Microsoft Advertising automation has reached the point where refusing it costs you money and accepting it uncritically costs you more. Automated bidding, broad match with smart matching, responsive assets, recommendations applied on your behalf: each of these can improve an account, and each of them can quietly rewrite decisions you made deliberately. The skill is no longer knowing how to set a bid. It is knowing which decisions are safe to delegate and which ones encode business knowledge the system cannot infer.
A useful test is whether the decision is about optimisation or about definition. Optimisation questions, such as how much to bid for this auction at this moment for this user, involve more variables than a human can weigh and belong to the machine. Definition questions, such as what a conversion is worth, which products are profitable, which queries represent the wrong customer, are business facts. Hand a definition question to automation and it will invent an answer from whatever data it can see, which is usually the wrong one.
What to hand over
Bid management first. Manual bidding on any account with meaningful volume is no longer competitive, because the signals available at auction time are not visible to you. Give it a target that reflects real economics, then stop adjusting daily. Accept that performance will fluctuate while the system learns, and judge over weeks. The most common mistake we inherit is an account where someone reset the target every few days and the strategy never stabilised.
Query matching is the second. Broad match with good conversion data finds phrasing you would not have thought of, particularly in the long tail where every search is slightly different. The same applies to asset combinations: responsive formats test permutations faster than any manual schedule. Import from your search campaigns elsewhere where it saves time, but review the import rather than letting it run unattended, since structures that suit one platform do not always translate.
What to keep
Everything that encodes what your business actually wants. Automation optimises toward the goal you give it, with total commitment and no judgement about whether the goal was sensible. If the goal is wrong, the better the automation, the worse the outcome, which is a genuinely counterintuitive property of these systems.
- Conversion definitions and values. If a newsletter signup and a sale both count as one conversion, the system will buy newsletter signups.
- Negative keywords, particularly brand, competitor, and job seeker terms that look like intent and are not.
- Budget allocation across product lines, which reflects margin and inventory the platform cannot see.
- Brand safety and placement exclusions on audience network inventory, where reach expands faster than relevance.
- The decision to pause something. Automation will keep spending on a losing structure indefinitely if the target is loose enough to permit it.
Keeping Microsoft Advertising automation honest
Automation needs supervision that is scheduled rather than reactive. We run a fixed weekly review that looks at the search terms report, the share of spend going to audience network placements, and any recommendation that was auto applied since the last check. Auto apply is the setting that causes the most surprises, because changes arrive without anyone requesting them and often look reasonable in isolation while conflicting with a deliberate decision made months earlier.
Keep a written change log. It sounds bureaucratic until the week performance shifts and nobody can remember whether the drop followed a bidding change, a landing page deployment, or a seasonal effect. A dated line per change turns a speculative argument into a five minute check. This matters more with automation, not less, because the system is also making changes and you need to separate its decisions from yours.
The account still needs a structure
There is a fashionable view that Microsoft Advertising automation makes account structure irrelevant. It does not. Structure is how you express priorities: what gets its own budget, what gets its own target, what is measured separately. Collapse everything into one campaign and you lose the ability to say that this product line must hit a tighter return while that one is being grown deliberately at a looser one.
What has changed is that structure should now be coarser. Dozens of tightly themed groups fragment the conversion data that automated bidding needs, so consolidate until each unit produces enough conversions to learn from, then stop consolidating. Used this way, Microsoft Advertising automation does the work humans were never good at, which is second by second auction judgement, while the account keeps expressing decisions only the business can make. That division of labour is what separates accounts that improve steadily from accounts that drift while everyone assumes the machine has it handled.
Keep reading: Microsoft Advertising (Bing Ads) · Advertising