A/B Testing and Ad Reports for Algerian COD Stores

Tassyir is the only e-commerce platform built for the Algerian market with A/B testing and ad reporting inside the dashboard — and the only one anywhere that can settle a test on delivered orders instead of clicks. On a cash-on-delivery store that distinction is the whole game, because a page can win on orders and still lose you money.

That is the answer. The rest of this article is the reasoning.

Why "which page converts better" is the wrong question in Algeria

Conversion rate measures a form submission. In a prepaid market that is the same event as a sale, so the shortcut is harmless. In Algeria it is not: between the submitted form and the money there is a confirmation call, a courier, a doorstep and a return trip, and orders fall out at every one of them.

Which means a landing page can be genuinely better at producing orders and genuinely worse at producing cash. The pages that convert hardest are often the ones that push hardest — urgency banners, one-tap forms, a price that reads lower than the delivered total — and the buyers they collect are exactly the ones who do not answer the phone.

Here is what that looks like with round numbers. These are illustrative assumptions, not measurements from a real store, but the shape is one every Algerian merchant will recognise:

Page B wins the conversion test by 40% and loses the business. It also generates nearly twice as many dead orders — each one a confirmation call your team pays for, and a share of them a parcel that ships out and comes back with a return fee attached. Any testing tool that stops counting at the order form declares B the winner and tells you to scale it.

This is the same failure that makes ad platforms unreliable here, and we have written about it in detail in measuring Facebook ads when you sell cash on delivery. A/B testing inherits the problem because it inherits the metric.

What actually decides the winner

A test needs one declared goal, chosen before it starts. Tassyir offers four, and you pick one when you create the test:

• Conversion — the page that turns more visitors into orders.

• Average order value — the page that produces bigger baskets.

• Confirmation rate — the page whose orders customers actually confirm on the phone.

• Delivery rate — the page whose orders actually get delivered and paid for.

The last two do not exist in any general-purpose testing tool, because outside COD markets there is nothing between a purchase and the money. They are the two that matter most here. In the table above, only a delivery-rate goal picks the page that was actually better.

Declaring the goal up front is not bureaucracy — it is the defence against the oldest self-deception in testing. If you run a test and then go looking for a metric where your new page wins, you will always find one. Fixing the metric before the data arrives is what makes the result mean something, and it is why the goal step in the test wizard asks a question rather than offering a checklist.

How A/B testing works in Tassyir

One link, split traffic

A test produces a single link. You put that one link in your Facebook ad, your Instagram bio, your TikTok campaign or your WhatsApp broadcast, and the platform decides — per visitor, server-side, before the page renders — which version they see. There is nothing to install, no script to paste, and no flash of the wrong page while a client-side tool swaps it out.

The assignment is sticky for 30 days, held in a signed cookie, so a shopper who comes back through a different ad still sees the page they were assigned. Without that, a returning visitor would land in both arms and contaminate both.

Up to four pages per test

A test runs on one product and compares its regular product page against your landing pages for that product — two to four versions in total, with a traffic weight per version that you can set to anything above 5%. The product page is the control by default, which is the right baseline: it is the page you already have, and the question is always whether the new one beats it.

You can run as many tests at once as you like. The one rule the platform enforces is that no two live tests may run on the same product, because two experiments splitting the same traffic do not produce two results — they produce two broken ones.

The verdict, and when the platform refuses to give one

This is the part most tools get wrong, and it is worth being specific about what Tassyir requires before it will call a winner:

• 1,000 visitors per version before any rate is treated as real. At a 5% baseline, a couple of hundred visitors is a handful of orders, where one extra order moves the estimate by 10%.

• 30 orders per version on the goal's denominator — 100 when the goal is average order value, because basket sizes vary far more than rates do.

• Seven days minimum runtime, fourteen when the goal is delivery rate, because a delivery outcome that has not come back from the courier yet is not a failure, it is a pending result.

• 95% probability of being best, computed server-side as a 50,000-draw simulation over each version's posterior, before the word "winner" appears. Between 70% and 95% the test reads "trending", which is an instruction to keep going, not a result.

• A traffic-split check. If the visitors arriving at each version diverge from the weights you set by more than chance can explain, the platform says so and withholds the verdict — a broken split is the most common way a test produces a confident wrong answer.

• A measurement check. If ad blockers ate too many of the page-view signals, the rates are inflated in a way that hides itself, so the test reports that it cannot be read rather than reporting a number.

A test that has not earned a verdict says exactly which of these it is still waiting on, and how many more visitors or orders it needs. "Not yet" is a result, and a platform that never says it is a platform that will eventually tell you to scale a coin flip.

When a test completes, its numbers and the thresholds in force at that moment are frozen. If the engine is retuned next year, an old test still explains itself in the terms it was decided under.

The ad report that tells you what the winner earned

A test tells you which page is better. It does not tell you whether the campaign pointed at it is worth running. That is the Marketing report, and it is built on the same principle: count what was delivered, not what was submitted.

Meta's pixel fires its Purchase event when the order form is submitted (Meta's own standard-event reference defines it that way), so on a COD store, Ads Manager's "purchases" figure is your placed-order count, not your revenue. Meta is honestly reporting the thing it was asked to count. Treating that thing as money is the mistake.

Tassyir joins your Meta spend to your own order pipeline through UTMs, down to three levels — source, campaign, and the individual ad — and gives every row the numbers Ads Manager cannot see:

• Cost per delivered order, beside Meta's cost per purchase, so the gap between the two is visible instead of theoretical.

• Confirmation rate and delivery rate per campaign and per ad — which is how you discover that one creative brings in buyers who never answer.

