How to scale an ecommerce business.

Almost every piece of scaling advice you will read is a list of tactics — increase ad spend, expand to new channels, improve your email flows, raise your prices, launch on marketplaces. All of them are sometimes right. None of them are useful without a diagnosis, because a business is never blocked by five things simultaneously. At any given moment, one constraint is binding, and effort spent anywhere else produces almost nothing. The skill in scaling is not knowing the tactics. It is identifying the constraint.

I run a paid media agency, so the honest disclosure is that I see this from a specific vantage point: hundreds of accounts where a brand arrived convinced that advertising was the bottleneck. Frequently it was not. The most valuable thing we do in a first engagement is often to tell a brand that their ad account is fine and their real problem is contribution margin, or repeat purchase rate, or a checkout that loses a third of the people who reach it. This guide is that diagnosis, written out.

The five constraints, in the order they usually bind

These are roughly sequential. A business rarely hits the fourth constraint before resolving the second, which is why scaling advice aimed at a stage you have not reached reads as irrelevant — it is.

1. Unit economics — you do not actually know whether an incremental order makes money, so you cannot safely spend more to get one.
2. Conversion efficiency — you are paying for traffic that your site fails to convert, which makes every acquisition channel look worse than it is.
3. Demand capture — there is existing demand for what you sell and you are not capturing all of it. The cheapest growth there is, and the stage most brands skip past too early.
4. Demand creation — you have captured the available demand and further growth requires making people want the product who were not already looking for it. Structurally more expensive, and the point at which creative becomes the main lever.
5. Retention and lifetime value — acquisition cost has risen to the point where the only way to keep growing profitably is to earn more from each customer you already paid for.

Constraint one: you do not know if an order makes money

This is the most common blocker and the least discussed, because it is an accounting problem wearing a marketing problem's clothes. The symptom is a specific and recognizable stall: revenue grows, ROAS looks acceptable, and profit does not move — or moves backwards. The cause is almost always that the number governing ad spend decisions is revenue-based rather than margin-based.

Platform-reported ROAS is computed on revenue. Your business runs on contribution margin — revenue minus cost of goods, payment processing, shipping and fulfillment, returns, and, in the UK, VAT. Those deductions are not small. A brand with 45% gross margin, 8% fulfillment, 4% payment and returns is not making money at a 2× ROAS, and if it is reporting VAT-inclusive revenue to the platforms it is not even making money at the 2.4× it thinks it sees.

The fix is unglamorous: compute contribution margin per order for your actual product mix, derive the breakeven ROAS from it, and set targets above that with a deliberate margin of safety. Then govern the account on a blended number — MER — rather than platform-reported ROAS, because MER cannot be inflated by attribution. The full arithmetic is in ecommerce unit economics, and the reason to prefer the blended view is in MER vs ROAS.

Resolving this constraint frequently unlocks growth by itself, in a way that feels counterintuitive. Brands discover that some of their spend was unprofitable and some categories had far more headroom than the blended average suggested. Reallocating toward the second group grows revenue and profit at once, with no new channel and no new budget.

Constraint two: the site loses the traffic you paid for

The symptom here is that every acquisition channel underperforms simultaneously. When Google, Meta, and TikTok all look mediocre at the same time, the common factor is rarely three separate media-buying problems. It is the destination.

Conversion rate is a multiplier on everything upstream, which is why it is the highest-leverage constraint when it binds. Taking a site from 1.5% to 2.2% conversion is a 47% improvement in the effective cost of every customer, across every channel, without touching a bid. No media optimization available to you is that large.

Where to look, in order of how often it is the answer: mobile page speed on the templates that actually receive paid traffic, not your homepage; the checkout, measured as a funnel step by step so you can see where people leave; whether shipping cost appears early enough to avoid being a nasty surprise at the final step; and whether the landing page a paid visitor arrives on actually addresses the promise made in the ad they clicked. That last one is the most common and the least examined — a great deal of paid traffic is sent to a category page that answers a different question than the ad asked.

Constraint three: existing demand you are not capturing

This is the cheapest growth available and the stage brands leave too early, usually because capturing existing demand is less exciting than creating new demand. The test is simple: are there people actively searching for what you sell, with commercial intent, whom you are not reaching?

