How to Improve Your ROAS: A Practical Guide for E-commerce Brands
By Jamie Slabber
ROAS is an output, not a lever. You improve it by pulling on the five things that actually produce it — creative volume and quality, offer and average order value, on-site conversion rate, tracking accuracy, and repeat purchase behaviour.
How to Improve Your ROAS: A Practical Guide for E-commerce Brands
By Jamie Slabber — Creative Director, Heir Digital Last updated: August 2026
Short answer: ROAS is an output, not a lever. You improve it by pulling on the five things that actually produce it — creative volume and quality, offer and average order value, on-site conversion rate, tracking accuracy, and repeat purchase behaviour. Cutting spend raises ROAS on paper while shrinking the business. Here is how to work on each input in order of impact.
Start by defining which ROAS you mean
Most arguments about ROAS are actually arguments about definitions. There are three numbers in play and they rarely agree:
Platform ROAS is what Meta or Google reports. It uses the platform's own attribution window and counts conversions it believes it influenced. Every platform over-attributes, and if you run Meta and Google simultaneously they will both claim the same sale.
Blended ROAS is total revenue divided by total ad spend across all channels. It cannot be gamed by attribution settings, which is precisely why it is uncomfortable — it is usually much lower than the platform numbers you have been reporting internally.
Break-even ROAS is the multiple you need to cover product cost, shipping, payment fees and overhead. It is set by your margin, not by your ambitions.
Work out break-even first, because it converts an abstract goal into a specific target. A brand at 70% gross margin breaks even around 1.4x. A brand at 30% margin needs roughly 3.3x before it makes a cent. Two brands can post identical ROAS and one is profitable while the other is quietly funding its own decline.
If you take one thing from this article: manage to blended ROAS, and know your break-even number before you set a target.
Lever 1: Creative volume and quality
This is the highest-impact lever in 2026 and it is not close.
Meta's shift to consolidated, algorithmic delivery removed most of the manual targeting levers agencies used to differentiate on. What remains under your control is what you feed the system. When delivery is automated, creative is the targeting — the ad itself determines who engages, and the algorithm follows.
The practical implication is that creative throughput has become a performance variable. An account shipping four new concepts a month and an account shipping thirty are running fundamentally different businesses, even with identical media buying.
This matters more in South Africa than in larger markets. The addressable audience for most consumer categories here is small enough that frequency climbs quickly and a winning concept can fatigue within four to eight weeks at meaningful spend. This is the mechanism behind a pattern most SA e-commerce brands recognise: a strong first two months, then a steady CAC climb that nobody can explain. The creative did not get worse. The audience simply saw it too many times.
What to do: Build a standing creative pipeline rather than reacting to fatigue after it shows up in the numbers. Define how many genuinely distinct concepts enter the account each month — distinct meaning different angles, hooks and formats, not colour variations of the same asset — and hold to it whether or not current performance is good.
Lever 2: Offer and average order value
ROAS is revenue over spend. Most brands only work on the denominator.
Raising AOV improves ROAS without touching the ads at all. The mechanisms are unglamorous and reliably effective: bundles, multi-buy pricing, subscription options, free shipping thresholds set just above current AOV, and post-purchase upsells that add margin without adding acquisition cost.
One counterintuitive finding worth testing: fixed rand-off thresholds tend to outperform percentage discounts, with one documented test showing a 21% AOV lift. Percentage-off trains customers to wait for sales. Threshold mechanics ("spend R800, get free delivery") push basket size up instead of pulling price down.
Also worth auditing: whether your discounting is masking a ROAS problem or creating one. Heavy promotional dependency inflates conversion rate while destroying the margin that determines your break-even multiple.
Lever 3: On-site conversion rate
Doubling conversion rate doubles ROAS at identical spend. Yet most brands respond to poor ROAS by changing agencies rather than fixing the page the traffic lands on.
Most e-commerce sites convert between 1% and 4%, with wide variation by category. The gap between two stores with similar traffic is usually driven by checkout friction, payment options, mobile usability and post-add-to-cart drop-off rather than traffic quality.
