What you'll learn in this article
- What happened to conversions in the 14 days after I raised bids on 50 keywords at once
- Why the first seven days told me almost nothing, and what the second seven revealed
- The three groups the 50 keywords split into, and what separated them
- The exact sequence I now use when changing keyword bids on a live account
- Which signals told me a bid increase was buying volume rather than value
Last quarter I raised max CPC on 50 keywords in a single manual CPC campaign, on the same afternoon, and then did nothing for 14 days. No ad edits, no budget changes, no negative additions. I wanted to see the effect isolated, because every time I've read about how to change keyword bids in google ads the article stops at the click path and says nothing about what the account does afterwards.
The headline result: spend rose 41%, clicks rose 28%, conversions rose 9%. Cost per conversion went from €38 to €48. That looks like a failure. It wasn't but understanding why took reading the 50 keywords as three separate populations rather than one average.
The setup, so the numbers mean something
Lead gen account, B2B services, single-country targeting, manual CPC with enhanced CPC off. Around €9,000/month in that campaign. The 50 keywords were all exact and phrase match terms with at least 40 clicks and 1 conversion in the preceding 90 days so nothing brand new, nothing statistically empty.
What I changed and by how much
I applied a flat +25% to max CPC across all 50. Flat, deliberately, because a percentage tailored per keyword would have made the outcome impossible to attribute. Google Ads lets you do this in one operation from the keywords table the official documentation on editing your bids describes selecting the keywords, opening the Edit dropdown and choosing to change max CPC bids, where you can set a new bid, increase or decrease the existing one, or raise bids to first page or top of page CPC. That last option is the one I avoided here, precisely because it applies a different multiplier to every keyword.
The baseline I measured against
I used the 14 days immediately before the change, not a month-long average. Two weeks against two weeks, same days of the week, no holidays in either window. Anyone comparing a post-change fortnight against a quarterly average is measuring seasonality as much as bidding.
One thing I fixed first: I ran a duplicate keyword check across the account. Two of my 50 existed in another ad group at a lower bid, which would have made the result unreadable. I paused the duplicates before starting.
The 14-day data, day by day
The aggregate hid everything interesting. Here's the shape of it.
Days 1-3: cost moved, conversions didn't
Spend jumped almost immediately by day two it was running roughly 35% above baseline. Conversions were flat. If I had judged the change here, I would have rolled it back, and I would have been wrong. Impression share gains show up in hours; the conversion lag on this account is four to six days.
Days 4-7: volume arrived, quality dropped
Clicks caught up and conversions started rising, but conversion rate fell from 5.1% to 4.2%. This is the part nobody warns you about. A higher bid doesn't just win more of the same auctions it wins auctions you were previously losing, and you were losing them for a reason. The incremental click is, on average, a worse click than your existing average click.
Days 8-14: the split appeared
By the second week the 50 keywords had separated cleanly into three groups.
Group A 11 keywords, clearly better. Conversions up 60%+, cost per conversion flat or improved. Every single one had been sitting below 60% impression share lost to rank before the change. They were genuinely underbid.
Group B 27 keywords, more expensive, no more conversions. Cost up 30-50%, conversions within noise of baseline. These were already winning most of the auctions they could win. I paid more for the same traffic.
Group C 12 keywords, actively worse. Volume up substantially, conversion rate collapsed by half or more. These pulled in query variants that don't convert. Raising the bid widened the door, and what came through wasn't buying.
Group A alone would have delivered a 22% conversion lift at a lower cost per conversion. Groups B and C together consumed most of the extra spend and returned almost nothing. The blended average 9% more conversions at a worse CPA was an artefact of averaging three different behaviours.
How I change keyword bids now
The mechanics take thirty seconds. The decision takes longer, and it happens before you touch the interface.
Step 1 Segment by lost impression share (rank) first
This is the single filter that predicted Group A membership. If a keyword isn't losing impression share to rank, a bid increase has almost nowhere to go. I no longer raise bids on any term below roughly 30% lost IS (rank) unless I have a specific reason.
Step 2 Check the search terms before, not after
Group C was predictable in hindsight. Every one of those keywords already had a messy search terms report at the old bid. Raising the bid amplified an existing matching problem. Fifteen minutes with the search terms data would have saved me a third of the wasted spend.
Step 3 Change in one batch, on one day
Staggering bid changes across a week makes attribution impossible. One batch, one timestamp, noted somewhere you'll find it in two weeks. This also matters for anyone reasoning about google adwords keyword rates over time: if you can't say exactly when a rate changed, you can't say what it caused.
Step 4 Size the increment to the gap, not to a round number
A flat +25% was right for an experiment and wrong for production. In normal work I move toward the top of page bid range rather than by a fixed percentage a keyword sitting 60% below the range needs a different move from one sitting 5% below. That reasoning is the same one I apply when setting keyword bids from scratch.
Step 5 Wait 14 days, then judge at keyword level
Not seven. Seven days on this account showed a bid increase looking like a failure that was actually working on a fifth of the terms. And judge each keyword individually the campaign average will tell you almost nothing useful.
What I infer from bid changes now
Cost responding faster than conversions is normal, not a red flag. Spend reacts within hours; conversions carry your full lag. Any judgement made in the first 72 hours is judging a metric that has moved against one that hasn't.
If conversion rate falls proportionally to the volume gain, the keyword was already at its ceiling. That's Group B. You bought the same traffic more expensively, which is the most common outcome of an untargeted bid increase.
A collapse in conversion rate alongside a big volume jump is a matching problem, not a bidding problem. Group C didn't need a lower bid it needed tighter matching. Bidding down would have hidden the symptom and kept the cause.
Flat conversions with flat cost means the bid never bound in the first place. A few of my 50 barely moved at all. Their old bid was already above what the auction demanded, so raising the ceiling changed nothing. This is worth checking before assuming an increase "didn't work".
Impression share lost to rank is the best single predictor of whether a bid increase will pay. Of the 11 keywords in Group A, every one had headroom by that metric. Of the 27 in Group B, almost none did.
What I stopped doing
Judging bid changes on campaign averages. The average said the test failed. The keyword-level data said a fifth of it worked well and should be kept. Those are different conclusions from the same numbers.
Applying flat percentages across a keyword list. Useful for an experiment, harmful in production. Different keywords sit at different distances from the auction price.
Using "raise to top of page" on a whole selection. It's available in the same bid editing dialog and it's tempting, but it applies an uneven and invisible multiplier you lose the ability to explain what you changed.
Reacting to day 3. Two weeks is the minimum on a lead gen account with a multi-day conversion lag. On accounts with longer sales cycles I extend it further.
Raising bids before checking search terms. A bid increase magnifies whatever matching behaviour already exists. If that behaviour is bad, you're paying to make it worse.
The practical takeaway
Knowing how to change keyword bids in google ads is a two-minute interface skill. Knowing which keywords deserve the change is the whole job, and the filter that worked for me is impression share lost to rank plus a clean search terms report.
Out of 50 keywords, 11 justified the increase. I kept those, reverted Group B, and rebuilt the matching on Group C instead of touching its bids. Net effect after the correction: conversions up 19% on roughly flat spend. The experiment that looked like a 9% gain at a worse CPA became a real gain but only because I read it at keyword level.