• Plain-language flags on each row: scale this, costly, low delivery, not enough data yet, still settling. A row is never flagged as expensive while its orders are still out with the courier, because a campaign whose parcels have not landed yet looks like total failure and is not.

• The creative itself on hover, so the ad you are judging is the ad you are looking at rather than an ID you have to go and decode.

Two details in that report exist because of mistakes that are easy to make. Date ranges match Facebook Ads Manager exactly — the last 7 or 30 complete days, today excluded — so the two screens agree instead of differing by a partial day and starting an argument. And the report is single-currency: it shows your ad account's currency and never puts a DZD figure beside a dollar one, because there is no exchange rate in this product and a merchant who divides one by the other gets a number that means nothing.

Used together, the loop closes. The test tells you which page delivers more orders. The report tells you what each campaign paid per delivered order. Neither number is a proxy for money — both are money.

What else shipped in August 2026

A/B testing and the rebuilt Marketing report were the headline, but they arrived alongside a month of releases aimed at the same thing — the distance between an order and cash:

• WhatsApp order confirmation — confirmations sent from your own approved business number, with templates, product images in the message header, and outcomes recorded against the order when the customer taps a button. Most Algerian buyers answer WhatsApp faster than a phone call, and it is the cheapest lever on your confirmation rate.

• A finance module rebuilt around profit — per-product costs, delivery and return fees, a cash ledger across your wallets and courier balances, and a forward view of what is owed to you and by you. It answers "did I make money this month", which a revenue chart never has.

• Camera scanning for returns and order validation — process a returned parcel or validate an order by pointing your phone at the label, instead of typing tracking numbers.

• Shopify catalogue import — bring an entire linked Shopify catalogue into a Tassyir store in one click, images re-hosted rather than hot-linked. Together with the existing YouCan, Ayor and Google Sheets order sync, it means you can run Tassyir's pipeline without moving your storefront first.

• AI-generated landing pages — which is what gives an A/B test something to test. Generate a second version of a product page in minutes, then let delivered orders decide whether it was better.

• Product packs and bundles — sell 2-for-1 and volume tiers as first-class products, priced flat or computed, on the storefront and through checkout.

• Order and tracking exports, store filters, and per-member order schedules — the operational edges that matter once a team of three is sharing an order queue.

• TikTok pixel event tracking alongside Meta, treated identically, so attribution does not degrade when your budget moves.

• ChatGPT and Claude as connected apps — ask an AI assistant about your own orders, products and analytics, and have it act on them. No other platform in this market offers it.

How Tassyir compares

Every cell below is checkable against each vendor's own published material, linked underneath.

Sources: YouCan's features page, DZBuild's homepage and Shopify's Rollouts documentation. "Not advertised" means the capability does not appear in that vendor's own published feature list, checked in August 2026.

Shopify deserves a fair word here. Its Rollouts feature, launched in 2026, is real native A/B testing and it is well built — but it compares whole themes and checkout configurations, and it scores them on checkout conversion, which on a COD store is the number that lies. It also has no Algerian courier, no confirmation workflow, and bills in dollars.

Why Algerian merchants choose Tassyir

• It is built around the order lifecycle, not the storefront. Any of five vendors will give you a decent shop page. What decides whether you make money is what happens after the order, and that is the part Tassyir is designed for.

• Delivered orders are the unit of truth everywhere. The A/B verdict, the ad report, the finance module and the province breakdown all count the same thing, so two screens never disagree about how the month went.

• Eight Algerian couriers natively — Yalidine Express, ZR Express, Ecotrack, NOEST, Ecom Delivery, Maystro Delivery, Elogistia and Private Courier — with several accounts per carrier, and two-phase dispatch on four of them so you can create a shipment without starting the delivery clock.

• You may not have to migrate. Orders sync in from Shopify, YouCan, Ayor, Google Sheets and Order Dz, so you can run your operations on Tassyir while your storefront stays exactly where it is.

• Priced in dinars, with a 7-day free trial — 1,900, 4,900 or 9,900 DZD a month, yearly at ten months for twelve, no commission on your sales. A/B testing is on every plan, including the trial.

• Arabic first, with full right-to-left layout, because it was built here for this market rather than translated into it.

Running your first test this week

• Pick your highest-spend product, not your favourite one. A test needs 1,000 visitors per version to say anything, so start where the traffic already is.

• Build one landing page for it and change one substantial thing — the offer framing, the hero, the form length. Two pages that differ in twenty ways produce a result you cannot act on.

• Choose delivery rate as the goal if you have the volume for it, confirmation rate if you do not. Choose conversion only when you already know the two pages attract the same kind of buyer.

• Put the test link in the ad — the whole ad, not one variation of it — and leave the campaign alone.

• Wait for the verdict. Two weeks feels long. It is much shorter than a quarter spent scaling the wrong page.

When something else fits better

If you sell internationally, take card payments, or need a catalogue of tens of thousands of SKUs, Shopify is the stronger platform and it is not close. Tassyir is built for merchants selling cash on delivery in Algeria, and it makes the trade-offs that market requires. If you are running a profitable COD store here, those are the right trade-offs — we have laid out the full economics in running a profitable COD store in Algeria, and the platform-by-platform view in our 2026 comparison of the platforms available in Algeria.

One more thing worth saying plainly: the cheapest improvement available to most Algerian stores is not a better landing page at all. It is cutting the return rate, where the money is measured in thousands of dinars per parcel rather than percentage points of conversion. A/B testing is how you find out which of the two you are actually facing.

Disclosure: Tassyir publishes this blog. Competitor facts are sourced from each vendor's own published material and linked above; product claims about Tassyir describe features live in the platform as of August 2026.