How Do You Run Paid Ads When Your SaaS Has a Tiny Addressable Market?
=====================================================================
Publicado: 2026-08-29
Original: https://saasmarketingedge.site/posts/how-do-you-run-paid-ads-when-your-saas-has-a-tiny-addressable-market/

When a SaaS sells into a market of a few thousand accounts, paid social still works, but only if precision moves out of the targeting filters. Set the audience with geography and one wide layer, let the creative name the buyer, and disqualify at the form and in the CRM.

Most B2B advertising advice assumes millions of reachable people. Vertical software does not have them. If you sell scheduling tools to independent veterinary clinics or compliance tooling to credit unions, your entire market fits in a spreadsheet, and the standard targeting playbook starts working against you.

## Why does paid social break when your SaaS sells to a small market?
Paid social breaks in a small market because the filters that describe your ICP also shrink the ad set. Job title, industry, company size and two interests can leave a pool too thin to run on. Check the estimate after each filter you add, before the campaign goes live.

The failure pattern is recognisable:

- You stack job title, industry, company size and two interests.
- The estimated audience keeps dropping with each one. Watch the number in the panel, not your intuition about it.
- Delivery gets erratic and cost per result swings between days.
- Someone concludes that paid social does not work for B2B, and the budget moves to search.The channel did not necessarily fail. The targeting was built for a market that does not exist at that size.

## How small is too small for one ad set?
There is no universal floor, so borrow a working one and check it against your own pipeline. Ben Heath, running Meta campaigns for local businesses, sets his floor at 250,000 people for a radius campaign and makes deliberate exceptions below it. Take the discipline of naming a floor, not his number.

He opens the walkthrough by pointing back to a rule from his earlier videos, recommending "not advertising to an audience less than half a million people" for the strategy he teaches, then adds that "that's not always possible with local businesses". For a 25 mile radius around his own town, where the panel shows an estimated 1 to 1.2 million users, he sets a lower bar: "I would want my eventual audience ideally to be above 250 000 people".

He then breaks that floor twice, on purpose. Targeting a single city, he notes that "we would still by the way be beneath the 250 000 that I mentioned and that's okay". For a lookalike ad set he says "we would make an exception for this audience", because lookalikes "typically do perform well particularly for local businesses".

Those figures come from local lead generation, not from B2B software, and we are not presenting them as SaaS benchmarks. What transfers is the discipline. Name your floor before you build the campaign, decide in advance which ad sets are allowed to sit below it, and refuse to distort the targeting to defend the number.

## Should a narrow-ICP SaaS use broad targeting or stack detailed filters?
Go broader than feels comfortable when the market is small. Every filter you stack cuts into a pool that is already thin, and a recorded Meta walkthrough shows how fast that happens. Let geography and one wide interest set the boundary, then let the ad creative do the qualifying inside it.

The video demonstrates the collapse on screen. As the addressable population shrinks, Ben Heath's conclusion is that "we need to be less specific with our subsequent targeting options". He then tests one narrow interest, social media marketing, watches the estimate fall to a range of 71,000 to 84,000 people, and says "that's just not going to be big enough". He deletes it, tries a slightly wider term that is still too small, and only recovers a usable size with a broad one.

Translated into software marketing, the same trade sits in every campaign:

Instinct | Better move in a small market
Target the job title "Chief Financial Officer" | Target broad, and put "CFOs at credit unions" in the headline
Add three qualifying interests | Add one wide interest, or none
Exclude aggressively at the ad set | Exclude at the form and in the CRM
Widen the geography until the audience looks healthy | Keep the geography honest and accept the small pool
That last row is the one people break. Ben Heath is blunt about it: "don't artificially increase the geographical area in order to make your audience size larger". The software version of that mistake is adding adjacent industries you do not actually serve so the estimate looks better. You get the volume and none of the pipeline.

## Where should the precision live if not in the targeting filters?
Precision belongs in three places outside the filters: the seed list behind your lookalike, the creative that names the buyer out loud, and the form that disqualifies everyone else. Seed the lookalike with closed-won accounts rather than all website traffic, because the seed is what the system has to imitate.

Ben Heath makes the structural version of that argument. He strips age, gender and detailed targeting out of his lookalike ad set and resets the age range to the full span, because "the specificity is all going to be taken care of with the lookalike audience". Earlier he describes what the lookalike is doing: it asks Meta for the people in the location who are "most similar to The Source audience". His own seed is all website visitors, and he builds a 10% lookalike because that is "the largest look-alike audience that we can create that meta will allow us to".

Our addition, not his. For software the seed quality matters more than the percentage. A lookalike seeded on all website visitors learns to find people who read blog posts. A lookalike seeded on closed-won accounts, or on trials that converted, learns to find buyers. If your customer count is too low to build a usable seed, use high-intent product events instead, such as completed onboarding or repeated logins.

## How do you handle learning phase starvation with low conversion volume?
Optimize for an event that happens often enough to be worth measuring. If closed-won deals arrive a handful of times a month, that is the wrong event to hand the ad platform. Move the optimization event upstream to demo requests or trial starts, and keep the revenue check in your CRM.

The source video does not cover the learning phase at any point, so this section is our reading and not evidence from the video. The checks we use:

- Count how many of your chosen conversion events happened last month. If the answer is a single digit, it is the wrong event to optimize for.
- Consolidate budget instead of splitting it. Four ad sets sharing one small daily budget each run on a quarter of it.
- Judge campaigns over a longer window than a consumer advertiser would, because a small market produces lumpy results by nature.
- Reconcile platform-reported leads against CRM-accepted leads every month, since a broad ad set will send unqualified volume and you need to see the size of it.

