Bid Adjustments in Google Ads 2026: What Still Works
Half the accounts that come across an audit desk still carry bid adjustments set two or three years ago: −20% on tablets, +15% for evening hours, +30% on a “site visitors” audience. The owner believes they’re working. In most cases they do nothing at all — because the campaign moved to an automated bid strategy long ago, and automated strategies ignore almost every one of those modifiers.
Here’s the factual version: how the modifier mechanic actually works, which adjustments still influence the auction in 2026, the two exceptions that survive under Smart Bidding, how to analyse mobile versus desktop properly (not from the default device report), and what to manage instead once bidding belongs to the algorithm.
What a bid adjustment actually does
An adjustment is a multiplier applied to your bid for a specific slice: device, location, time of day, audience, demographic. The math is trivial: final bid = base bid × (1 + adjustment). A +20% modifier turns a $1.00 bid into $1.20; −50% turns it into $0.50; −100% means you stop buying impressions in that slice entirely.
The operative word is bid. Adjustments were designed for a manual world where you set the click price and the system multiplies it. In an automated world you don’t set the click price, so there’s nothing left for the multiplier to act on.
The core fact: under Smart Bidding, most adjustments do nothing
Automated strategies — Target CPA, Target ROAS, Maximise Conversions, Maximise Conversion Value — recompute a bid in every auction using hundreds of signals: device, time, location, language, interaction history, query context, and many more. Every slice you were manually multiplying is already priced inside the model.
So the system simply doesn’t apply your multipliers. The summary:
| Adjustment type | Manual CPC | Smart Bidding |
|---|---|---|
| Device | Applies | Two cases only — see below |
| Location | Applies | Ignored |
| Ad schedule | Applies | Ignored |
| Audience | Applies | Ignored |
| Demographics (age, gender, income) | Applies | Ignored |
| Interaction (calls) | Applies | Ignored |
| Topics and placements | Applies | Ignored |
| Content type | Applies | Ignored |
The practical consequence: if your campaigns run automated strategies, those percentages in the interface are archaeology. They don’t hurt, they don’t help, and worst of all they create a false sense of control. The damaging version is a specialist “optimising” by nudging modifiers, watching numbers fluctuate (normal noise), and drawing conclusions from it.
The two exceptions that still work
1. Device exclusion at −100%
A −100% device modifier works under every strategy, automated included. It isn’t a price change — it’s an exclusion: the campaign stops entering auctions on that device type.
This is the only remaining hard lever for “don’t serve us on tablets” or “mobile only” at campaign level. A switch, not a dial.
2. Device adjustments under Target CPA
With Target CPA, device adjustments still apply — but they mean something different. They modify the CPA target for that device, not the bid. A +20% mobile adjustment says “for mobile conversions, an acceptable cost per conversion is 20% higher.”
That’s a legitimate tool when device economics genuinely differ — for instance, mobile leads close at a lower rate and you’re willing to pay less for them. A negative mobile adjustment then expresses business reality rather than an attempt to outsmart the algorithm.
Note that changing a target changes the strategy’s inputs and triggers recalibration. The mechanics are in the Smart Bidding learning period, and this year’s shift in target-based strategy behaviour is covered in the change to target-based bid strategies.
Manual CPC: adjustments multiply, they don’t add
If you still run manual campaigns — usually tight brand campaigns or heavily controlled tests — all eight adjustment types apply. And here’s the arithmetic that quietly burns budgets.
Adjustments compound multiplicatively. Example: a $1.00 base bid with +30% mobile, +40% audience, and +25% evening schedule.
Intuition says +95%, so $1.95. Reality: 1 × 1.30 × 1.40 × 1.25 = $2.28. Your bid is 2.3× base, not 2×. Across three layers the gap is tolerable. Across five or six layers of +20–30% each, you can easily be bidding three times your base and genuinely not know where the overspend came from.
The rule that follows: in manual campaigns, keep no more than two or three adjustment layers active at once, and compute the compound multiplier by hand before adding another.
To sanity-check what a bid change does to volume, use the built-in forecasting tools — covered in bid, budget and target simulators.
Mobile vs desktop: how to read the difference correctly
The most common conclusion drawn from the default device report is “mobile CPA is 50% worse, cut mobile.” That conclusion is usually wrong, for three reasons.
One: cross-device behaviour. Someone discovers you on a phone at lunch and buys that evening on a laptop. Under last-click, the conversion lands on desktop while mobile carries only cost. The report says mobile is bad; mobile was the entry point.
Two: different conversion types. On phones people call and message; on desktop they fill forms. If you only track form submissions, mobile conversions physically cannot appear in the report. That’s a tracking gap, not a device problem.
