Creaboost

neo · skills

The methods of the best experts on every channel, kept current.

A skill tells Neo what to look at, and how to judge it. Creaboost has written 43 of them, one per deliverable, each built on an identified expert source. Your team can create its own.

4,000 advertisers and agencies already use Creaboost.

why methods

A 2019 rule applied in 2026 gets expensive.

The model reasons very well. But trained on thirty years of content published online, it gives a 2019 tip as much weight as a 2026 one, and still recommends lookalikes and interest targeting. A general-purpose model makes the same mistake in law when it misses the latest case law. A skill gives it the rule that applies today, drawn from an identified expert source.

  • 2019 · meta

    Lookalikes and interest targeting.

    Audiences were built by hand, ad set by ad set, and creative followed.

  • 2026 · meta

    The creative is the targeting.

    Meta reads each ad and finds the people who will respond to it. Clearly distinct concepts open new audiences: ten variations of one idea reach the same people.

  • 2026 · google

    Google mostly captures demand that already exists.

    Brand and non-brand are always read apart, and on Shopping, the product feed is the strongest lever.

how they are written

Creaboost writes each skill on an identified expert source.

  • the source

    One expert source per skill.

    Each skill starts from an identified expert source for its channel or its job. A rule that does not come from it is labeled a working benchmark, to calibrate on each account.

  • the check

    Meta and Google, cross-checked.

    The Meta and Google skills were checked against the platforms’ reference documentation. No skill takes a platform’s own recommendations as a source.

  • the deliverable

    One skill writes one deliverable.

    The Meta review, the diagnosis of a drop, the paid media audit, the structure map: each deliverable has its skill, its format and its rules. A skill can call other skills.

At the top of every deliverable, one line gives the period, the skill that wrote it and the data sync date.

catalog

43 skills, sorted by job.

Here are the skills Neo applies, one line each. A skill carries its platform’s name when it depends on it, like meta_pilot.

reviews
  • business_review

    Did ad spend bring new customers at a cost the business can afford, and where should the next dollar go. Anchored on the store.

    Reallocate budget →

  • meta_pilot

    Meta: answer a piloting question, analyze an ad set or an ad, review any period.

    The Monday review →

  • google_ads_pilot

    Google Ads, with the same three modes. Brand and non-brand are always read apart.

  • tiktok_pilot

    TikTok, built on the Meta model. Never judged on its 7-day click ROAS alone.

  • microsoft_ads_pilot

    Microsoft Advertising, built on the Google model. Its small volumes are read over a month.

    as soon as the account is connected

  • ad_pilot

    Snapchat, Pinterest, LinkedIn, Reddit… First question: does the platform deserve the budget.

    as soon as the account is connected

diagnosis and audits
  • performance_diagnosis

    The cause of a drop, drilled from the business down to the trigger, then one targeted fix. Optionally, a weekly watch that stays silent when nothing moved.

    Explain a drop →

  • paid_media_audit

    What holds growth back across all your platforms: three pillars, one bottleneck, next steps ranked by impact.

    Audit an account →

brand memory
  • brand_kit_generation

    The Brand Kit: logo, colors and fonts, applied to every export.

  • brand_platform_generation

    What the brand stands for, and how it speaks.

  • products_customer_voice_generation

    Ranges, hero products, permanent offers, your customers’ objections and own words.

  • persona_generation

    Highly specific personas for paid social, drawn from your customers’ voice.

    Brand memory →

  • events_generation

    What happened, what is live, what is coming: sales, launches, stockouts.

  • economics_generation

    What an order really earns: break-evens, discounts, cohorts, targets.

account memory · structure

every week

  • meta_account_structure_mapping

    How your Meta accounts are built, and what is off in their structure.

  • google_ads_account_structure_mapping

    Feed, brand exclusions, asset groups, bidding: the Google Ads structure.

  • tiktok_account_structure_mapping

    Identities, expiring Spark Ads, creator share, overlap with TikTok Shop.

  • microsoft_ads_account_structure_mapping

    Imports from Google Ads, rejected products, LinkedIn targeting.

    as soon as the account is connected

  • ad_account_structure_mapping

    The structure of other platforms: Snapchat, Pinterest, LinkedIn, Reddit…

    as soon as the account is connected

account memory · metrics & funnel

every month

  • meta_metrics_funnel_mapping

    Meta baselines, three attribution readings, the funnel, the hit rate, seasonality.

