PPC judged on profit per order, not on ROAS screenshots.

A fixed monthly PPC management fee scoped to the work, quoted before anything starts and kept separate from your media spend. We avoid percentage-of-spend pricing for stores because it pays the agency to spend more rather than to spend well, and it makes the honest recommendation — cut this campaign — expensive to give. Scope follows catalog size, the number of campaigns and how much catalog work is needed. After a first review you get a plan and a target before you commit to anything.

The one that clears your contribution margin, which is why the question is usually asked the wrong way round. A 4x return on a 20% margin product loses money once shipping, payment fees and returns are counted, while 2.5x on a high-margin item with strong repeat purchase is excellent. For calibration, the ecommerce averages across 500+ accounts are 3.4x on Search, 5.1x on Shopping and 4.8x on Performance Max — useful context, never a target on its own.

Usually both, with clear jobs. Performance Max is efficient across a catalog and, at a $43.16 average cost per acquisition against $50.71 on Search, hard to argue with on cost — but it hides where the money went and happily re-buys brand demand. We keep brand in its own campaign, run Performance Max on the catalog with a feed built to steer it, and keep Standard Shopping where a product line needs direct control. Then we read the placement and search term data monthly instead of trusting the black box.

Three usual causes, and most accounts have all three. Conversions are counted by more than one platform, so the same order is claimed twice; brand searches sit inside an acquisition campaign, so demand you already owned is billed as new; and refunds, cancellations and shipping subsidies never make it back into the number. We rebuild tracking server-side, deduplicate, subtract returns, and split brand out, which usually lowers the reported figure and makes every decision after it trustworthy.

It is the campaign. Shopping and Performance Max match your data to a query, so a product whose title starts with an internal code is invisible for the phrase people actually type, and a missing size or GTIN can drop it from the auction altogether. Most accounts we take over gain more from a week of feed work than from a month of bid tuning. We rewrite titles around search behaviour, fix attributes and availability, and document the rules so new products inherit them.

By deciding what a first order is worth and refusing to guess. We measure new-customer contribution separately from repeat, look at second-order rate and time to repeat, and set acquisition targets that allow a thinner first sale where the data supports it and a strict one where it does not. On the spend side we protect margin with product groups, so the clearance line and the hero product are never bidding under the same target, and we cut the long tail that never repays its click.

They overlap, and the overlap should be deliberate. Paid gives you same week evidence on which queries and product sets convert, which sharpens the ecommerce SEO roadmap; search then takes over terms where click costs are punishing. With organic search at 35-55% of revenue-generating sessions in established stores, the goal is a blended cost per order across both rather than two teams claiming the same sale.

Half the outcome, and the half most accounts ignore. If the ad promises free returns and the landing page mentions them in the footer, the media was wasted regardless of how well it was bought. We match pages to ads, surface shipping cost and delivery time before the cart, and test offers rather than headlines. With documented cart abandonment averaging 70.22%, post-click work is usually the cheapest performance available.

Enough for the platform to learn on, which in practice means a budget that can produce a steady flow of conversions each week rather than a trickle. Below that, campaigns spend their life in learning and every read is noise. If the budget is genuinely small we narrow the ambition instead of spreading it: brand protection plus one product group with real demand, run properly, beats four half-funded campaigns. We will tell you plainly if paid is the wrong first investment for your store.

By margin and by demand, in that order, then by what the data says once it is running. We start from a contribution view of the catalog rather than a best-seller list, because the products that sell most are often the ones that earn least. Products with genuine search demand and healthy margin get their own campaign and budget; the long tail rides in a catalog-wide campaign; loss leaders are advertised only where the repeat data justifies it. That view is revisited every month, not set once.

Ask any PPC agency what they will measure and listen for whether the answer is platform ROAS or contribution per order. Ask who touches the account day to day and how much of the work is automated. Ask to see a weekly change log from a current client, redacted. And ask what they would switch off in your account tomorrow — a good answer names something specific and explains the reasoning. Anyone promising a return figure before seeing the data is selling a number, not a service.

Automated Google Ads bidding, informed properly, with human control over what it is optimising toward. Smart bidding is very good at picking the auctions once it is fed clean conversions, correct values and sensible groupings, and very bad at all of it when the inputs are wrong — which is why our work concentrates on the inputs. We keep manual control over structure, budgets, targets and negatives, and we review the search terms and keywords weekly because that is where automated spend goes strange first.

A weekly PPC log of what changed and why, and a monthly written read on orders, revenue net of refunds, contribution and new-customer share, split by campaign and product group. No dashboard-only reporting, and no report that lives behind a login only we can access. If a month was poor you will read why in plain language, along with what we are changing. Clients tell us the change log is the part they keep, because it is the only record of what actually caused a movement.

Frequently, and it is often the best value. A common split is that we own the catalog, the structure, measurement and the weekly optimisation while your team owns creative production and promotions, or the reverse where you have a strong buyer but no analyst. We document naming, budget rules and the feed checklist so nothing depends on us being in the room, and we are happy to train the team to take the account back entirely.

Catalog and tracking fixes tend to show inside two to three weeks because they change what the auction sees. Structural rebuilds need a learning period, so expect noisier numbers for a fortnight before they settle, and we sequence changes so the whole account is never learning at once. A fair judgement point is the end of the second month with the full picture at ninety days. We report from week one, including the weeks where the honest answer is that it is still settling.