Two things deserve attention before anything else. First, mobile — it carries the majority of e-commerce traffic and typically converts below desktop, which makes mobile checkout the highest-value place to spend an optimisation cycle. Second, cart abandonment, which Baymard Institute documents at roughly 70% on average, a figure that has barely moved in five years.
If your paid media is being judged on ROAS while your product page has not been touched in eighteen months, you are measuring the wrong thing.
Lever 4: Tracking accuracy
A meaningful share of South African Meta accounts still run on the browser pixel alone, or with the Conversions API firing values that do not match it. The consequence is not just bad reporting — it is bad optimisation. Duplicate or mismatched conversion signals teach the algorithm to trust campaigns that are not working, and it then spends more there.
What to check: Is server-side tracking live? Are events deduplicated? Do the conversion values match your actual order values, or are they firing at a default? Does the number of purchases in the ad platform bear any relationship to the number in your back end?
Fix measurement before scaling spend. Scaling on top of broken data multiplies the error rather than the revenue.
Lever 5: Retention and repeat purchase
First-order ROAS is a poor measure of a business that sells consumables, subscriptions or anything replenishable. If a customer's second purchase arrives at zero acquisition cost, your economics are set by lifetime value, not by the first transaction.
This changes what you can afford to bid. A brand with a 40% repeat rate can profitably acquire at a first-order ROAS that would bankrupt a one-purchase business. Email and WhatsApp flows, subscription mechanics and post-purchase sequencing are therefore acquisition levers, not retention afterthoughts — they raise the ceiling on what you can spend to win a customer.
Three things that lower ROAS while appearing to raise it
Cutting spend. Reducing budget concentrates delivery on your warmest, cheapest audiences. ROAS rises, revenue falls, and the brand shrinks. This is the most common self-inflicted wound in performance marketing.
Over-weighting retargeting. Retargeting posts spectacular ROAS because it takes credit for people who were going to buy anyway. Shifting budget toward it improves the dashboard and starves the top of funnel. Within a few months there is nobody left to retarget.
Chasing platform-reported numbers. Optimising toward the metric with the most generous attribution window is optimising toward a measurement artefact.
Frequently asked questions
What is a good ROAS for a South African e-commerce brand?
There is no universal figure, because the answer is determined by gross margin, repeat purchase rate and fixed costs. A high-margin supplement brand and a low-margin apparel brand need very different multiples to survive. Calculate break-even from your own numbers and treat any target quoted before someone has asked about your margins as guesswork.
How long does it take to improve ROAS?
Ninety days is a fair window for structural work. The first month covers tracking correction, account restructuring and initial creative production. The second produces readable test results. The third is where compounding starts showing in blended numbers. Quick wins do exist — usually in tracking or checkout — but sustained improvement comes from the creative pipeline, and that takes a quarter to demonstrate.
Should I lower my target ROAS to scale?
Often, yes. There is an inverse relationship between ROAS and volume: the more you spend, the further into cold audiences you reach, and the lower efficiency goes. The right question is not "what is our highest possible ROAS" but "what is the most profit we can generate," which usually sits at a lower multiple and a much higher spend.
Does creative testing hurt ROAS in the short term?
Yes, slightly, and it is the cost of not stagnating. Testing budget produces losers by definition. Ring-fence it — commonly 10–20% of spend — so that experimentation does not get cut every time a weekly report looks soft.
About the author
Jamie Slabber is Creative Director at Heir Digital, a Cape Town paid media and performance marketing agency working with DTC and e-commerce brands across Meta, Google and TikTok. He leads the agency's creative and account teams, and his background spans agency ownership, festival and events marketing, and community-led brand building. He holds a BA in Visual Communication from Stellenbosch Academy of Design.
[LinkedIn] · [Email]
Heir Digital is a Cape Town paid media agency working with DTC and e-commerce brands across Meta, Google and TikTok. [Request a free account audit.]