  • google_ads_metrics_funnel_mapping

    Google baselines, brand and non-brand, impression share, Performance Max by channel.

  • tiktok_metrics_funnel_mapping

    The funnel from 2-second views to purchase, creators, how long videos last.

  • microsoft_ads_metrics_funnel_mapping

    The Google model applied to Microsoft, compared with Google.

    as soon as the account is connected

  • ad_metrics_funnel_mapping

    Other platforms. Under ~30 events, a number is too thin to read.

    as soon as the account is connected

  • shopify_metrics_funnel_mapping

    The store’s business view: MER, acquisition MER, nCAC, attribution gap per platform.

creative
  • creative_concept_generation

    Persona × angle × offer concepts, with their platforms, formats and 5 to 8 hooks.

    Find concepts →

  • ad_script_writing

    Nine-beat scripts from a concept, adapting a winning ad, reviewing a script line by line.

    Write a brief →

market
  • competitor_watch

    Your competitors’ ads: launches, long-runners, saturated angles, open white space.

    Track competitors →

measurement and budget
  • incrementality_testing

    Designs and reads a geo holdout, and gives the incrementality factor to apply to platform numbers.

    Measure incrementality →

  • channel_launch

    Are you ready for a new platform, which one, and its first 90 days through go / no-go gates.

    Launch a new channel →

  • peak_sales_planning

    One big sale period: offers, allowed CAC, budget phasing, how to read results mid-sale, calendar.

    Plan a peak sale →

beyond paid
system
  • review_orchestration

    Runs every review of a brand in one go: one per platform, then the business review, and a summary with every link.

    Scheduled tasks →

  • scheduling_management

    Creates, changes, pauses or resumes your tasks, and explains a run that failed.

  • unattended_run_safety

    The rules of every run with nobody at the screen: only the task’s own instructions count.

    Security and access →

  • deliverable_export

    Any deliverable as slides, spreadsheet, document or shareable page, in the brand’s look.

    Deliverables and exports →

  • analyze_report

    Rebuilds a report of the Analyze module as a deliverable, with the numbers shown on screen.

    Client reporting →

  • connector_setup_assistant

    Connects your work tools and checks them, without ever handling your credentials.

    Integrations →

cadences proposed by each skill · you choose your own

two memories per account

Every skill reads the account’s two maps first.

For each connected source, the brand’s memory keeps two maps up to date. A skill reads them before looking at a single number: it knows how the account is built, and what it usually delivers.

  • structure · every week

    How the account is built.

    Campaigns and ad sets with their role, audiences, naming convention, findings ranked by severity, a 90-day change log. Updated every week, after a restructure and when the account is connected.

  • metrics & funnel · every month

    What the account usually delivers.

    The account’s baselines (the median of the last three closed months), attribution, the funnel, the hit rate, seasonality. Written early in the month, on the closed month.

  • store · every month

    The business view on top.

    For Shopify, a metrics & funnel map that opens with MER, acquisition MER and nCAC next to the number of new customers.

When a map is missing or stale, Neo says what its answer cannot rely on, and offers to update it.

written rules

A skill writes down its rules, with their thresholds.

A skill sets its rules before looking at the numbers. Every decision Neo makes cites the rule it applies, and your team can push back on it.

situationthe ruleskill
A winning ad It spent at least twice the account’s average spend per ad within 30 days of launch, at a CPA at or below target.meta_metrics_funnel_mapping
A test that doesn’t take off It is judged once it has spent its planned budget, (target conversions × expected CPA) ÷ test length, per day. Below target, stop it and record what it taught.meta_pilot
A number on low volume Under ~30 purchases, it is too thin to read. Never decide on a single day, ad-level ROAS or a breakdown.meta_pilot
A concept wearing out Fatigue is read over 4 to 6 weeks: two weeks in a row, frequency up, reach flat, hook rate and link CTR down, CPM and CPA up. Starting point: frequency +20%, CTR or hook rate −15%.meta_pilot
A fatigued winner New hooks on the same body first. When the concept itself is worn out, a new concept, changing the angle first.meta_pilot
An ad set to scale +20% budget after five days on target: a budget increase does not reset learning. Stop when the extra spend only returns break-even ROAS (1 ÷ gross margin).meta_pilot
An ad that opens a sequence Never cut it on its last-click CPA when its ad set holds its target.meta_pilot
Two ROAS to compare Compare them within the same role: catalog, retargeting and Advantage+ without exclusions always show a better attributed ROAS.meta_pilot
Saturated retargeting Frequency above 7 over 30 days, or 8 over 7 days: cap it or cut it.meta_pilot
Creative volume About one new ad a month for every $1,000 of monthly spend.meta_pilot
Platform advice Never a source: neither Meta’s opportunity score nor Google’s optimization score.google_ads_pilot
Measuring incrementality A platform A/B test never measures it. Neo designs and reads a geo holdout.incrementality_testing
Launching a new platform First-order gross profit ÷ nCAC under 1: the launch waits.channel_launch
Before a big sale period Fund the weeks before: cutting spend shrinks the pool of people who know the brand.peak_sales_planning

