhatsApp campaigns don't perform. They perform when measured properly.
Working: Long-form video creative for high-AOV products. Bangladeshi consumers, particularly outside Dhaka, increasingly watch 2-3 minute product videos before purchasing high-value items. The 15-second-ad orthodoxy works for impulse purchases but actively underperforms for considered purchases. Test long-form.
Working: User-generated content over brand-produced content for most categories. The lift varies by category but the pattern is consistent: UGC-style creative outperforms polished brand-produced creative in cost-per-result terms for most consumer categories. This isn't unique to Bangladesh but the magnitude of the lift is often larger here because polished international-style brand creative reads as inauthentic to many Bangladeshi audiences.
The operational challenge is producing UGC at scale. Brands that systematically partner with micro-influencers and content creators for UGC production end up with significant creative cost advantages over brands that produce all creative in-house.
Working: Aggressive remarketing for cart abandoners using detailed product data. Dynamic product ads showing the exact products customers viewed and abandoned, served within 24-72 hours, with reminders about stock or pricing, recover 15-25% of abandoned carts for most Bangladeshi e-commerce brands. This requires proper catalog integration and event tracking — work most brands haven't fully done.
Working: TikTok Shop and Spark Ads. TikTok's commerce features have matured significantly. Brands integrating product catalogs with TikTok and running Spark Ads against creator content are seeing performance that often beats their Meta numbers in fashion, beauty, and FMCG categories.
The metrics that actually matter
A surprising number of Bangladeshi e-commerce brands optimize toward metrics that aren't actually proxies for business success. Common examples:
CTR as a primary KPI. Click-through rate matters but optimizing for it directly often degrades conversion rate. The audiences that click most aren't necessarily the audiences that buy most. Care about CTR as a diagnostic; don't optimize toward it as a goal.
CPM minimization. Cheaper impressions aren't better impressions if they're reaching the wrong audience. Brands obsessing over low CPMs often end up with high CPAs.
Conversion volume without value consideration. Two hundred conversions averaging BDT 1,200 each is a worse business outcome than 150 conversions averaging BDT 2,500 each, even though volume is lower. Optimize toward value where the platform supports it, not toward conversion counts.
ROAS at first conversion. Reported ROAS based on first-purchase value misses lifetime value entirely. A brand with strong repeat purchase rates should accept lower first-purchase ROAS than a brand with one-time purchases, because the actual customer value is higher. Most Bangladeshi e-commerce brands don't calculate or use lifetime value data in optimization decisions.
The metrics worth caring about:
Delivered ROAS (after CoD return rate accounting). Real revenue, not order-placement revenue.
Cost per delivered customer, segmented by acquisition channel and category. Tells you which channels actually produce paying customers, not just clicks.
Repeat purchase rate by acquisition channel. Customers acquired through different channels often have dramatically different lifetime values. Channels that produce one-time deal hunters look efficient in first-purchase metrics but produce poor lifetime value. Channels that produce loyal customers look expensive in first-purchase metrics but compound value over time.
Time-to-second-purchase. For categories with repeat purchase potential (FMCG, beauty, fashion), the gap between first and second purchase predicts long-term customer value. Tracking and optimizing toward shorter time-to-second-purchase often improves overall profitability more than chasing more first purchases.
These metrics require proper data infrastructure to calculate. Most Bangladeshi e-commerce brands don't have that infrastructure built. Building it is generally a higher-leverage investment than another round of creative testing.
The honest reality of what's hard
A few things about Bangladeshi e-commerce performance marketing that I rarely see discussed honestly:
Attribution is genuinely broken for most accounts. Between iOS privacy changes, ad blockers, CoD cancellation gaps, multi-device journeys, WhatsApp dark conversions, and offline component of many purchase decisions, the attribution data most brands work from is probably 40-60% accurate at best. Anyone claiming precise ROAS attribution in this environment is either using tools that don't actually measure what they claim, or they're not being honest about uncertainty ranges.
Creative production is the binding constraint for most accounts. Brands with adequate budgets and decent strategy often hit performance walls because they can't produce enough creative variations to feed the algorithms properly. A typical Meta campaign benefits from 8-15 fresh creative concepts monthly. Most Bangladeshi brands produce 2-3. The creative gap explains performance gaps more than any other single factor for established accounts.
Most agencies don't actually optimize accounts daily. The pitch describes daily monitoring and optimization. The reality is usually weekly review with mostly-passive management between reviews. This isn't inherently bad — over-optimization is also a problem — but the disconnect between pitched service and actual service creates trust problems later.
Brand and performance aren't really separate. The performance marketing playbook treats brand as a separate concern. In practice, brands with strong organic awareness, content presence, and earned media run substantially more efficient paid campaigns than brands without those layers. The dollar-for-dollar comparison is misleading when comparison brands have different underlying brand strength.
Where I'd start with a new e-commerce client
If a new Bangladeshi e-commerce brand came to Ngital tomorrow with BDT 8 lakh monthly to spend on paid acquisition and asked where to start, my actual answer would be roughly this:
Month 1: fix tracking first. Server-side Conversion API for Meta. Enhanced Conversions for Google. TikTok Events API. CoD-adjusted conversion events firing only on delivery. Most brands resist this because they want to start running ads immediately, but spending Month 1 fixing tracking infrastructure produces dramatically better Month 2-12 performance than spending Month 1 on campaigns running against broken tracking.
Month 2: rebuild creative production. Set up systematic UGC sourcing through micro-influencers. Produce 12-15 creative concepts in the first 30 days. Test ruthlessly. Most accounts can lift performance significantly through creative refresh alone if they haven't done it systematically before.
Month 2-3: diversify channel mix. Move from Meta-only to Meta + TikTok + Google as primary channels. Specifically activate Google Shopping if catalog integration permits. Move budget allocation toward whatever shows early traction.
Month 3-4: implement WhatsApp Business integration for categories where pre-purchase conversation is valuable. Connect WhatsApp threads to CRM. Attribute sales back to source campaigns.
Month 4-6: build proper measurement. Delivered ROAS reporting. Lifetime value calculation by acquisition channel. Repeat purchase rate tracking. Use this data to reallocate spend toward channels producing actual customers rather than channels producing cheap conversions.
This timeline isn't aggressive. It's deliberate. Most agencies will pitch month-1 results because that's what clients want to hear. The honest reality is that the first 90 days are usually about building infrastructure that makes months 4-24 perform well, and brands that try to skip that infrastructure investment usually plateau by month 6 wondering why their early gains stopped compounding.
If you're running e-commerce in Bangladesh and the description above sounds different from what your current setup looks like, you're not alone — most accounts I audit fall short on the same fundamentals. The fixes are operationally heavy but well-mapped. The brands that do them outperform the ones that don't, consistently and substantially.
Ngital runs paid acquisition across Facebook Ads, Google Ads, and TikTok Ads for Bangladeshi e-commerce brands, with the tracking, creative production, and measurement infrastructure that makes paid campaigns actually work.