Taboola Targeting Options Explained: Geo, Device, Audiences & SmartBid
Taboola targeting is a different animal from social: geo, device, audiences, publisher controls and SmartBid. Here is what each lever does and a campaign structure to start from.

Realize, Taboola's ad platform, lets you target by location (country, region or state, DMA, city and postal code, with sub-country targeting limited to one country per campaign), platform (desktop, smartphone, tablet), operating system and version, browser and connection type. Its audience options include pixel-based retargeting, customer files, US-only lookalikes, predictive audiences, Taboola first-party and third-party marketplace segments (including age, gender, interest and intent), contextual categories and custom topics, and search-keyword and mail-domain segments in selected markets. On top of that come publisher controls (block lists, by-site bid adjustments on manual strategies, and approved site lists through an account manager), scheduling by day and hour, and bidding through Maximize Conversions, Enhanced CPC or Fixed Bid. Realize's own advice for a new campaign is to start broad and narrow later, because on a native network your creative, landing page and publisher mix still do much of the audience selection.
Taboola targeting is not Facebook targeting#
If you arrive from Meta, recalibrate first. Taboola ads run inside sponsored feeds and recommendation widgets on publisher pages, and Realize does offer demographic and interest targeting: its marketplace includes age and gender segments (Taboola recommends its "declared" ones for precision) plus interest and intent segments, and you can combine segments with AND logic. But availability varies by country, every overlay narrows reach, and Realize recommends that new campaigns skip audience targeting at first because it limits reach and raises the CPC needed for scale. Stacking narrow segments on day one is an easy way to burn a first budget. Three mental shifts:
The publisher context is part of the audience. An article page about retirement planning attracts retirement-curious readers, and you can act on that with Realize's 200+ contextual categories and custom topic segments as well as publisher bids and blocks. The creative is a filter. A headline about sciatic nerve pain draws sciatica sufferers on a general news site. This is why native buyers spend much of their effort on hooks and angles; the mechanics are covered in how Taboola ads work. Targeting is mostly subtractive. You launch reasonably broad, then cut: geos that never pay back, devices that break your funnel, publishers that eat clicks without converting.
Here is the stable core of the targeting stack. Taboola ships changes to campaign settings regularly, so treat this as the durable skeleton and check the current campaign-setup documentation for the exact options in your account:
| Dimension | What you can set (Realize help center, checked 7 October 2026) | Practical use |
|---|---|---|
| Location | Country, region/state, DMA, city, postal code; sub-country targeting one country per campaign | One campaign per geo or tight geo group |
| Platform | Desktop, smartphone, tablet | Realize recommends one platform per campaign |
| Operating system | OS family and version | Offer compatibility, app installs |
| Browser | Browser families, including in-app webviews | Cleanup tool; browser-specific products |
| Connection type | Wi-Fi only | Heavy landing pages, large app downloads |
| Audiences | Pixel retargeting, customer files, US lookalikes, predictive, marketplace (age, gender, interest, intent), In-Market For (beta), search-keyword and mail-domain segments | Retargeting and exclusions first |
| Context | 200+ contextual categories; custom topic segments | Include or exclude contexts |
| Publishers | Block lists; by-site bid adjustments on Fixed Bid and Enhanced CPC; approved lists through an account manager | Your main optimization surface |
| Schedule | Days and hours; bid modifications by day and hour on manual strategies | Only after data shows patterns |
| Bidding | Maximize Conversions (optional target CPA), Enhanced CPC, Fixed Bid | Automation vs control trade-off |
Geo targeting: the highest-leverage split#
Geography is the first thing to get right because it moves everything else: CPCs, competition density, language, and payout on the back end. A common rule, and one Realize's location-targeting guide points toward, is one campaign per country (or per small group of same-language countries), not one worldwide campaign. Costs and auction pressure differ so much between a US desktop click and a Tier-3 mobile click that blending them makes your reporting unreadable and lets cheap low-intent traffic soak up the budget.
The classic Tier 1 / Tier 2 / Tier 3 framing applies directly: advertiser demand and click prices differ widely by country, so a US desktop click and a Tier-3 mobile click are different products. This guide has no representative CPC dataset, so it gives no range; get current estimates for your geo, device and vertical, and see our native CPC benchmarks for the network-level picture. If you target regions, DMAs or postal codes, Realize suggests bidding roughly 30% to 50% above your run-of-network CPC to keep volume.
Sub-country targeting (state, region, DMA, city or postal code) matters for offers with genuine geographic economics: insurance (rates vary by state), home services, local lead-gen, and anything with licensing boundaries. For a national ecommerce offer, sub-geo targeting usually just fragments your data.
One trap worth naming explicitly: targeting geo and offer geo must agree end to end. If your lead-gen form only accepts US ZIP codes, every Canadian impression is a wasted click no matter how tidy "North America" looked in campaign settings; if your affiliate payout table pays differently per country, a blended campaign hides which geo is actually carrying the profit. Match each campaign's geo to the narrowest unit your economics care about, and verify the lander, the form and the payout all agree with it.
