Case studies / five accounts

    The numbers, and how they happened.

    Real accounts, written up the way we would write them internally: what was actually wrong, what we changed, what it returned, and where the answer went against what the client expected.

    $1.24M
    Probiotics, one 14-day deal event
    +30%
    EU reseller, profit year on year
    $2.42
    Fire starters, per new customer
    +22%
    Skincare, Prime Day across 10 markets
    +131%
    Home textiles, profit in three months
    01 / US probiotics brand / 8 figures / Amazon US

    $1.24M in a single two-week deal event, with TACOS improving to 10.6%.

    The situation

    One flagship product carried roughly 80% of portfolio revenue. Ad efficiency had stalled at a level the team had learned to live with, and the plan for the next deal event was the same plan as the last one: a flat discount across the catalogue and a budget increase on the day.

    What we changed
    01
    Audited SB against SP on identical search terms
    On the same terms, Sponsored Brands was returning about twice the efficiency of Sponsored Products and was being funded like a secondary format. Budget moved accordingly.
    02
    Closed six-figure coverage gaps
    Terms with real category volume that the account had never bid on, found by working impression share rather than the search term report alone.
    03
    Designed the deal per SKU, not per catalogue
    Discount depth and bid multipliers set per SKU against its own margin and rank position, with AMC competitor audiences layered on the hero product for the event window.
    $1.24M
    revenue, 14-day event
    10.6%
    TACOS, while spend scaled
    2x
    SB efficiency vs SP, same terms
    02 / EU reseller / 6,000+ SKUs / 5 marketplaces

    +30% profit year on year, and 20 hours a week of manual work gone.

    The situation

    A portfolio too large to plan by hand. Long-tail demand was guesswork, bundle decisions were made on instinct, and the team was spending most of a working week keeping campaigns tidy across five marketplaces.

    What we changed

    A semi-automated campaign stack that creates, prunes and negates on rules the team can read, so structure stops decaying between reviews.

    On top of it, a demand model calibrated on the client's own sales history rather than category averages. Bundle potential is now scored before a unit is stocked, and Q4 planning starts from a forecast instead of last year plus a feeling.

    +30%
    profit, year on year
    13%
    blended ACOS
    20h
    manual work removed weekly
    1.6 to 2x
    Q4 baseline, forecast ahead
    03 / US outdoor CPG / fire starters / Amazon US

    $2.42 per new customer, and an honest verdict against running the deal again.

    First Best Deal on a thin-margin consumable. The client wanted to know whether it had worked. The dashboard said yes, so we ran the full post-mortem instead of answering from it.

    01
    Unit economics per promo type: deal, coupon, and both together, after fees and returns.
    02
    TACOS decomposed into the part driven by the discount and the part driven by the ad spend.
    03
    Rank tracked on 12 core keywords through the event and for the four weeks after it, to see what held.
    04
    LTV payback modelled on the new-to-brand buyers the event actually acquired.
    The verdict
    Coupons win for steady-state margin. Deals are a rank instrument, not a profit instrument, and should be run when you want position, not revenue. We told them not to repeat it as a margin play.
    $2.42
    cost per new-to-brand customer
    +131%
    sessions in deal week
    60 to 76%
    new-to-brand share
    04 / Global skincare / 10+ EU and UK marketplaces

    +22% Prime Day revenue with TACOS held near 10% across ten marketplaces.

    The situation

    Ten marketplaces, ten different competitive positions, one shared budget and one team. Peak events were being run as ten separate improvisations, and the marketplaces that mattered least often got the most attention because they were the loudest in the reporting.

    What we changed

    Budget allocated across marketplaces by expected marginal return rather than by historical share, with per-market ceilings set before the event instead of adjusted during it.

    On the day, one war room and the hourly tracker: sales, spend and an end-of-day forecast refreshed every 30 minutes, so reallocations happened while they still mattered rather than in the post-mortem.

    +22%
    Prime Day revenue
    ~10%
    TACOS across the event
    10+
    marketplaces, one plan
    30min
    decision cycle on the day
    05 / US home textiles / commodities niche

    +131% profit in three months, discount dependency down 88%.

    The situation

    A commodity category where every competitor was discounting, so the account had been discounting too, continuously, for long enough that the promotional price had become the real price. Revenue looked stable. Contribution margin did not.

    What we changed

    We withdrew the permanent discount in steps rather than at once, holding each step long enough to read conversion and rank separately, so a drop in one was never mistaken for the other.

    Ad spend was retargeted from defending price to defending position on the terms where the brand actually converted better than the category, and the listing was rebuilt around the two attributes that justified paying more.

    What surprised the client
    Units fell, and profit still nearly doubled. In a commodity niche the discount was buying volume that had never been worth having.
    +131%
    profit, three months
    -88%
    discount dependency
    3 mo
    to full effect

    Send us a storefront. We will tell you what we would change.

    Thirty minutes, no deck. If we are not the right fit we will say so on the call.

    Book a 30-minute auditTry the free tools first