Blended ROAS without lying to yourself

Platform ROAS is not a lie, exactly. It is an answer to a narrow question — “how much revenue did this platform’s pixel see after someone interacted with an ad?” — using that platform’s attribution window and its own definition of a conversion. The problem starts when we treat four of those narrow answers as one wide one.

Why the platforms add up to more than 100%

A customer sees a Meta ad on Monday, clicks a Google search ad on Wednesday and buys on Thursday. Meta claims the sale under its 7-day click / 1-day view window. Google claims it under its 30-day click window. Your Shopify store recorded one order. Your dashboards now show two.

Notice that the two lines do not even move together. Platform ROAS peaked in June when the team ramped retargeting — an audience that would mostly have bought anyway. Blended ROAS improved later, after budget moved to prospecting.

Start from net revenue, not pixel revenue

The numerator of an honest ROAS comes from your store or billing system, not from an ad platform. And it should be net: after refunds, chargebacks and discounts, and ideally excluding shipping and tax. A 12% refund rate is invisible to every pixel ever written.

1234567891011121314select
    date_trunc('week', o.created_at)                         as week,
    sum(o.subtotal - o.discounts - coalesce(r.refunded, 0))  as net_revenue,
    sum(s.spend)                                             as ad_spend,
    round(
        sum(o.subtotal - o.discounts - coalesce(r.refunded, 0))
        / nullif(sum(s.spend), 0), 2
    )                                                        as blended_roas
from shopify.orders o
left join shopify.refunds_by_order r on r.order_id = o.id
join ads.daily_spend s on s.date = o.created_at::date
where o.test is false
group by 1
order by 1;

The three numbers your CFO actually wants

Finance does not care which platform gets credit. They care whether the next euro of spend will come back, and how fast. Give them these three, every week, from the same model:

What about incrementality?

Blended metrics tell you whether the whole machine works, not which part to cut. For that you need experiments: geo holdouts, conversion lift studies, or simply pausing a channel in one region for two weeks. Kimo’s comparison view lets you overlay a holdout region against the rest and read the gap directly.

Setting it up in Kimo

Connect your ad accounts for spend and your store for revenue. Kimo Marketing ships a paid_media model with the three metrics above pre-defined, including the provisional refund adjustment. Adjust the margin assumption, set an alert on CAC payback crossing your threshold, and share the dashboard with finance. It is the one number you will both agree on.