starting thresholds · each skill calibrates them on your account

piloting

Neo steers by concept, ad set and creative surface.

Skills read results where decisions get made. The concept is the idea being tested: a persona, an angle, an offer, picked up from your naming conventions and tags. The ad set carries the budget and the optimization. The creative surface is the creative as people see it: the same video counts once, however many ads run it, and is read by format and placement.

Analyze creatives →
  • concept · new ocre collection

    26% of spend at a $59 CPA, with link CTR down 21%.

    Its ad sets only have bottom-of-funnel creatives, no opener: the new creatives will fill that step.

  • ad set · ocre drop 2

    Its CPA goes from $70 to $60, and it still loses ~$2,200 of budget in the week.

    The campaign budget moved to four duplicates created in late September, at a worse CPA: those are the ones to cut.

  • surface · founder video

    Hook rate 33% (median 22%), hold rate 3.0% (median 8.7%).

    The video hooks people, then loses them after 3 seconds: the weak link is the bridge.

  • decision · check on oct 20

    “Keep the hook, rework the bridge after 3 seconds, with ‘3 reasons’ on screen and a counter.”

    One decision per level, with its rule and check date, for your team to approve.

three modes

A skill serves your question, onboarding and your tasks.

  • your question

    In the chat.

    “Should I cut this ad set?” Neo picks the matching skill and answers in order: the decision, the reason, the exact steps, what to watch.

  • onboarding

    At setup.

    The memory skills fill each section of the brand, one after the other, from your accounts and the files you upload.

  • a scheduled task

    On schedule.

    The task runs the skill on the period it names, asks no questions, and sends you the summary. Neo suggests the task; it is created only once you say yes.

your skills

Your team creates its own skills, in its own words.

An agency has its own way of running a review, an advertiser has its house rules. Write them once: Neo then applies them to every deliverable they cover.

  1. 01

    Start fresh or from a skill

    In the chat: “Create a review skill from meta_pilot that looks at margin first, then CPA, and never at ROAS alone.”

  2. 02

    Review

    Neo writes it in the same format (goal, inputs, steps, rules, deliverable), and you read it before it goes live.

  3. 03

    For one brand

    A skill belongs to a brand, like its memory, its templates and its custom metrics.

  4. 04

    Evolve it

    To change a skill, create a new one from it. Any skill, from Creaboost or from your team, can be turned off.

Scheduled tasks →

questions

What people ask us about skills.

Can I see the method behind an answer?

Yes. At the top of every deliverable, one line gives the period, the skill that wrote it and the data sync date. Every decision cites the rule it applies and the numbers it rests on.

Can I edit a skill?

You create a new skill, from scratch or from an existing one, Creaboost’s or your team’s, then turn off the one you no longer want. Creaboost’s skills stay maintained by Creaboost.

Where do the rules in the skills come from?

Creaboost writes each skill on an identified expert source. A rule that does not come from it is labeled a working benchmark. The Meta and Google skills were cross-checked against the platforms’ reference documentation.

Are skills shared across clients?

Creaboost’s skills are the same for every brand. The ones your team creates belong to one of your brands and stay in your workspace, like its memory and its data.

Security and access →
Can a skill change my campaigns?

None can. Skills read your accounts and recommend, with precise steps your team applies in Ads Manager or Google Ads.

What does Neo do when no skill fits?

It answers with your data and the brand’s memory, and tells you no skill applies. If the request keeps coming back, create a skill for it.

See a Creaboost skill applied to your accounts.

Book a demo: Neo runs the week’s review of your Meta account, step by step, right in front of you.