How much should a South African estate agency spend on Meta ads?

For most South African residential agencies advertising one office in one local market, a workable Meta budget starts at R2,000 to R3,000 a month for brand visibility, or R3,000 to R5,000 a month where the campaign is asking homeowners for a valuation or seller enquiry. Listing promotion is budgeted per property rather than per month. Those are working levels, not recommended ceilings, and they are separate from any management fee. The more useful question is not what the number is, but what makes it move.

Three costs that should never be added together

The first thing to get right is not the amount. It is the accounting.

Agency advertising involves three distinct costs, and most quotes blur at least two of them:

When those three are quoted as one figure, the agency loses the ability to answer the question that matters at review time: is this underperforming because the work is weak, or because the media budget was never enough to test it properly? A blended number makes that undiagnosable.

The billing model also sets the incentive. Where a supplier's fee is a percentage of platform spend, the supplier earns more as the budget rises. That model is normal in larger media buying and plenty of firms using it do good work, so this is not an accusation. It does mean that a recommendation to increase spend comes from a party who benefits from it, and a principal should know which model they are buying before that conversation arrives. Realead charges a fixed monthly fee and takes no percentage of ad spend.

The account arrangement matters for the same reason. Spend is paid directly to Meta from the agency's own account, and the agency keeps ownership and administrative control of the Meta business assets involved: the ad account, the dataset that records pixel activity and conversions, the audiences built from it, and the campaign and spend history. A supplier is granted partner access to work in those assets, and that access can be withdrawn. The pricing page sets out all three costs separately.

Everything below is about the third number only.

Why there is no universal number

There is no correct budget for an estate agency, because "an estate agency" is not a media planning unit. A four-agent office working two suburbs and a thirty-agent office working a metro are the same business type and completely different advertising problems.

Note what does not change the number: how many agents you have. This runs against how a lot of property marketing is sold, so it is worth being blunt about. Agency advertising buys attention in a geographic market. The market does not get bigger because you hired three more agents, and it does not shrink when someone leaves. Realead prices Advertising for the agency, its office and its agreed local market, and agent count never moves that price. If a supplier's advertising quote scales with headcount, ask what is being bought per agent, because it is not market coverage.

What does change the number is the size of the audience you need to reach, what you are asking that audience to do, how long you intend to keep asking, and how much data the campaign needs before it settles.

The four things that actually set the number

1. The objective

Asking a stranger to remember your agency is cheap. Asking a homeowner to submit their address and phone number for a valuation is not. The action you optimise toward sets the floor, because rarer actions cost more to generate.

This is why brand visibility and seller enquiries carry different working levels. They are not the same product at different intensities. They are different asks.

2. The size of the local market

South African agency markets are usually measured in suburbs, not cities. That is a strength for relevance and a constraint on budget mechanics. A tightly drawn audience around three suburbs may only be a few tens of thousands of adults, and a large share of them are not property-relevant in any given quarter.

Meta's guidance cautions against drawing audiences too narrowly, noting that overly narrow targeting can increase costs and lead to creative fatigue. A small audience shown the same creative repeatedly gets expensive, and the honest response is usually to widen the geography slightly or refresh the creative more often, not to keep pushing budget into a shrinking pool.

3. How long the campaign runs

A budget is a rate, not a total. R3,000 a month for six months and R18,000 in a single month are the same money and completely different campaigns. The first builds familiarity that survives between listings. The second buys a spike the market forgets.

Property decisions do not arrive on a schedule you control. A seller moves when a bond changes, a job relocates, a family grows or a neighbour gets a good price. None of those are events an agency can time, which is the argument for continuity over intensity: you are covering a window you cannot predict.

4. How much data the campaign needs

Meta's delivery improves as it accumulates results, and until it has enough, an ad set sits in what Meta calls the learning phase. For ad sets optimising toward a specific event, such as a lead submission, Meta's guidance is to structure the ad set so it can accumulate roughly fifty of those events within a seven-day window, which is where delivery tends to stabilise. Ad sets that stay below it may be reported as learning limited.

Three qualifications matter more to a small agency than the number itself.

It applies to event-optimised ad sets. A campaign optimising for reach or impressions is not measured against that threshold. It is not a universal requirement for every Meta campaign, and it should not be quoted as one.

The window is rolling. It is fifty events in any seven days, not fifty events eventually. A campaign does not leave that state simply by running longer at the same budget, so any advice implying that patience alone resolves it has misread the mechanic.

