Food brand · Growth

Come Verde

I work with a food brand on how it speaks, where it advertises, how it sells on Amazon and Mercado Libre, and which part of that work AI can take. The goal is to plan all of it together instead of each area on its own.

Come Verde home page
Real screenshot of the project
Work for a brand
type of project
Meta Ads · Google Ads
main tools
5
real screenshots of the project

01 · The challenge

What was going on

Context

Come Verde sells healthy snacks in physical stores and on marketplaces. The logic there is different from running your own store: what matters is that people remember you at the supermarket and that the product leaves the shelf, not just that they click an ad.

The problem

Get brand, advertising, and measurement to agree on three things: that people remember you, that the product sits where they look for it, and when in the day it gets eaten. Without reducing all of it to how many sales yesterday’s ad closed.

My part

I handle the ecommerce and media side, coordinate with their brand and commercial teams, and build the systems used to plan, launch, and learn from every campaign.

02 · How I solved it

How I decided to solve it

  1. 01

    Separate three things that usually get mixed: building the brand, pushing a specific promotion, and selling on marketplaces.

  2. 02

    Advertise where the product is actually on the shelf, and in the season when people eat it.

  3. 03

    Use AI to speed up the repetitive parts, such as briefs, ad variants and summaries, inside a strategy the team defines.

03 · What I made

What got built

  • The handbook for how digital advertising gets done, so it does not depend on who is around that month.
  • Guides by campaign type, which channel serves which purpose, and who the customer is.
  • A dashboard showing what is working, what is being tested, and what we learned.

04 · Charts

The numbers as charts

CV-MIX

Digital budget layers

Way of Work investment framework — methodological shares, not amounts.

  • Brand building75%

    Mental availability; long flights; reach & frequency.

  • Tactical activations20%

    Promo, launch, geotarget to chains with stock.

  • Performance / marketplaces5%

    Amazon/ML spillover; never at brand’s expense.

Documented ranges ~70–80 / 15–25 / 5–10; chart shows range midpoints.

CV-ROL

Brand → AI → Media flow

Claude executes against guardrails; human validation is a checkpoint, not an afterthought.

  • Brand and commercial inputs20%

    Product, target, occasion, chain, objective.

  • Brand strategy layer (Claude)25%

    Brief, copy, tone, hooks with guardrails.

  • Human validation20%

    Internal review plus leadership approval.

  • Creative production15%

    On-brand assets for each format.

  • Media amplification20%

    Buying, targeting, delivery reporting.

Relative operating-flow weights; not headcount or budget.

CV-MED

Measurement across three horizons

Useful substitutes when in-store ROAS is not visible in ad platforms.

  • Early signals35%

    Reach, frequency, VTR, engagement.

  • Intent35%

    Brand and brand+retailer search lift.

  • Velocity30%

    Sell-out / test & control with Commercial.

Reading framework; no published time series.

06 · How it turned out

What changed in the end

  • A way of working built for shelf products, not copied from an online store.

  • It became clear who decides what across brand, media, and commercial.

What I would not compromise on

  • First be remembered; fine-tune the ad after.
  • Do not advertise where the product is not.
  • AI speeds up the work; it does not decide the strategy.

07 · Technology

What it is made with

  • Meta AdsMeta Ads
  • Google AdsGoogle Ads
  • YouTubeYouTube
  • TikTokTikTok
  • Amazon Ads
  • Mercado Libre Ads
  • OdooOdoo
  • ClaudeClaude
  • Google DriveGoogle Drive
  • Looker StudioLooker Studio

Want something like this for your business?

If you sell in physical stores and online and each channel goes its own way, they can be planned together.

Write to me