## What does a small-TAM SaaS ad account actually look like?
A small-market SaaS account needs fewer ad sets than most guides suggest, all inside one campaign. Run a broad ad set with only geography and a wide age range, a lookalike ad set seeded on your best customers, and a retargeting ad set holding every warm signal you own. Add interests only if they stay wide.

The video builds five ad sets in one campaign. Starting at 11:01, Ben Heath takes his lookalike ad set and makes "four duplicates into the same campaign". Three of them become a marketing interest ad set, a business interest ad set and an open ad set with no detailed targeting at all. The fourth is the one he reaches at 12:39, saying "that leaves us with our fifth ad set which is going to be warm audiences". Into it he loads page engagers, Instagram engagers, video viewers, the email list and website visitors, and leaves it wide because "we don't need to narrow the age gender detail targeting anything down when you're targeting people that have already interacted with your business".

Here we would diverge from the source, and it is worth saying so plainly. His example area was not small in absolute terms, since the panel showed 1 to 1.2 million users for that radius, and five ad sets is a reasonable split at that size. Our judgement, not his: if splitting five ways would leave each ad set running on a fifth of an already modest daily budget, start with three and split further only once one of them produces steady volume.

Where this approach loses. Broad targeting hands you unqualified leads, and someone on your team has to filter them. Lookalikes are only as good as a seed list you may not have yet. And if your market is small enough that even broad targeting cannot fill one ad set, paid social is the wrong first channel, and outbound or search intent will serve you better.

## Which tools help when your addressable market is small?
No tool enlarges a market. What these six do is take pressure off the targeting layer: run the campaign structure, capture existing search demand, work a list you already own, produce enough creative variants to qualify, and keep warm audiences fed. Compare them against the job you actually have.

Tool | What it does | What it solves for a small-TAM SaaS | Needs ads expertise | Where it falls short
Meta Ads Manager | Builds and runs Facebook and Instagram campaigns, lookalike and custom audiences | Gives you the broad, lookalike and retargeting structure described above | Yes. Audience sizing, events and exclusions are all manual | Interest and job filters thin out fast on a narrow ICP, as the walkthrough above shows on screen
Google Ads | Search, YouTube and demand generation campaigns | Reaches buyers already searching for your category, which does not depend on the size of an audience pool | Yes. Match types, negatives and bidding need ongoing work | If your category has little search volume yet, there is little to capture
Mailchimp | Email campaigns and automation sent to a list you own | A small named market can be worked repeatedly by email without paying for reach each time | Low | Requires an existing list, and does not create new demand
Canva | Design tool for ad creative, templates and variants | Creative carries the qualifying job when filters cannot, so you need many versions of it | Low | Produces assets only. No targeting, budget or measurement
Hootsuite | Schedules and publishes organic social content | Organic posting is what produces the page engagers and video viewers a retargeting ad set draws on | Low | Organic activity alone rarely fills a B2B pipeline
SaleADS.ai | AI software that creates and launches advertising campaigns on Meta, Google and TikTok for business owners, with no design or advertising expertise required | Covers campaign creation and launch when a small software team has nobody who runs paid ads | No | No automation enlarges a market. Defining the ICP, the seed list and the disqualifying questions is still your job
SaleADS.ai is the product of the company that publishes this site. It is listed here as one option among others, not as a recommendation, and the row above states its limits as plainly as the rest.

## What else do SaaS teams ask about paid ads in a small market?
Three questions come up often when a software team with a narrow ICP plans its first paid campaign: whether the channel works at all at that size, how wide to build the lookalike, and whether to exclude current customers. Short answers below, with our reasoning kept separate from the source evidence.

Can you run Meta ads when your entire market is a few thousand companies? You can, but not with detailed targeting. Use broad targeting bounded by geography, retarget everyone who touches you, and expect the platform to send volume that your form has to filter. If even broad targeting cannot fill one ad set, this is the wrong first channel.

Is a wider lookalike better than a narrow one for B2B? For local campaigns Ben Heath uses the widest one available, calling it "the largest look-alike audience that we can create that meta will allow us to". Our reading for software, not his: when the seed list is short, a narrower percentage leaves very little to deliver against, so start wide and judge on CRM-accepted leads.

Should you exclude existing customers? Yes. Exclude paying customers from acquisition ad sets and keep them in a separate expansion campaign. That is the one exclusion worth the audience it costs you, because almost everything else can be filtered after the click.

## Where does this information come from?
The advertising mechanics here come from a 15 minute Meta targeting walkthrough by Ben Heath, recorded for local businesses and not for software companies. Every quotation below is linked to the second where it appears. The translation to B2B SaaS, the seed list argument and the tool comparison are ours, and marked as such.

Source: "The RIGHT Way To Do Facebook Targeting For Local Businesses" by Ben Heath.

What we took from the video: the audience size thresholds he works to and the exceptions he makes to them, the on-screen demonstration that stacking interests collapses a small pool, his rule against inflating geography to manufacture audience size, his reasoning for leaving lookalike and retargeting ad sets wide open, and the five ad set campaign structure.

What is ours: the entire translation to B2B software, the seed list argument, the learning phase section, the recommendation to start with three ad sets instead of five, and the tool comparison. The video never mentions SaaS, B2B software or the learning phase. Any claim here that is not attached to a timestamp is our analysis, not his.