Learning limited is a status, not a verdict. It describes what the delivery system knows about an ad set. It says nothing about whether the advertising is doing anything for the business. A small suburban office optimising toward valuation enquiries may sit in that state more or less permanently, which is an ordinary condition for a small local advertiser rather than a fault. What it genuinely tells you is that results will be less stable month to month, so the campaign should be judged over a quarter rather than a fortnight. Treat it as a prompt to review the objective, the volume of events available, the ad set structure, the audience and the budget. It does not by itself prove the campaign has failed, and it does not make increasing spend the automatically correct response.

The broader guidance Meta gives around learning is more useful to act on than the threshold: give an ad set enough budget and at least seven days before judging it, consolidate ad sets with similar audiences rather than proliferating them, and avoid unnecessary edits, because significant changes restart learning.

Why a split budget is the more common mistake

Underfunding gets discussed. Fragmentation does not, and it is the more frequent error.

Meta's documentation recommends consolidating ad sets with similar targeting, and warns that too many ad sets with overlapping targets can cannibalise data collection and hurt each one's ability to learn. Every additional ad set divides the same pool of results into smaller piles, and each smaller pile takes longer to reach the point where delivery is optimised. Three ad sets at a third of the budget do not each run at a third of the quality. They each run worse than that.

For an agency this shows up in a recognisable way. The principal, reasonably, wants brand presence, seller leads, a listing pushed and a top agent supported. Four sensible wants. Funded from one modest budget, all four underperform, the campaign gets judged a failure, and the conclusion drawn is that Meta does not work for property. The budget was not necessarily too small. It was too divided.

This is why Realead's standard Advertising scope runs up to two campaign objectives live at once rather than the full set. Two funded objectives beat four starved ones, and the discipline of choosing is part of what the plan is for. The advertising service page sets out the four standard objectives and how the choice is made.

The practical rule: if your total budget cannot fund one objective at its working level, run one objective. Do not run two at half strength.

Working spend levels

These are the platform spend levels Realead publishes on the pricing page, reproduced without adjustment. They describe where the work tends to function, not what it will return.

Two cautions about numbers of this kind, including these.

Auction costs move. Cost per thousand impressions and cost per click in South Africa shift with season, competition, audience and creative, and they move again when a large advertiser enters a market. Any figure of that sort quoted as a permanent benchmark should be treated as marketing rather than measurement, which is why the table is expressed as working budget levels rather than expected costs per result.

And a working level is not a promise. There is no lead count, no cost per lead, no mandate count and no return figure attached to any row above, because none of those can be honestly attached in advance.

A decision framework

Four questions, answered in order, because each answer constrains the next.

1. What single commercial job is this money doing over the next quarter? If you cannot choose, the answer is usually that the agency needs recognition in its market before it needs enquiries, because a valuation request from an unfamiliar name is a harder ask.

2. What is the smallest geography that still contains enough people? Start from the suburbs you genuinely work and can service properly. If that audience is very small, widen the boundary before you increase the budget. Widening costs nothing. Over-spending into a tiny audience costs a great deal.

3. Can I fund the working level for that objective, continuously, for at least three months? If yes, that is the budget. If no, apply the rule above and drop to the objective you can fund properly. Three months is roughly the shortest window in which continuity, learning and a full reporting cycle can all occur.

4. What am I prepared to judge this on at the end of the quarter? Agree it before the first rand is spent. Reach and frequency in the target area, enquiry volume and enquiry quality, cost per result and its direction of travel, and what changed month to month. Agreeing the measure afterwards is how good campaigns get cancelled and poor ones get renewed.

A worked illustration

The following is hypothetical, used only to show how the questions interact. It is not a client, a case study or a result.

Consider a hypothetical six-agent independent office working three adjoining suburbs, with an owner who wants more seller mandates and about R4,500 a month available for platform spend.

The primary job is seller enquiry. The geography is already tight, so it stays as drawn rather than being narrowed further by demographic filters. R4,500 sits inside the published seller-lead level, so one funded seller objective is affordable while funding a second properly at the same time is not.

Change one variable and the answer changes. Drop the available spend to R2,500 and the correct decision is not a smaller seller campaign. It is a brand visibility campaign at its own working level, because a properly funded easier ask is more likely to build something useful for that office than an underfunded harder one.

What the budget does not buy

Platform spend buys delivery. It does not buy conversion. An enquiry that arrives and sits unanswered for two days has cost the same as one answered in ten minutes and is worth considerably less. Response speed, local knowledge, pricing advice and follow-up decide what happens next, and none of those are things a budget can purchase.

It also does not buy certainty. Any supplier willing to quote a fixed monthly lead count in advance is guessing in public. Realead does not quote one, which is stated plainly on the advertising page and in the FAQ.