In practice this means non-brand search coverage — the categories, problems, and competitor comparisons your buyers type before they know your name — and Shopping coverage across your full catalog rather than the fraction of SKUs a neglected feed manages to serve. Feed quality is the single most under-invested lever in ecommerce paid media: titles, attributes, and images decide which auctions you are even eligible for, and a catalog with weak feed data is invisible in queries it would win. That work is laid out in the Google Shopping operator's guide, and the choice between Shopping and Performance Max for it in Shopping vs Performance Max.

One important distinction while you are here: brand search is not demand capture, it is demand harvesting. People typing your brand name were already coming. Counting those conversions as acquisition makes your paid performance look better than it is and, worse, teaches Smart Bidding to spend more on the cheapest conversions in the account — the ones you would have won for free. Separating brand from non-brand is the prerequisite for knowing whether any of this is working.

Constraint four: you have to create demand now

Eventually you capture the available demand, and growth requires reaching people who were not looking. This is structurally more expensive — you are interrupting rather than answering — and it is the point at which the nature of the work changes. Targeting stops being the lever, because on Meta and TikTok the algorithm has largely absorbed targeting. Creative becomes the lever.

Concretely, that means volume and variety of genuinely distinct concepts, not variations of one idea, tested against each other with a framework for deciding what wins and when to retire an asset. Brands that plateau on Meta are almost always creative-constrained rather than budget-constrained, and the diagnostic is straightforward: if your top-performing ad has been your top performer for four months, you are not testing enough to find its replacement. The system for this is in the creative testing system that scales and scaling Meta creative production.

Expect your efficiency metrics to get worse here, and plan for it rather than panicking. Demand creation converts at lower rates than demand capture by definition, so a blended target that was appropriate at constraint three will look like failure at constraint four. This is where a MER floor with an explicit payback window earns its keep: it lets you accept a worse immediate ROAS on prospecting because you have decided in advance how much future value you are licensed to spend against.

Constraint five: acquisition cost has caught up with you

At some point acquisition costs rise to where growing profitably on first orders alone stops working. Every competitor is bidding, the cheap demand is captured, and creative gains are incremental. The remaining lever is earning more from each customer you already acquired.

The mechanics are the familiar ones — email and SMS lifecycle flows, subscription where the product supports it, post-purchase cross-sell, and genuine product-line extension. What matters strategically is what higher lifetime value does to your acquisition ceiling: if repeat purchase behavior means a customer is worth 2.2× their first order within twelve months, you can profitably pay materially more to acquire them than a competitor optimizing on first-order ROAS. That is how brands out-bid competitors for the same customer without losing money, and it is why retention work is an acquisition strategy rather than a separate discipline.

The discipline required is to license this deliberately rather than hopefully. "Our LTV is high so we can afford a bad CAC" is the sentence that precedes a cash crisis when the payback window is longer than the business can fund. Decide the window explicitly, measure cohorts against it, and revisit quarterly with whoever owns the cash flow in the room.

How to find your binding constraint this week

1. Compute contribution margin per order for your real product mix and derive breakeven ROAS from it. If you cannot, you are at constraint one and nothing else matters yet.
2. Compare your paid conversion rate on mobile against desktop, and walk your own checkout on a phone. If mobile is dramatically worse, or you find friction you had forgotten about, constraint two is binding.
3. Check your non-brand impression share and what share of your catalog is actually serving in Shopping. Meaningful gaps in either mean available demand is going uncaptured, which is cheaper to fix than anything else on this list.
4. Look at how long your best-performing creative has held that position, and how many genuinely distinct concepts you tested last month. A stale winner and a low test count is constraint four.
5. Pull first-order versus twelve-month cohort value. If they are close, retention is your ceiling and constraint five is where the remaining growth lives.

Work them in order, and only move on when the current one genuinely stops binding. The reason this sequence matters is that solving a later constraint while an earlier one holds produces work that looks productive and changes nothing — scaling spend into a site that cannot convert, or testing creative when the real problem is that incremental orders lose money.

If you want the diagnosis done on your actual data rather than from a list, that is what our free audit is for: it reads your Google Ads or Meta account and computes these numbers on your account rather than on benchmarks. For what changes once you are spending past six figures a month, scaling ecommerce ads past $100k picks up where this leaves off, and our ecommerce PPC practice is how we run it.