模拟 · Minimal Organization · Digital platforms

Search engine

A search and advertising group as a walkable model: auctions in milliseconds, experiments over weeks, and a new ad campaign that waits for people.

输入密码后开始如有兴趣，请索取密码 →

All data are fictional; QUAVIRO is not modelled on any provider. The model can be fitted to your real data and then shows what is invisible today: review rates and waiting times by ad category and language, or ranking experiments in progress.

The ad campaign approval process at a search company, as a running model. QUAVIRO is a fictional search and advertising group living on two time scales: every search query is an auction in milliseconds, every ranking improvement an experiment over weeks. The main process follows a new campaign through automated review, advertiser verification and manual review to the learning phase of the bidding strategy.

The site has seven buildings: pavilions for search, ads and customers, the auction tower, two data centres and a substation. Stricter ad rules, or new rules before an election, send more ads to human reviewers – and the queue in front of manual review grows where everyone can see it. Alongside, someone searches for hiking boots, and a ranking change passes external raters and a weekly launch committee.

This page is for people who run or explain operations at a search or ad-tech company: ad review and policy, sales, product management, infrastructure, compliance and communications.

A new ad campaign starts

01
Advertiser creates the campaign: A hundred small advertisers set up campaigns.
02
Automatic review: The ads are checked automatically in seconds.
03
Advertiser is verified: Trust and safety checks who advertises.
04
Advertiser provides documents: The advertiser provides proof.
05
Manual review: Reviewers look at the ad.
06
Rejected.
07
Bidding strategy learning phase: The bidding strategy learns for one to two weeks.
08
Running.

How a search and advertising company can use the model

Internal communication

Search, ads, customers and infrastructure on one site: the collaboration view shows which teams touch a campaign, from sales to trust and safety.

适用对象Leadership, internal comms
Onboarding

The diagram ‘Year, week, day’ explains the company's two clocks to new teams: the auction in milliseconds next to experiments that run for weeks.

适用对象New teams in search and ads
Training

The disruption ‘Automatic review fails’ leaves every new ad waiting for two hours. Trainees watch the queue form and see how long manual review needs to clear it.

适用对象New ad reviewers, advertiser support
Explaining to customers

Advertisers and agencies see why their campaign is waiting – not for computing time, but for verification and manual review. The diagram ‘Where the time goes’ separates the two.

适用对象Advertiser support, large-customer sales, agencies
Strategic overview

Before new rules apply, play out how many reviewers they will need: set ‘Ad rules’ to strict, or run ‘New rules before an election’ for a month.

适用对象Leadership, policy team, compliance

What is shown – why – for whom

展示内容	为什么	适用对象
Queue in front of verification and manual review (Resources view)	为什么Shows where new campaigns get stuck.	适用对象Head of ad review, sales
Diagram ‘Where the time goes’ for one campaign	为什么Separates computing time from time spent waiting for people.	适用对象Ads product management
Diagram ‘What the work depends on’: data centres, substation, systems	为什么Makes the dependencies of search tangible.	适用对象Head of infrastructure, risk management
Diagram ‘Year, week, day’: millisecond auctions next to experiments over weeks	为什么Explains two time scales within one company.	适用对象Communications, onboarding
Control ‘Ad rules’ with the disruption ‘New rules before an election’	为什么Shows the staffing new rules require before they take effect.	适用对象Policy team, compliance
Collaboration view between search, ads and customer support	为什么Shows who works on a campaign.	适用对象Managers, new teams
Diagram ‘How a case flows’: automated review, verification, manual review	为什么Shows the share of each path an ad takes.	适用对象Ads quality assurance

Departments in the model

Board

Leads search, ads and infrastructure.
Ranking and index

Finds the best pages in milliseconds.
Search quality

Checks every change with raters and experiments.
Experiment platform

Thousands of tests at once on small shares.
Auction and ad selection

Runs an auction on every query.
Models and bidding

Learns which ad fits.
Ads policy

Reviews ads automatically and by hand.
Invalid traffic

Filters click fraud and credits.
Large customer sales

Serves the largest advertisers.
Advertiser support

Helps small customers with self-service accounts.
Measurement and billing

Clicks, conversions, invoices.
Trust and safety

Verifies advertisers and abuse.
Legal and privacy

Consent, supervision, proceedings.
Operations

Keeps latency within the error budget.
Data centres

Own halls, own chips, own network.

All processes in the model

A new ad campaign starts

01Advertiser creates the campaign: A hundred small advertisers set up campaigns.
02Automatic review: The ads are checked automatically in seconds.
03Advertiser is verified: Trust and safety checks who advertises.
04Advertiser provides documents: The advertiser provides proof.
05Manual review: Reviewers look at the ad.
06Rejected.
07Bidding strategy learning phase: The bidding strategy learns for one to two weeks.
08Running.

Someone searches for hiking boots

01Query arrives: A million queries for hiking boots arrive.
02Results and ads are fetched at the same time.
03Organic results: The index delivers candidates, ranking orders them.
04Select eligible ads: Eligible ads with budget are found.
05Candidates ready.
06Auction: bid times quality: The auction calculates: bid times expected quality.
07Deliver result page: The result page goes out.
08Collect clicks (hourly): Clicks are collected and processed hourly.
09Filter invalid clicks: Invalid clicks are filtered out and credited.
10Billed.

Ranking is improved

01Quality team proposes a change: The quality team wants to weight local results better.
02External raters compare results: External raters compare old and new results.
03Discarded.
04Live experiment on a small share: A live experiment runs on one percent of queries.
05Launch committee (weekly): The launch committee decides, once a week.
06Stopped.
07Staged rollout: Rollout in stages, with a way back.
08Live.

What you can change

Controls

Advertisers: Before Christmas many more campaigns start.
Ad reviewers: How many manual reviews run at once.
Ad rules: Stricter rules: more cases go to people (model assumption).
Compute for models: Larger models rank better but take time (model assumption).

Disruptions

Data centre fails: For two hours search runs on only half the halls.
Click fraud wave: For a week filtering takes twice as long.
New rules before an election: For a month more ads go to people, reviews take longer.
Automatic review fails: For two hours every new ad waits.

What the model gives a search and advertising group

01

Two time scales, one picture

The millisecond auction and the weeks-long experiment run side by side on the same site, instead of on two separate slides.
02

Waiting for people, made visible

The simulation shows queues in front of verification and manual review. Controls for reviewers and ad rules show what each lever does to waiting time.
03

Dependencies you can point at

Two data centres and a substation carry everything. If one data centre fails, search runs on half the halls for two hours; ‘What the work depends on’ shows the chain.
04

For workshops and big screens

Runs in the browser and on large projections, with 14 diagrams calculated live from the model.
05

With your own figures

Review rates by language, data-centre load and live experiments from your systems turn QUAVIRO into your own company.

Basis and limits

Fictional search and advertising group · model based on researched industry structures · no forecast for real companies

1 case = 100 campaigns, 1 million queries or 1 ranking change · 1 figure ≈ 50 employees

All volumes, durations, rates and capacities are model assumptions, not values of a real group.

The auction sources describe a common principle; QUAVIRO is not modelled on any provider.

Documented: auction by bid and quality, external raters by public guidelines, supervision of large platforms.

Public sources the model is based on:

Gesetz über digitale Märkte, VO (EU) 2022/1925

Your search and ads business as a model

Would you like to see your ad review, data centres or experiments like this? Get in touch by email or phone and we'll work out together which data the model needs. Access to the simulation is by password, available on request.

输入密码后开始

如有兴趣，请索取密码 →

cube4tec.de/spiele/minimal-organization/