Bidding strategy

Choose by data, not by taste. As a rule of thumb:

• No conversion history: use manual CPC, or maximise clicks with a maximum CPC cap, for the first two to four weeks. The goal is to learn which queries show up, not to win.
• Some history, a handful of conversions a month: use maximise conversions with no target.
• Around 30 or more conversions in the last 30 days: move to target CPA.
• Conversions with values attached: use maximise conversion value, then target ROAS once the values are reliable.

These thresholds are guides, not rules Google enforces. They exist because the algorithm needs enough outcomes to find patterns.

The algorithm optimises for what you tell it to count. If every newsletter signup counts, you will get newsletter signups. Set one primary conversion for a qualified lead, such as a demo booked by a company that fits your profile. Make softer actions secondary so they report but do not steer bidding.

The full measurement set-up is in If you cannot measure the right outcome, you are flying blind.

Start from what a customer is worth, then work down:

1. Take the gross profit per customer over its lifetime. Say 6,000.
2. Divide by three, so each customer has to return three times what you spend to win it. That gives 2,000.
3. Divide by the share of qualified leads that become customers. If one in five closes, 2,000 divided by five gives 400.

So 400 is the most you can defend paying for one qualified lead. Use your own numbers. If your close rate is a guess, say so and revisit the target after ten closed deals.

A target far below your current cost per conversion starves the campaign of traffic. Look at the actual cost per conversion over the last 30 days. Set the target at or just above that number. Then lower it by 10 to 15 percent every two weeks until you reach the figure you can defend.

The algorithm needs several conversions a week to learn. If your target is high, a small daily budget produces one conversion a month and nothing to learn from.

When the sums do not work, optimise for an earlier step that happens more often, such as a booked call. Then import the later outcome from your CRM, such as a won deal, so the campaign gradually learns what a good lead looks like. HubSpot can hold that data, and an offline conversion import carries it back into Google Ads.

After you change the strategy, the target or the budget by a large amount, the campaign goes through a learning period. That usually lasts one to two weeks. Do not edit during it. Move targets and budgets in steps of no more than 20 percent, and one at a time.

Bid adjustments for device, location and audience belong mainly to manual CPC. Under smart bidding Google largely handles these signals itself, so your adjustments carry little weight. With smart bidding, your levers are the target, the conversions you count and how you split campaigns. Use audiences to observe, not to restrict, until you have the volume to learn from them.

Searches for your own name convert cheaply and make any average look healthy. Put them in a separate campaign with its own target, so the non-brand campaigns are judged on what they earn. The layout is in Campaign structure and account setup.

• Setting the target from a hopeful number instead of from unit economics.
• Switching strategy every week because the last one has not worked yet.
• Counting every form fill as a conversion.
• Raising the budget and cutting the target on the same day.
• Judging a campaign during its learning period.
• Mixing brand and non-brand under one ROAS target.

After four weeks, the campaign spends most of its daily budget and cost per qualified lead sits within about 20 percent of the target. The number of conversions per week is steady, not jumping around. Your sales team confirms that the leads are the right kind. Only then lower the target by the next 10 to 15 percent. The defended-target logic is explained in Bid to a target you can defend, then leave the machine alone.