Book a call

Paweł Skibiński, Growth Strategist, Paid Social

Eleven years in marketing. About $100M in ad budgets. I test what would have happened if you had switched the channel off.
Book a call with PawełLinkedIn
Paweł Skibiński, Growth Strategist, Paid Social

Snapshot

About $100M in budgets and hundreds of campaigns over eleven years. Meta is my main platform, with TikTok, AppLovin, Pinterest, Snapchat, Reddit and Nextdoor beside it. For measurement: Haus (geo-lift, multi-touch attribution, cMMM), Northbeam (multi-touch attribution) and Triple Whale. I run the campaigns, and I talk to CMOs about what every dollar really brings in.

11Years in marketing
~$100MIn ad budgets, all told
$0.5–2MMonthly Meta budgets on large accounts
iROASThe number I trust

What I do

I join paid social with incrementality measurement. I check whether the spend actually brought in extra revenue, because what the platform dashboard shows is not enough.

  1. Geo-holdout tests, multi-touch attribution and cMMM

    I design the experiments and the analyses.

  2. Meta on large accounts

    Budgets of $0.5–2M a month.

  3. Tracking and data audits

    CAPI, deduplication, identity parameters, attribution windows.

  4. Results into budget decisions

    What it means for you: Every report ends with a decision: scale, hold, cut or test again.

When I use which channel

  1. Meta

    When the main goal is acquiring new customers at scale.

  2. TikTok

    When the brand has a steady supply of native creative and agrees to measure results with a holdout. Platform ROAS there can be heavily inflated.

  3. AppLovin

    An extra channel to test when Meta starts to saturate.

  4. Pinterest

    For brands with a long path to purchase and a visual product.

  5. Snapchat, Reddit, Nextdoor

    When the audience is clearly there and the brand wants to diversify beyond Meta.

Who calls me

Industries

DTC footwear and apparelPet techFitness apps (subscriptions)Passport and visa servicesGrillsAir care

Mostly the US market, partly the EU and Poland.

Hot takes

  • Hot take 1

    Platform ROAS and real impact rank channels differently

    In my work, one client's best channel in the ads manager (TikTok, ROAS 6.01) was the worst in a holdout test: $0.23 of extra revenue per dollar spent.

  • Hot take 2

    Another attribution tool will not fix it

    It only adds one more number that does not agree with the others.

How I work

  1. 1

    First checks

    Event quality: duplicates from server-side and CAPI (I have seen an account show a ROAS of about 66x because of them), event parameters such as email, phone and fbc, deduplication. Then attribution windows and models. Then the dates of breaks in the data, such as tagging changes and account restructures; without them, week-on-week comparisons mean nothing. And the prospecting and retargeting split, where the biggest money usually sits.

  2. 2

    The first week

    A tracking and account-structure audit, a list of what to cut straight away, the break-even point and the metrics that really mean something to you.

  3. 3

    The first month

    A fixed structure and the first budget changes, a reporting rhythm, and a plan for incrementality tests: what we test, in what order, and which decision follows each result.

Tools

  1. Meta, TikTok, Reddit, Nextdoor and Pinterest ads managers

    Daily work on the accounts.

  2. Haus

    Geo-holdout tests: the only numbers that speak about cause, not correlation.

  3. Northbeam

    Multi-touch attribution, and new customers at creative level.

  4. Triple Whale

    One view of Shopify revenue, whatever the platform reports.

  5. Adnova

    Creative analytics: which ads, hooks and formats are working.

  6. Claude Code

    I build dashboards and analyses with it, such as dayparting and creative scorecards, that used to take days.

I built automated client dashboards: a weekly Meta creative scorecard that refreshes itself under a fixed link, a partnership ads dashboard joining Meta with Northbeam data, a TikTok Spark Ads dashboard per creator, and an hour-by-hour dayparting analysis. And an approach: incrementality testing as continuous work, not a once-a-year event. According to the incrementality testing platform, the team I worked in was in the top 5% of its users who test the most.

Lessons

  • The costliest mistake

    I approved a service provider for a client too quickly. Getting rid of them took a year, and they charged about $370k a year.

  • Favourite metric

    iROAS, ideally marginal iROAS. It says what the next dollar will do, not the average one.

  • Most overrated

    ROAS from the platforms.

  • Quick tip

    Before you cut or scale a channel, check its incremental impact.

Off the clock

  • I dance Argentine tango, and teach it.
  • I play chess.

FAQ

A rider in a red cape and bicorne hat on a rearing yellow plush horse, after David's Napoleon Crossing the Alps

What is a geo-holdout test?

You switch a channel off, or keep it running, in selected regions and compare them with the rest. The difference is the revenue the channel really adds.

Why not trust the ROAS in the ads manager?

Because it can rank channels in the opposite order to a holdout test. In my work, the best channel in the dashboard was the worst in the test.

What does a report from you start with?

A decision: scale, hold, cut or test again. The tables come after.

Which platforms do you run?

Meta first, plus TikTok, AppLovin, Pinterest, Snapchat, Reddit and Nextdoor.

Next step.

Bring the channel you are not sure you could switch off. Book a call with Paweł.

Book a call with Paweł