Set up Google Ads properly and keep it running well

What must be in place before the first campaign?

A company needs its own administrative access, documented manager account links, billing that is properly set up and controlled access to the website, Tag Manager, Analytics and CRM. Avoid shared logins. For every role, record what it is needed for, which permission it carries, who is responsible and how access is withdrawn.

Google grants ownership privileges over a client account differently depending on whether the account was newly created or linked to a manager account later. Check who originally created the account and which manager account holds which ownership privileges over the client account. Those rights remain access rights: they do not transfer the client’s ownership of its data. Access to an account does not yet make it a properly handed-over company account.

What measurement readiness is required?

Define the business outcome to be measured before launch. For lead generation this is often a genuine enquiry, correctly recorded in the agreed system, not a click on “Send”. Agree the primary and secondary actions, the counting rule, consent and test cases, and set out how failed submissions are handled. Decide which later CRM stages are reported back and who is accountable for their quality.

If the signal is missing or counted twice, a campaign may still generate traffic technically, but it cannot be steered reliably towards actual business results. The launch decision must make that state visible: ready, conditionally ready or blocked. “We will fix it later” is not a measurement strategy.

How should the account structure be planned?

A campaign needs a clear role: offer, market, the campaign’s budget and the person accountable for it, conversion goal and landing page. Closely related search intents belong together in one ad group; not every wording variant needs its own ad group. Consistent naming makes operations easier but does not replace a sound structure.

Create a map with campaign, audience, search intent, landing page, conversion goal, budget source, exclusions and the person responsible. Separate brand, generic demand and experiments wherever the budget or decision logic differs. Shared budgets, automated bidding and account-wide settings must remain visible in the operating documentation.

What does implementation mean?

Implementation is a verifiable chain:

1. discovery and current-state capture;
2. access and measurement readiness;
3. goal and structure design;
4. configuration in a controlled environment;
5. test failure cases and unexpected inputs, then sign off the setup;
6. limited launch with an observation plan;
7. transition into recurring operations;
8. documented handover.

For each step, record what is needed, what is delivered, who is responsible, who approves and how a change can be reversed. A partner who only “sets things up” but delivers no acceptance, no handling of errors and no documentation leaves the team with an account but no operating model.

How is an existing account taken over?

A takeover is not a greenfield setup. Secure roles, billing, history, conversion actions, manager links, scripts, rules, data connections, audiences and open experiments. Document active budgets and automated strategies before any change.

Then decide what is kept, repaired, paused or rebuilt. A complete restructure can affect comparability with earlier periods and learned bidding behaviour, even though the change history remains. Taking over an account without correcting recognised faults carries those faults forward. Use a findings list with impact, risk, dependency and a rollback path instead of a blanket “best practice” migration.

What belongs in the ongoing routine?

A daily check does not mean changing every keyword every day. A routine can fetch data, document the current state of the data, trigger activated workflows and surface exceptions. When an analysis cannot run, the reason is recorded: missing or stale data, an excluded account, a campaign without usable data, a timeout or an unsupported bidding mode. A daily run does not guarantee fresh values for every individual search term.

A possible review rhythm, adapted to the available data and the agreed scope, separates:

• ongoing alerts for access, outages, spend or broken measurement;
• daily account review for urgent deviations;
• weekly search term, budget, bidding and tracking decisions;
• monthly structure, goal, pipeline and cost reviews;
• event-driven reviews after a launch, migration or larger change.

Not every task belongs in the same daily loop. Which data is required, which period it covers and how large the minimum sample must be depend on the decision being made.

How is Google Ads performance improved systematically?

Performance does not improve through an unsorted list of “optimisations”. Operations need a repeatable cycle: record the baseline, name the bottleneck, form a hypothesis, assess impact and risk, propose a bounded change, decide, implement separately and compare the result with the baseline. If tracking data or a reliable basis for comparison is missing, repair the measurement first.

Prioritisation follows the business goal. Search terms and negative keywords address waste and relevance. Spending is adjusted against the agreed budget, and funds move only where the available results justify it. Keyword, ad and landing page checks look for growth or quality problems. None of these procedures guarantees better results; each must show its own evidence, its limits and the reason when no proposal is made.

How can Google Ads costs be optimised sensibly?

Cost optimisation does not mean cutting the advertising budget across the board. Separate at least media budget, management cost, internal time and third-party tools. Inside the account, four levers are assessed separately: exclude irrelevant search queries, limit bids or goals, move the budget closer to the plan, and shift budget towards demand whose results are already demonstrated. A saving only makes sense when the consequences for valuable leads or conversions are checked.

Every cost decision needs an amount, a period, the account or campaign level, a clear measurement basis and an assessment of the risk the change carries. How the management fee and the media budget are composed is shown in the cost comparison for Google Ads management.

How do operations stay current with Google Ads developments and features?

New campaign types, bidding features, measurement options and AI features are triggers for a review, not automatic recommendations. Keep a small feature register with publication date, affected accounts, prerequisites, possible benefit, data requirements, loss of control, test design, the person responsible, rollback and review date. First check whether the feature even fits the market, goal, conversion volume and existing governance.

A bounded test with a clear comparison baseline follows. A new feature becomes standard only once the team can trace its effect and its operational consequences. That keeps the account adaptable without chasing every Google Ads development unchecked. The team reviews results at agreed intervals; interim findings do not replace formal sign-off.

How do alerts and quality assurance work?

An alert is a prompt to investigate, not a diagnosis. Every rule needs a metric, a threshold, a comparison basis, data freshness, a severity level, the person responsible, a rule for temporarily muting an expected alert and documented action steps. Examples include missing current data, a broken conversion signal, unexpected spend, disapproved ads or a change outside the documented process.

Good QA also checks where apparent confidence is misleading. A green dashboard can rest on a missing denominator. An alert can be expected after a planned budget change. Store the context and, after the correction, confirm whether the signal has actually returned.

What role do AI and automation play?

Automation can collect, classify, prioritise and prepare proposals. Many decisions are deterministic; optional AI must not invent missing evidence. Smart Bidding remains one Google layer, external analysis another, and human business responsibility a third. Name the layer in every decision.

Testing a new feature within limits

Suppose your team wants to extend keyword matching: first clarify which additional search intent should be reached, whether conversion measurement is reliable enough for the test, and which irrelevant searches should remain excluded. Availability and specific settings are checked in the account concerned.

Then record the starting position, the test scope and the success criterion. Observe qualified leads, cost per qualified lead, demand coverage and conversion delay. Do not change matching, landing page and success measurement at the same time if you later want to trace the effect. A higher optimisation score or more clicks alone does not justify scaling up.

maitiq connects such tests with ongoing account analysis: review search terms, examine the spend on irrelevant or poorly performing searches, identify relevant long-tail queries and study campaigns or ad groups that already produce useful results for growth potential. Your team or the named contact assesses the proposals in the business context, evaluates new features and develops existing campaigns using the evidence, instead of merely switching on new settings.

How are decisions and handover documented?

Separate six states: data basis, analysis, proposal, decision, implementation, verification. A decision names the actor, the time, the scope and the reason. Where an implementation requires its own authorisation, it is issued separately. A verification record states what actually happened and which source proves it.

Which documents and training the handover includes is agreed at the start. They can include the structure overview, role list, measurement inventory, open risks, decision log, documented action steps and known limits. Agreed training works with realistic, controlled cases and also shows when it is better to change nothing.