Two more field-tested details:
- Match language to geo, always. An English creative served in a French geo is likely to convert badly, and a low CTR also hurts delivery: Realize's auction recommends items with higher CTR more often, so with a weak CTR you need a higher bid for the same reach.
- Expand geo by cloning, not by editing. When a US campaign works, clone it into UK or AU with localized creative rather than adding countries to the working campaign. The playbook for that expansion is in scaling to new geos.
Device, OS and browser targeting#
Consider splitting desktop and smartphone into separate campaigns. Realize's platform-targeting article says costs per platform may differ significantly and recommends targeting one platform per campaign; its new-campaign guide, by contrast, suggests leaving targeting broad at first to gather data, so a reasonable compromise is a short broad read followed by a device split. Landing-page behavior also differs: mobile readers move through advertorials differently and convert at different rates.
Beyond the split:
Tablet. Check tablet volume and performance in your reports before giving it its own campaign; you can bundle it with desktop or exclude it until everything else is optimized. Operating system targeting earns its keep when the offer cares: app installs, carrier-billing flows, and offers where iOS and Android audiences monetize differently. Realize lets you target OS family and version. For a standard lead-gen or ecommerce flow, run both and check the split in reporting before acting. Browser targeting is a cleanup tool, not a strategy, though it matters for browser-specific products such as extensions. If reporting shows one browser converting far below the rest (sometimes a sign of in-app webviews or tracking problems on cookie-blocking browsers), exclude it then. Don't pre-optimize.
The honest hierarchy: geo and device splits are structural (settle them in the first weeks); OS and browser are reactive (act on data).
Audience targeting: what's actually available#
Realize's audience layer is broader than many native buyers assume, but every overlay narrows reach and segment availability varies by country. Treat audiences as a supporting act at first:
Retargeting. Install the Taboola Pixel before you spend a dollar: it is your conversion pixel and it builds site-visitor audiences you can target or exclude. Retargeting pools on native are often small, so judge retargeting campaigns on your own CPA rather than assuming they will be the most profitable line. First-party audiences. Customer files (hashed emails or mobile IDs, at least 1,000 records) can be uploaded for the countries Realize lists, mainly for suppression and re-engagement, and Realize can exclude already-converted users automatically. Marketplace audiences. Taboola first-party and third-party segments based on demographic, interest and intent signals, plus In-Market For segments (in open beta, for selected verticals). Test them skeptically: they narrow scale, Realize advises raising manual bids when you use them, and the creative may filter the audience more cheaply. Lookalikes and predictive audiences. Realize's lookalikes are built from a customer file and are available only for US targeting (2,000 to 750,000 records); predictive audiences are built from your pixel or server-to-server conversions once you have 100 of them on Realize. Search and mail signals. In selected markets, Realize offers segments of users who searched for chosen keywords or received mail from chosen domains.
The practitioner stance: broad prospecting campaigns with strong creative, plus a retargeting campaign, plus converter exclusions. Audience layering beyond that should have to prove itself against the broad baseline.
Publisher targeting: block and adjust, don't hand-pick#
Publisher targeting on Taboola is mostly subtractive. Campaigns start across the network within your geo and device settings, reporting shows per-site performance, and you shape traffic from there. You can block sites from the start, and Realize says an account manager can apply an approved list of sites, but it recommends that new campaigns leave blocking clear so they gather site-level data first:
Let spend accumulate per site. Native publisher traffic is long-tail; most sites will send you a handful of clicks. Resist blocking on tiny samples. Act decisively where data justifies it. A site that has consumed several conversions' worth of clicks with nothing to show for it goes on the blocklist; Realize's own tip is to lower the bid on an underperforming source before removing it entirely. This weekly loop of reviewing the site report and cutting the bleeders is a core Taboola optimization task. Bid-adjust the middle. On Fixed Bid and Enhanced CPC campaigns, sites that convert at marginal cost can get a downward bid adjustment instead of a block, and proven converters an upward one. On Maximize Conversions the algorithm adjusts by site itself, and you control sites with block lists. Graduate to approved-site campaigns later. Once you know your best publishers, a separate campaign limited to them (set up through your account manager) can concentrate budget where you have proof. Run it alongside, not instead of, discovery campaigns.
One caution: aggressive early blocking is an easy self-inflicted wound. Block thirty sites in week one on statistical noise and the campaign may struggle to spend; Realize also notes that a long block list can lengthen the learning phase.
SmartBid, Maximize Conversions and fixed bids#
SmartBid is Realize's core bidding algorithm. It powers two strategies: Maximize Conversions, which sets your bids automatically within your budget (with an optional target CPA), and Enhanced CPC, which moves a baseline bid you set up or down according to each impression's predicted value. Fixed Bid leaves the CPC, or a CPM, entirely in your hands, and a separate Maximize Value strategy optimizes for revenue. Realize recommends Maximize Conversions, including for new advertisers; check the current documentation, because the lineup evolves.
