There is a version of this article that opens with the media strategy, and it would be the wrong one. So here is the sequence in the correct order.
Gutta på Haugen, a specialty food shop in Oslo trading since 1994, had a website that was a form. A customer described what they wanted, an employee phoned them back, took the order verbally, arranged payment offline and shipped the groceries.
You cannot buy traffic for that. Not usefully. Every krone of paid media landing on that site arrived at a page whose best possible outcome was a callback request, at a cost per manually processed order the published benchmarks put at $30–80 — the arithmetic we work through in your order form is costing you $30–80 an order.
The store had to exist before the media could mean anything. That is the whole point of this piece.
The Numbers
From the results page, all figures from Kilowott project reporting:
| Metric | Figure |
|---|---|
| Ecommerce revenue | 6.8M+ NOK |
| Blended ROAS, Meta + Google | 4.24x |
| Year-over-year revenue growth | 3.1x |
| Conversions | 2,967 |
| Conversions YoY, at stable CPA | +101% |
| Click-through rate growth | +50% |
| Meta ROAS range, by campaign type | 5.36 – 8.44 |
Two things to say about that table before anything else, because they are the difference between a case study and a claim.
These are Kilowott project reporting figures, not platform exports. We do not have direct access to the client’s analytics or advertising accounts, so none of them carries a GA4 or Ads Manager label. We say so on the results page itself.
There is no pre-rebuild baseline. The shop was making sales through the form-and-phone process, but that volume was never quantified. So the growth figures are real growth from a real starting point — they are simply not growth from a number we can show you. Do not read “3.1x” as “from zero,” and do not let anyone else read it that way either. The cost of that missing baseline is the subject of measure the process you are about to delete, and it is our own mistake, not a client failing.
Why the Blended Figure Is the Honest One
The single most abused number in performance marketing is a channel ROAS quoted in isolation.
Meta ROAS on this account ranged 5.36 to 8.44 across Advantage+, catalog sales and web campaigns. Quote the top of that range on a slide and you have an 8.44x agency. Quote the blended figure across Meta and Google together and you have 4.24x, which is the number that corresponds to money the business actually received.
The gap between those two is not dishonesty in the platforms. It is attribution overlap, prospecting versus retargeting mix, and the simple fact that the cheapest conversions are the ones that were most likely to happen anyway. A brand-search campaign will always post a spectacular ROAS. It is also the campaign doing the least incremental work.
Blended ROAS is a worse number and a better metric. If a media report only shows you per-channel figures, ask for the blended one, and notice how long it takes to arrive.
The honest caveat on our own figures: a ROAS without a spend base cannot be fully audited by a reader. 4.24x on 200k NOK and 4.24x on 2M NOK are very different achievements. That figure is not currently published on the results page, and we would rather flag its absence than let the ratio imply a scale it has not earned.
The Structure
The programme covered full-funnel demand across both platforms, with seasonal scaling around the three peaks that matter for a Norwegian delicatessen — Christmas, Easter and Summer.
Meta — awareness and discovery. Advantage+ campaigns and catalog sales carried the top of the funnel. For a specialty retailer this is doing something specific: most of the addressable market does not know the shop ships nationally. That is not a demand-capture problem, it is a demand-creation problem, and search cannot solve it because nobody is querying for a shop they have not heard of.
Performance Max — mid-funnel. Category-level intent, product discovery, and the long tail of queries too specific to structure manually.
Branded and non-brand search. Branded search protects the demand the awareness work created. Non-brand competes for category intent. Keeping them separated in reporting is what stops the first from flattering the second.
Remarketing — high intent. People who reached a product page, a recipe or a cart and did not finish.
Feed and creative discipline. Campaign structure, creative rotation and feed optimisation aligned to the shop’s own commercial priorities rather than to whatever the platform’s automation would optimise toward by default. On a catalogue with single items, variable products, catering bundles and subscriptions in it, feed quality is not housekeeping — it is the input to every automated campaign type above.
One honest limitation on this section: the reasoning behind the paid-media choices was documented as strategy, not captured as a designer interview the way the build decisions were. So we can describe the structure with confidence and will not invent a rationale for each choice after the fact. The build-side reasoning, which was captured in detail, is in the whitepaper.
What the Store Contributed to the Media Result
This is the part most performance reports leave out, and it is where the two halves of the project meet.
A product page that could answer questions. Producer, variants, delivery estimate, ratings, and full product, nutrition and allergen detail — replacing what staff used to answer live on the phone. Paid traffic converts against page content, not against ad copy.
Recipe pages wired to the basket. Existing content with existing traffic, made buyable. Benchmarks put shoppable recipes at roughly 30% higher basket size and about three extra items — see the recipe page is a storefront. Higher average order value raises ROAS without touching the media plan at all.
Subscribe & Save on monthly staples. Revenue that recurs without reacquisition cost is the cleanest ROAS improvement available to any account, and it comes from the build rather than the bidding: solve replenishment before you chase new customers.
A checkout that finished the job, with handling fee and VAT itemised, delivery notes, and the subscription split shown before payment.
Every one of those is a build decision that shows up in a media report. Which is why “our ROAS is disappointing” is so often a store diagnosis wearing a media costume.
The Market Context Worth Knowing
Norwegian online grocery reached only 5–10% of its category in 2025, against 10–15% for e-commerce overall. Strong physical retail means the categories where in-person expertise matters most have digitised slowest. That same category is now the fastest-growing segment of Norwegian e-commerce at roughly 14.8% CAGR through 2030. (Market data, not this project’s results.)
For a specialty retailer, that combination is the opportunity in one line: a national audience increasingly willing to buy this way, and few specialty competitors who have built for it properly. The window is the gap between those two facts, and it closes.
The Uncomfortable Order of Operations
If your paid media is underperforming, the instinct is to change the media. New creative, new structure, new agency, tighter audiences.
Sometimes that is right. But run the cheaper check first: take your top-spending campaign, click the landing page, and try to complete a purchase like a customer who has never heard of you. Count the questions you cannot answer from the page. Count the steps between intent and payment.
If that exercise is uncomfortable, no amount of bidding sophistication will fix the number — because the store, not the account, sets the ceiling. What the eight jobs of a counter business have to do with that ceiling is the argument in the counter problem, and how the storefront itself was rebuilt is documented screen by screen in the design case study.
