Lesson 4 of 5 · 9 min read · advanced
Automation rules and bid strategies: letting machines mind the shop
How if-this-then-that rules pause losers and protect budgets, how platform bid strategies differ, and why delayed revenue makes naive rules dangerous.
An arbitrage account can hold hundreds of ads across several platforms, each spending money every minute of the day and night. No person can watch them all. Two kinds of automation fill the gap: rules you write, which act on your data, and bid strategies the platform runs, which decide what to pay for each impression.
Think of a thermostat and a pilot. A thermostat follows one simple rule: if it is colder than 20 degrees, switch the heating on. That is an automation rule. A pilot makes thousands of small judgements you never see, guided only by the destination you gave. That is the platform's bid strategy. You need both, and you need to know which one is flying.
Automation rules
Automation rules are conditions and actions: if a campaign, ad or placement meets these conditions over this time window, do this. They live in the ad platform itself, in a tracker (Voluum calls its version Automizer) or in a dedicated tool such as TheOptimizer, which reads cost from the ad account and revenue from the tracker or feed and then pauses, restarts or re-budgets.
| Rule type | Example (illustrative thresholds) | Purpose |
|---|---|---|
| Stop-loss | If an ad has spent $30 and earned under $10 in settled hours, pause it | Cap the cost of a failed test |
| Placement block | If a native site has sent 300 clicks with a lander CTR under 10%, add it to the blocklist | Remove sources whose visitors do not engage |
| Budget step-up | If ROI over the last 3 days exceeds 30%, raise budget by 20% | Grow winners gradually (scaling) |
| Budget step-down | If ROI over the last 2 days is below 0%, cut budget by 30% | Shrink before stopping |
| Schedule | Lower bids between 1am and 6am local time | Act on dayparting patterns |
| Alarm | If spend is rising and revenue has been zero for 3 hours, pause and notify | Catch a broken link or feed outage |
The trap: revenue that has not arrived yet
Cost shows up within minutes. Feed revenue shows up over hours (revenue reporting delay). A rule that compares the two over 'today so far' will see full cost and partial revenue, conclude that everything is losing, and pause your best campaigns by lunchtime.
A worked example with made-up numbers: by 2pm a campaign has spent $200. The feed has so far reported $150. A rule set to pause below break-even fires. But the revenue for the last four hours is incomplete; once it settles, the day so far is worth $240, a 20% return. The rule killed a profitable campaign because it compared a finished number with an unfinished one.
- Exclude unsettled hours. Evaluate a window that ends several hours ago, matched to how long your feed takes to firm up.
- Use both cost and volume conditions. Require a minimum spend and a minimum number of clicks before any rule can act.
- Prefer leading indicators for speed. Lander CTR is known immediately. A source whose visitors never click a search term can be judged without waiting for revenue.
- Adjust estimates by history. If your estimated revenue usually finalises 8% lower, build that haircut into the rule.
- Limit the blast radius. Cap how much any rule can raise a budget in one day, and have rules notify you when they act.
Platform bid strategies
A bid strategy is the instruction you give the platform's own bidding system. Names vary by platform and change often, but the choices fall into four families.
| Family | You tell the platform | Good for | Watch out for |
|---|---|---|---|
| Spend the budget | Get as many results as possible for this budget | Early testing, simple setup | It will spend the budget even when results are expensive |
| Cost goal (cost cap, Target CPA) | Keep the average cost per result near this figure | Holding margin when you know your break-even point cost | Set too low and the campaign barely spends |
| Bid cap | Never bid more than this in any auction | Strict control | Needs frequent manual adjustment |
| Value goal (Target ROAS) | Bring back this much revenue per unit of spend | Accounts sending reliable value events | Only as good as the values you send |
How do you choose a cost goal? Work backwards from the feed. Say, for illustration, that a visitor is worth $0.20 on average (RPV) after your usual haircut for finalisation, and you want a 20% margin on spend. Then you can pay at most $0.20 ÷ 1.2, about $0.167 per visitor. If the event you optimise for is a monetised click, and one visitor in four produces one, the most you can pay per event is about $0.67. The goal comes from your unit economics, not from what the platform suggests.
Your rules do well at
- Hard stops and safety limits
- Blocking placements and sources
- Budget changes on settled data
- Alerting you to outages
Platform bidding does well at
- Choosing which person to show the ad to
- Setting a price per auction
- Shifting delivery through the day
- Learning from value signals you send
The two can fight. A rule that pauses an ad set at noon because its morning looked poor denies the platform's bidding the afternoon in which it would have recovered. Decide which decisions belong to rules and which to the platform, and write that down.
Key takeaways
- Automation rules are if-then instructions acting on your own data; bid strategies are instructions to the platform's bidding system.
- Never compare complete cost with incomplete revenue: evaluate rules on hours that have settled.
- Require minimum spend and click volume before a rule may act, and cap how far it can raise budgets.
- Derive cost and ROAS goals from your own revenue per visit and target margin.
- Frequent edits reset platform learning; automation never replaces a manual policy check of creatives.
Questions people ask
What are automation rules in media buying?
They are if-then instructions that act on campaigns without a person present, for example: if an ad has spent $30 and earned under $10, pause it. They can live in the ad platform, in a tracker or in a tool such as TheOptimizer. In search arbitrage they are mainly used for stop-losses, blocking weak placements, stepping budgets up or down and raising alarms when revenue stops.
Why do my rules pause profitable campaigns?
Almost always because the rule looks at today so far. Cost is reported within minutes, but feed revenue for the last few hours is incomplete, so recent performance looks worse than it is. Set the evaluation window to end several hours in the past, require a minimum spend before acting, and compare what the rule saw with what the revenue later settled at.
Which bid strategy is best for search arbitrage?
There is no single best one. New campaigns often start with a simple spend-the-budget strategy to gather data. Once the cost you can afford per result is known, a cost goal protects margin. Accounts that send reliable revenue values back can use a value or ROAS goal. Whatever you pick, the target should come from your own revenue per visit, not a platform default.