Three: the landing page. Very often “bad mobile” is just a slow or awkward mobile experience. A two-to-three-second speed gap destroys more conversion rate than any bid modifier could recover. Before cutting traffic, check what happens on the page — the method is in landing page speed and Core Web Vitals.
The correct analysis order:
- Confirm every conversion type is tracked on both devices — calls, forms, chats, checkouts.
- Look at multi-touch data rather than last click, or entry and purchase will always split across devices.
- Evaluate on-page behaviour separately for mobile and desktop: speed, bounce, scroll depth.
- Compare full economics rather than CPA: how many mobile leads reach a closed deal, and at what value.
- Only then decide whether a device-level CPA target adjustment is warranted — or whether the device was never the issue.
When excluding a device is genuinely right
A full −100% exclusion is a blunt decision, and the valid reasons are few:
- The product is technically unusable on that device. Desktop software, large-screen interfaces, long B2B forms nobody completes on a phone.
- No mobile version of the site exists. Until it does, mobile traffic is a paid bounce.
- Tablets with negligible volume and consistently poor results across at least several hundred clicks. On a small sample, “bad tablet” is noise.
- Campaigns deliberately built for app installs or for phone calls only.
In every other case, fix the target, the landing page, and the tracking before amputating an audience. Exclusion is irreversible in the sense that you’ll never learn what it would have produced.
What to manage instead when you’re on automated bidding
- Conversion data quality. An automated strategy is exactly as smart as the signals you feed it. Duplicates, untracked conversion types, and junk micro-goals distort bids far more than any modifier could correct.
- Conversion value instead of conversion count. If different conversions are worth different amounts, send the value — that lets the algorithm do natively what modifiers used to approximate. See conversion value rules and value-based bidding.
- Audiences in observation mode. Even when modifiers are ignored, observation gives you the segmentation you need to make structural decisions. Covered in audience observation.
- Campaign separation instead of multipliers. If a slice needs fundamentally different economics, give it its own campaign and its own target. Under automated bidding that’s the only reliable way to control segment economics.
- Location targeting instead of location modifiers. Since geo multipliers are ignored, manage the targeting itself — see location targeting: presence vs interest.
- Ad schedule as a serving restriction, not a multiplier. Time-of-day modifiers are ignored, but the schedule itself still gates delivery. See ad scheduling and dayparting.
Calculating the real economic gap between devices
If you do want to justify different CPA targets per device, calculate down to revenue rather than reading the ads report. A lead generation example:
| Metric | Desktop | Mobile |
|---|---|---|
| Leads per month | 400 | 900 |
| Spend | $12,000 | $18,000 |
| Cost per lead (ads report) | $30 | $20 |
| Lead-to-deal close rate | 22% | 9% |
| Deals | 88 | 81 |
| Average deal value | $1,400 | $900 |
| Revenue | $123,200 | $72,900 |
| Ad spend as share of revenue | 9.7% | 24.7% |
Look at what happens. In the ads report mobile looks better — leads are a third cheaper. In revenue terms mobile is markedly worse: ad spend eats 2.5× more of the revenue it produces. This is precisely why “mobile is cheaper, let’s push more budget there” is such an expensive mistake.
From here there are two fundamentally different scenarios, and conflating them is costly.
Scenario A: the gap is real. Mobile leads genuinely produce less revenue — different audience profile, smaller basket, weaker intent. A device-level CPA target adjustment is justified. Work backwards from the same acceptable ad share of revenue: at 9.7%, mobile revenue of $72,900 supports $72,900 × 0.097 ≈ $7,071 in spend. Divide by 900 leads and the allowable mobile cost per lead is about $7.90 — meaning your actual $20 is nearly triple what the economics support. The device adjustment should be negative and substantial.
Scenario B: you created the gap. Mobile leads close worse because the form is longer than it needs to be on a phone; mobile calls aren’t tracked so some deals never appear in the report at all; sales follows up on mobile leads a day later than desktop ones. In that case a bid modifier just freezes the problem in place — the fix belongs to the landing page, the tracking, and the sales process.
How to tell them apart: compare close rates by lead source (form / call / chat), not only by device. If mobile calls close as well as desktop forms, you’re in Scenario B and no bid adjustment is warranted.
Auditing your adjustments, step by step
- Export every active adjustment across campaigns and ad groups — device, location, schedule, audience, demographics.
- Note each campaign’s bid strategy. Anything that isn’t Manual CPC or Target CPA is a candidate for zeroing out.
- Remove adjustments that provably don’t apply. This won’t move performance directly, but it removes the illusion of control and simplifies future diagnosis.
- Review every −100% separately: is a device still excluded that stopped being a problem years ago?
- For Target CPA campaigns, recalculate device adjustments from real economic difference rather than reported CPA gaps.
- For manual campaigns, compute the compound multiplier and confirm it’s within a sane range.