When automated bidding works, it is because it can price individual impressions faster than a human can. But it is only as good as the conversion signal feeding it, which gives you a clean decision rule:
Realize's default is Maximize Conversions from launch. It can run before your pixel is live, using Taboola's network data to optimize toward page views and engagement, and shifts to your own conversion events once tracking is in place. The manual alternative: start with Fixed Bid or Enhanced CPC if you want hard cost control, run an arbitrage-style flow where you know your click value precisely, or need by-site bid adjustments and day-and-hour bid modifiers, which only the manual strategies support. Add a target CPA only after at least 50 conversions in seven consecutive days, as Realize advises. Don't thrash. Every bid-strategy change restarts learning. Pick a lane, give it enough spend to judge (measured in conversions, not days), then decide.
Automated bidding also interacts with budget: the algorithm needs room to explore, so a starved daily budget produces erratic delivery and slow learning. Realize recommends a Maximize Conversions daily budget of at least 10 to 15 times your expected CPA during learning (or $50 a day if your expected CPA is under $5), and advises against more than doubling the daily cap at once. If you cannot fund that, stay on a manual strategy until you can.
Whichever mode you run, feed it honest events: fire the conversion on the action that matters economically (lead, sale), not on a pageview you can hit cheaply. Automated bidding optimized toward a cheap proxy event will happily buy you thousands of worthless proxy events.
Day-parting and scheduling#
Realize supports scheduling campaigns by day and hour, with a per-campaign timezone, and on manual bid strategies it also lets you raise or lower bids by day and hour. Treat it like OS targeting — reactive, not structural. Run full weeks first; if reporting shows a consistent dead zone (common for B2B offers overnight, or call-based offers outside call-center hours), carve it out. Check which timezone your account reports in before setting schedules, and remember that day-parting shrinks scale — every hour you exclude is auctions you don't enter. For call-driven verticals it's essential; for most ecommerce and lead-gen it's a late-stage refinement.
A campaign structure that uses targeting properly#
Pulling it together, the standard account shape looks like a matrix of geo × device prospecting campaigns, plus a retargeting layer:
| Campaign | Geo | Platform | Bidding | Role |
|---|---|---|---|---|
| US-DESK-prospecting | US | Desktop | Maximize Conversions, or manual bids while pruning | Discovery + scale |
| US-MOB-prospecting | US | Smartphone | Maximize Conversions, or manual bids while pruning | Discovery + scale |
| US-ALL-retargeting | US | All | Maximize Conversions | Harvest warm traffic |
| UK-DESK-prospecting | UK | Desktop | Maximize Conversions, or manual bids while pruning | Geo expansion |
Launch sequence that respects how the network actually behaves:
Pixel installed and verified, conversion events firing on the economically meaningful action. One geo, split by device after an initial broad read, broad publishers, and either Maximize Conversions with a daily budget of 10 or more times your expected CPA or a manual bid reasoned from the value of a click. Weekly publisher review: block bleeders, bid-adjust the middle where your strategy allows it. A target CPA once you have 50 conversions in seven days; approved-site campaigns once you have a proven-site list. Expand geo by cloning winners with localized creative.
The anti-pattern is stacking every lever on day one — narrow geo, narrow audience segment, day-parting, and a hand-tuned site list on a fresh account. Each restriction shrinks the auction pool and raises your effective cost before you have any data proving the restriction earns its keep. Full setup mechanics, screenshots and account walkthrough live in how to advertise on Taboola.
The targeting mistakes that burn first budgets#
Common ways first budgets get burned, all avoidable:
Launching worldwide "to see what works." Cheap clicks in low-cost markets can absorb the budget before you learn anything. Start with one geo you understand. Stacking every restriction on day one. Narrow geo plus a marketplace segment plus day-parting plus a hand-built site list, on an account with zero data. Each layer shrinks the auction pool and raises effective cost before anything has proven its value; Realize's own set-up guide recommends starting broad. Blocking publishers off ten-click samples. Native traffic is long-tail; statistical noise looks like signal for the first week. Set a spend threshold per site before you judge it. Expecting conversion optimization without conversion data. Without the pixel or server-to-server events, Maximize Conversions optimizes toward page views and engagement using network data, not toward your leads or sales, and a buyer can wrongly conclude that automated bidding doesn't work. Optimizing targeting before creative. If CTR is poor, no geo or device split rescues the campaign; on native the creative does much of the targeting, and fixing it often moves results more than any settings change. Ignoring language. English creative in non-English geos is likely to depress CTR, and with it delivery and effective cost across the campaign.
Reverse-engineer competitors' targeting before you spend#
Every targeting decision above gets easier when you can see what incumbents in your niche already run. OpenAdLibrary, which publishes this guide, records the geo and device each captured Taboola ad was seen on, its first and last observation dates and the traced landing page behind it. On the Founders plan ($29.99 a month) you can search and filter those captures on /spy/taboola; the free plan shows the 20 newest ads. Capture context is not the advertiser's campaign settings: seeing an ad in the UK on mobile tells you it ran there, not how it was targeted, bid or budgeted. A creative observed for 30 or more days is a candidate to study, not proof that its geo, device and creative combination pays; brand budgets, testing, slow optimization and capture gaps can all keep an ad visible. The full research workflow is in the Taboola ad spy guide.





