Bid Adjustments by Device & Geo: Modifiers That Move ROAS

Bid adjustments by device and geo are modifiers, usually expressed as a percentage or multiplier, that raise or lower your base bid depending on whether a user is on mobile or desktop, and which country or region they're in. Used well, they're one of the fastest ways to move ROAS without touching creative at all, because they let you pay more where conversion value is highest and less where it isn't, instead of one flat bid smeared evenly across every device and geo your targeting reaches.

Why one bid across device and geo leaves money on the table#

Every native network's auction is blind to your margin by default. It optimizes toward the bid and targeting you set, not toward where your actual conversion value concentrates. If mobile converts at half the rate of desktop for your offer, but you're bidding the same amount on both, you're either overpaying for mobile traffic or underpaying for desktop, and usually both at once. The same logic applies across geo: a Tier-1 country with higher average order values can absorb a higher bid than a Tier-3 country where the same click is worth less to you.

Device and geo modifiers exist specifically to correct for this without requiring separate campaigns for every combination.

Where the levers live#

Check each network's current documentation for the exact field names and whether the modifier is a hard multiplier or a soft signal the algorithm weighs alongside other factors; this changes periodically and the mechanics differ slightly by network. See how Taboola ads work and how Outbrain works for the base bidding mechanics these modifiers sit on top of.

Building your first device modifier#

Before setting any modifier, pull a device-split conversion report from your own tracking, not the network's click report. You need conversion rate and average order value by device, not just clicks or CTR, because a device can have a high CTR and a low conversion rate, which would lead you to bid the wrong direction if you only looked at engagement.

A simple starting approach:

1. Calculate effective CPA by device over your last meaningful sample (at least 2-3x target CPA in spend per device).
2. If mobile CPA is running 30% worse than desktop, start with a mobile bid modifier around -20% to -30%, not the full gap; auctions respond nonlinearly and an aggressive first cut can remove you from the auction entirely rather than just reducing volume.
3. Watch volume, not just CPA, for the first week. A modifier that's too aggressive can starve a device of impressions entirely, which sometimes makes CPA look artificially good on a tiny remaining sample.
4. Adjust again after a full week of data under the new modifier, since the mix of who sees your ad on each device shifts once the modifier is live.

Building your first geo modifier#

Geo modifiers follow the same logic but interact with geo tiers more directly. Tier-1 countries generally support higher bids because average order value is higher there; Tier-2 and Tier-3 geos usually need lower bids to stay profitable given typically lower order values, but can also have far less competition, meaning a modest bid still wins plenty of volume. This is the same dynamic behind scaling into new geos as a way to find cheaper, less-contested inventory once a Tier-1 campaign matures.

Don't set geo modifiers off assumption alone. Pull actual performance by country if your volume supports it; assumptions about which geos "should" convert well are wrong often enough that the data is worth the extra reporting step.

Combining device and geo modifiers#

The two interact. A geo that converts well on desktop might convert poorly on mobile within that same country, and treating "geo modifier" and "device modifier" as fully independent settings can produce a combined bid that's wrong in either direction. Where a network supports it, check combined device-by-geo breakdowns, not just each dimension separately, before finalizing modifiers. This is also where publisher-level bidding adds a third dimension: a publisher's audience skews toward a particular device and geo mix, so publisher, device and geo modifiers aren't fully independent levers even though the network dashboards present them as separate settings.

A worked example#

Say a campaign's overall CPA is sitting right at target, $35, blended across device and geo. Pulling the split shows US desktop converting at $22 CPA, US mobile at $48, and a secondary Tier-2 geo at $30 on both devices combined. The blended average is hiding a US desktop segment that's already very profitable and a US mobile segment that's actively losing money relative to target.

A reasonable first pass: bid up US desktop by 25-30%, bid down US mobile by 25-30%, and leave the Tier-2 geo alone since it's already near target. After a week, the blended CPA on the same total budget typically moves meaningfully toward target, purely from reallocating spend toward the segment that was already working, without touching creative, landing page or offer at all. The lesson generalizes: a blended metric that looks acceptable can be masking one segment subsidizing another, and modifiers are how you stop the subsidy without shutting off traffic entirely.

Common mistakes#

• Setting modifiers off industry assumptions instead of your own data. "Mobile always converts worse" is not universally true; it depends on your landing page, offer and checkout flow.
• Overcorrecting on a small sample. A -50% device modifier based on three days of data is usually reacting to noise, not a real pattern.
• Forgetting to revisit after a landing page or creative change. A mobile-unfriendly landing page fixed later means your old mobile modifier is now stale and probably too conservative.
• Ignoring interaction effects. Optimizing device and geo independently without checking the combined breakdown can leave real profit on the table or, worse, actively suppress a profitable combination.

Rolling modifiers into your reporting cadence#

Device and geo modifiers aren't a set-once configuration. Build them into the same weekly cadence you'd use for publisher-level bid reviews: pull the device and geo split alongside the publisher report, check whether last week's modifiers moved the blended CPA in the right direction, and adjust incrementally rather than resetting from scratch each time. Because the three dimensions, publisher, device and geo, interact, changing all three at once makes it hard to tell which adjustment actually caused a result. Where possible, change one dimension per week until you have a feel for how sensitive your specific account is to each lever.

Checking the plan against real market behavior#

Before finalizing bid adjustments, it helps to see whether your assumptions match how competitors in your vertical are actually behaving across device and geo. If competing creative in your niche runs primarily in Tier-1 geos on desktop and rarely appears on mobile in Tier-3 countries, that's a signal about where the profitable inventory tends to sit, gathered from real, currently-running campaigns rather than guesswork. OpenAdLibrary's ad intelligence index lets you check that pattern directly across networks before you commit budget to a modifier scheme built on assumption alone.

The takeaway#

Device and geo bid modifiers are a low-effort, high-leverage lever precisely because they don't require new creative, new offers or new landing pages, just a more accurate map of where your existing campaign's value already concentrates. Most accounts running a flat bid across device and geo are unknowingly subsidizing their worst-performing segments with their best-performing ones. Fixing that split is usually worth more to ROAS in a week than most creative iteration cycles are worth in a month.