- Log the change date. When you’re diagnosing a future dip, that log saves hours.
Bid adjustments across campaign types
The available levers differ by campaign type, not just by bid strategy. A short map:
| Campaign type | What you actually get |
|---|---|
| Search | Full adjustment set on Manual CPC; on automated bidding, only the two exceptions |
| Shopping | Device and location adjustments under manual bidding; on automated bidding, control moves to targets and product group structure |
| Display | Adjustments exist nominally, but the same ignore rules apply on automated bidding; real levers are placement exclusions, frequency, and audiences |
| Video | Control comes from campaign subtype, objective, reach and frequency; classic multipliers barely matter |
| Demand Gen | No conventional adjustments; control is audience signals, creative, targets, and frequency caps |
| Performance Max | No adjustments; control is targets, signals, exclusions, and asset group structure |
| App | No adjustments; control is target cost per install or per action, plus creative assets |
The pattern is unambiguous: the newer the campaign type, the fewer manual multipliers it exposes and the more weight sits on inputs — conversions, values, audience signals, and creative quality. That’s the direction worth building skill in, rather than hunting for one more place to add twenty percent.
Common mistakes
| Mistake | What actually happens |
|---|---|
| Tuning geo and audience modifiers on automated bidding | Nothing changes; normal noise gets read as effect |
| Cutting mobile based on the default CPA report | You amputate the entry point of a cross-device path |
| Adding modifiers together mentally on Manual CPC | Actual bid ends up well above what you planned |
| Setting −100% on tablets “just in case” | Segment lost with no data on its real economics |
| Treating a Target CPA device modifier as a bid change | Wrong math: it moves the target, not the click price |
| Leaving three-year-old adjustments in place | False sense of management, extra noise during diagnosis |
A bid adjustment answers the question “how much more should I pay for this slice?” If you’re not setting the bid, the question no longer has meaning. What you manage instead is data, campaign structure, and landing pages.
Related reading
If you’re cleaning up bid management, the adjacent surfaces are worth the same pass: portfolio bid strategies and shared budgets, where the boundaries of a single economic unit get decided; frequency caps and Target Frequency if you run video or Display alongside Search; and the Local Services Ads migration into Google Ads, where manual cost-per-lead bidding disappears for good.
The structured version of account management sits in the Google Ads guide, and the campaign infrastructure side is what PPC Rebels handles.
FAQ: bid adjustments in Google Ads
Do bid adjustments work with Maximise Conversions?
No, with one exception: a −100% device modifier still functions as a device exclusion. Location, schedule, audience, and demographic modifiers are ignored.
What about Target CPA?
Device adjustments apply, but they modify the CPA target for that device rather than the bid. Other adjustment types are ignored.
Should I delete old adjustments that no longer apply?
They cause no direct harm, but they create a false sense of control and complicate diagnosis. Zero them out during an audit and log the date.
How do I stop serving on tablets?
Set a −100% device adjustment for tablets at campaign level. It works under any strategy. Make sure the decision rests on a meaningful sample, not a few dozen clicks.
Why is my Manual CPC bid higher than I calculated?
Because adjustments multiply rather than add. Three +30% modifiers don’t produce +90% — they produce a 2.2× bid. Compute the product, not the sum.
Mobile CPA is worse than desktop. Should I bid mobile down?
Not immediately. First check tracking completeness (calls and chats are frequently missing), cross-device paths, and mobile landing page quality. In most audits the problem lives in one of those three places, not in the device.
Can I set an adjustment at ad group level?
Device adjustments can be set at campaign and ad group level, with the ad group value taking precedence where present. All of it still obeys the strategy rules — under automated bidding only the two exceptions above apply.
How do I control a segment if modifiers don’t work?
Move the segment into its own campaign with its own target. Under automated bidding that’s the only dependable way to give a segment different economics.
Do adjustments apply to Performance Max?
PMax doesn’t expose the familiar adjustment set. Control comes through targets, audience signals, exclusions, and asset group structure. Don’t expect Search-campaign behaviour there.
Should I go back to Manual CPC for the control?
Usually not. Manual bidding gives you modifiers but gives up auction-time bid recalculation across dozens of signals. It stays justified in narrow cases: small brand campaigns, controlled tests, or accounts with extremely thin conversion data.
How can I verify an adjustment applied at all?
Compare segmented data before and after at comparable volume — and check the bid strategy first. If the strategy is automated and the modifier isn’t one of the two exceptions, it didn’t apply, and any difference in the numbers has another cause.
What do I do with demographic adjustments that “used to work”?
They did work in the manual bidding era. Moving to an automated strategy ended their effect, whether or not they remained in the interface. If the demographic difference is real and economically meaningful, express it through conversion value or a separate campaign.
