We are the Revenue Growth Partners

Buys attention and answers for what it returns. In practice that means media strategy, campaign architecture, audience and bidding, creative testing, and the measurement sitting underneath all of it. The weak version of the job stops at the click and reports ROAS. The useful version stays accountable for what that traffic does next.

There is no universal number, because ROAS is meaningless without your margin. A 3x return is excellent at 70% gross margin and loss-making at 25%. Work backwards instead: what can you afford to pay for a customer and still make money? That figure is yours alone, and it is the only one worth optimising against.

Usually one of four things — auction competition has increased, creative has fatigued, the audience you were harvesting is exhausted, or tracking has degraded and the platform is now optimising on worse signal. The last is the most common and the least suspected. Diagnose before adding budget; more spend behind a broken signal raises cost faster.

Search captures demand that already exists. If your category has more potential buyers than current searchers, search alone caps your growth. Programmatic and DSP reach people before they start searching. Add them when search is efficient but volume-limited — never as a fix for search that is performing badly.

Global ecommerce sits around 2.5–3%, and clearing 3.2% puts a store in the top fifth. But the category spread is enormous: food and beverage runs 4.9–6.2% while luxury and jewellery sit at 0.8–1.5%. Your own trend against your own baseline tells you more than any benchmark will.

Partly real, partly measurement. Mobile typically converts at 1.8–2.8% against desktop's 3.2–3.9%, but some of that gap is people researching on a phone and buying on a laptop, which looks like mobile failure and is not. The real part is almost always form friction, costs appearing late, and layout that shifts while someone is typing.

More than most people assume. At a 2% conversion rate, detecting a 15% relative lift needs roughly 15,000–20,000 visitors per variant. At 5% or above, around 1,000 per variant can be enough. Below that threshold, testing is the wrong instrument and a diagnostic rebuild is the right one. We will tell you which you are.

A test runs a minimum of two weeks to cover two full business cycles, however fast it reaches significance, and anything past four to six weeks risks contamination from seasonality or campaign changes. Expect a prioritised roadmap in three weeks, and the first trustworthy reads a month or two after testing begins.

No, and anyone who does is either guessing or planning to cherry-pick the reporting. What we commit to is the process: tests designed properly, sized properly, and reported with their duration, sample size and confidence whichever way they went. A meaningful share of well-designed tests lose. We publish those too.

They never will, and the gap is not a fault. Platforms count conversions they claim credit for, inside their own attribution window, often including view-through. Analytics counts what it can still see after consent and cookie loss. Both are partial. The question is not which is right, but whether total revenue moved.

It measures what a channel actually caused, by holding a group back and comparing. It matters because last-click and platform-reported numbers credit channels for customers who would have bought anyway — brand search and retargeting are the usual offenders. It needs real volume to run cleanly, so below a certain scale we use simpler holdouts rather than claim precision we do not have.

One page a month, in plain language, written against the single number the engagement is held to. Wins and losses both, with test design, duration and sample size attached. No dashboard export, and no metric that cannot be traced back to revenue.

The evidence is genuinely mixed, and anyone quoting a single multiple is selling something. Published studies range from 1.3x organic to more than 20x, while one peer-reviewed study of 973 ecommerce sites found ChatGPT traffic converting around 13% worse than organic. The pattern: considered, research-heavy purchases benefit most, impulse categories barely at all.

Too early to move real budget, not too early to stop losing ground. The practical work — clean structure, answerable content, being citable rather than merely rankable — is largely the same work that helps conventional search, so it is rarely wasted. At most volumes we would not yet fund it out of paid media.

A performance marketing firm that is held to revenue rather than to media metrics — which means we do not hand the traffic over at the click. Buying the visit and converting it are the same problem measured at two points, and splitting them across two suppliers is how the loss in between ends up owned by nobody.

By engagement, not by the hour. The audit is three weeks at a fixed fee — the smallest commitment, and the one that tells you whether anything larger is worth it. Media spend is always separate and stated as such.

Roadmap in three weeks, testing the month after. Read speed depends on your traffic, and some sites do not have enough to test at all. No payback date before we have seen your data.

Alongside them, usually. We audit the accounts they run and the measurement underneath — the findings are yours either way. If the buying is sound, we take the site and they keep the media. Replacing an agency is rarely the right first move.

Read access to analytics, ad accounts and the site, one person who can answer questions about the business, and a decision-maker who can approve what gets shipped. Nothing else. If your developers are shipping the changes rather than us, we need their queue too, because that usually sets the real pace.

They are the operating record of the person you would work with. References on request before you sign anything. Every case study names what failed, and every report we send carries a losses section.