Skip to content
ThinkMedia

Ecommerce

Nine documented engagements. Every one of them started the same way — an account buying transactions instead of customers.

Get a free auditOur results
Documented ecommerce engagements
9
Average blended ROAS lift across these engagements
2.3x
Ad spend managed
$20M+
Years operating
3

Mean blended ROAS across these engagements moved from 1.79x to 4.17x. Every figure is taken from the case studies below, so the arithmetic can be checked against them.

Every ecommerce account we have taken over was optimised for the first purchase. That is why none of them were profitable at scale.

The mechanism

Why the first purchase is the wrong target

A platform optimises toward the event you give it. Tell Google or Meta that a purchase is the goal, and the algorithm will find you the cheapest available purchase. That is not the same as the most valuable customer, and over a few months the difference compounds into a real problem.

The cheapest purchase almost always comes from a discount. So the account learns to find discount buyers. Those buyers convert well, look excellent in the platform's reporting, and never come back. Your blended ROAS holds up for a quarter, then slides, and the usual response is a bigger discount.

We have watched this play out in every one of the nine ecommerce accounts on this page. In four, a discount-led creative angle was the biggest source of new customers and the worst source of repeat ones.

We cut it in each case, usually within six weeks. Each time, the short-term numbers got worse before they got better.

The second failure is broader and quieter. A prospecting audience defined by interest or demographic is not an audience of your buyers. "Women 25 to 45". "Luxury lifestyle". "Pet owners".

It is an audience of people the platform thinks resemble your category. It produces volume. It produces clicks. And it produces a customer base that does not resemble the one you already have.

The third is structural. Most catalogues have eight to twelve products that carry the business, and most accounts spend as though every SKU deserves equal budget. Two engagements here moved hero product share of revenue from 36% to 61%, and from 41% to 63%. Not by finding new demand. By stopping the spread.

Proof

The engagements.

Every case study

In order

What we change, and in what order

  1. We fix measurement before we touch a campaign.

    If the platform cannot see which purchases came from where, every decision after that is a guess. It also has to tell a first order from a fourth. So: server-side tracking, Conversions API with proper deduplication, and value-based signals rather than a flat purchase event. One account here had measurement broken by iOS changes for over a year.

  2. Then we change what the algorithm optimises toward.

    Value-based bidding on contribution rather than revenue, purchaser lookalikes rather than interest audiences, and where the business supports it, optimisation toward the second order rather than the first.

  3. Then the feed, if there is one.

    Google Shopping rewards specificity — material, dimensions, finish, compatibility. Two of the accounts here doubled Shopping's share of revenue on feed work alone, with no change in spend.

  4. Then creative, continuously.

    Targeting is largely automated now. Creative volume is the remaining lever, and it is the one most accounts under-resource. Weekly rather than monthly is usually the change that matters.

  5. Then retention, treated as a channel rather than an afterthought.

    Repeat purchase rate is the number that decides whether paid media is an investment or a treadmill. Across these nine accounts it moved from a mean of 15% to a mean of 28%.

Honest scoping

Who this is for, and who it isn't

This works for brands spending upwards of $5,000 a month with a product people buy more than once, or a catalogue with clear hero products, or both. It works best where there is enough order volume for the platforms' models to learn from — roughly 50 conversions a month per campaign as a floor.

It does not work on a genuine one-time purchase with no adjacent range. Everything above depends on a second order existing.

It does not work if the margin cannot survive a learning period. That is six weeks where performance is worse before the account stabilises.

And it does not work if you cannot know what a customer is worth over twelve months. Every decision here is made against that number.

If any of those apply, we will say so on the call rather than after the proposal.

Asked and answered

Common Questions

  • Usually the mix rather than the total. Most accounts at this stage are profitable on new customers and leaving the repeat revenue on the table. Or profitable overall, carrying two or three campaigns that lose money quietly. The first audit finding is almost always a campaign running at a loss for months, inside a blended average that looks fine.

  • Six to eight weeks before the account stabilises, and three to four months before the trend is clear enough to act on. Anyone promising faster is either restructuring nothing or counting a seasonal spike as a result.

  • No, but you should know what discounting costs you. Discount-acquired customers in the accounts on this page had materially lower repeat rates than full-price ones. That may be an acceptable trade for you. It should be a decision rather than a default.

  • Platform-reported ROAS counts the purchases the platform can see and takes credit for the ones it influenced. Blended ROAS — total revenue over total spend — is the honest version, and it is usually lower. Every figure on this page is blended.

  • Then that is the engagement, at least at first. We will not optimise against numbers we do not trust, and we will tell you if the first month is measurement work rather than campaign work.

  • No, but most of the accounts here are Shopify. The feed and tracking work differs by platform; the argument does not.

What ran

Delivered across these accounts.

Taken from the services-delivered table of each case study above, so this lists what actually ran rather than what could.

Get in touch

Tell us what you’re running.

Channels, monthly spend, and the part that isn't working. That is enough for us to tell you whether there is anything here worth a call — and we would rather say so now than after a proposal.

marks a required field.

Including https://

As much or as little as you like.

A person replies from a real address. We never sell or share your details.

Keep going

The rest of it.