Estimated Action Rate: The Meta Ads Setting That Quietly Decides What Your Leads Cost

Here is the short answer: estimated action rate is Meta’s prediction of how likely a specific person is to take the action you told the campaign to optimise for, and it is weighed alongside your bid every time your ad enters an auction. Meta’s own documentation says the auction winner is decided by total value, which combines the advertiser’s bid, estimated action rates and ad quality. So the event you choose to optimise for is not a reporting preference. It changes who Meta predicts will act, which changes which auctions you win and what you pay to win them. Choose a shallow event like a raw form fill and Meta becomes brilliant at finding people who fill in forms. Choose a qualifying event and it hunts for people who behave like buyers.

Almost everything written about this is written for ecommerce, in the language of add to cart and initiate checkout. If you run a service business in New Zealand, the trades, professional services, health, home improvement, the events look different but the mechanism is identical, and it is arguably the single highest-leverage decision in your whole account. This post explains the mechanism and then works through the decision properly for lead generation.

What estimated action rate actually is

Every time someone opens Facebook or Instagram, an auction runs to decide which ads they see. Meta’s explanation of the ad auction is that the winner is the ad with the highest total value, made up of three things:

1. Your bid. What you are willing to pay, whether you set it manually or let Meta bid for you.
2. Estimated action rate. Meta’s prediction of how likely this particular person is to take the action your campaign is optimising for.
3. Ad quality. Signals about the ad itself, including feedback from people who see it and low-quality attributes like engagement bait or withheld information.

Meta states that estimated action rates and ad quality together measure ad relevance, and that a relevant ad can beat a competitor with a higher bid. That sentence gets quoted a lot. The part that gets skipped is what “action” means in “estimated action rate”. It means the specific optimisation event you selected when you built the ad set. Not conversions in general. Not good outcomes in the abstract. The one event you pointed the campaign at.

So the prediction being weighed alongside your bid, thousands of times a day, is a prediction about one behaviour that you chose. Two identical campaigns with identical creative and identical budgets, one optimised for a form submission and one optimised for a booked call, are not running the same auction strategy. They are asking Meta two different questions about every person on the platform, and they will be shown to different people at different prices as a result.

Why the optimisation event decides your cost, not just your reporting

Follow the mechanism through and the consequences fall out on their own.

It decides who sees your ad. Meta scores people on their predicted likelihood of taking your chosen action. People with high predicted likelihood generate high estimated action rates, which generate high total value, which wins auctions. Your ad flows towards whoever is most likely to do the specific thing you asked for.

It decides what you pay. When your estimated action rate on a given person is high, you need less bid to win that auction. When it is low, you need more. Optimise for an event your best prospects rarely trigger and you are paying a premium in every auction to reach them, because the model does not rate them as likely to act.

It decides what your account learns. Every completed event becomes training data for the next round of predictions. The system compounds in whatever direction you pointed it. This is the same recency-weighted learning we covered in our post on the Meta Pixel’s 180-day window: delivery reflects who has been converting recently, and the optimisation event defines what “converting” means.

The uncomfortable version of this for lead generation: there is a population of people on Meta who are unusually likely to fill in forms. Some are your buyers. Plenty are bored scrollers, quote collectors, people burning time in a queue, and bots that have learnt to look human. If your optimisation event is a raw form submission, every one of those completions is a success as far as the algorithm is concerned, and the model dutifully goes and finds more people like them. We see the end state of this regularly in accounts that come to us: a campaign that has spent months optimising for lead volume, a CPL the owner is quite proud of, and a sales team that has quietly stopped ringing the leads because so few of them answer. The ad account looks healthy. The business is not getting customers from it. Nothing is malfunctioning, which is exactly the problem. The machine was pointed at the wrong target and hit it.

The lead-gen event ladder

Ecommerce advertisers get a ready-made ladder of events with obvious commercial weight: view content, add to cart, initiate checkout, purchase. Service businesses have an equivalent ladder, but nobody hands it to you. You have to build it.

Rung 1: the raw form submission. Someone completes your lead form or website contact form. Highest volume, cheapest per event, weakest correlation with revenue. This is the default most accounts sit on forever.

Rung 2: the qualified enquiry. A submission that passed some bar: answered qualifying questions sensibly, matched your service area, gave a real phone number, was not caught by your spam filtering. Fewer events, each one worth more as signal.

Rung 3: the booked call or appointment. They picked a time in your calendar. Dramatically stronger intent, because booking a call costs the prospect something a form fill does not: a commitment to actually show up and talk.

Rung 4: the closed or qualified job. They became a customer, or your team spoke to them and confirmed they were real, in budget and ready. This is the event that actually pays your invoices.

Each rung up the ladder means fewer events, and each event carries more information about what a buyer looks like. That trade-off is the whole game, because of a constraint Meta is explicit about.

The 50-events constraint, and why you cannot just optimise for revenue

Meta’s published guidance is that an ad set needs roughly 50 optimisation events per week to exit the learning phase and deliver predictably. Below that, the model never gets enough data to stabilise, delivery stays erratic, and costs usually run higher than they should.

This is why “just optimise for closed jobs” is bad advice for most NZ service businesses, even though rung 4 is the truest signal. If you close eight jobs a month from Meta, an ad set optimised on closed jobs will sit in the learning phase permanently. You would be feeding the model the best possible signal at a volume too low for it to learn from.

So the real decision is: the highest rung on the ladder that can still produce somewhere near 50 events a week at your budget. That is the event where signal quality and signal volume balance.

In practice, for the local service businesses we work with, that usually lands on rung 2 or rung 3. A qualified enquiry or a booked call is a meaningfully better predictor of revenue than a raw form fill, and a business spending enough to generate leads daily can often still clear, or get close to, the volume the learning phase wants. Rung 1 is where you start when the account is new and budget is small, not where you settle. And the closer you can operate to the qualifying event, the more every other quality lever works with you instead of against you. The form friction and screening tactics in our guide to reducing spam leads do double duty here: they filter junk out of your inbox, and they stop junk being counted as success by the algorithm.

Two practical notes from running this across accounts:

Expect CPL to rise before quality shows. Move a campaign from optimising on form fills to optimising on a qualifying event and the cost per lead in Ads Manager will usually go up. It has to: you are now asking for a rarer, more valuable behaviour, and the model is temporarily short of data while it relearns. The number that matters is cost per qualified lead, and then cost per customer. Judge the change on those, over weeks rather than days, or you will retreat to the cheap metric at exactly the wrong moment.

Consolidation matters more, not less. Because qualifying events are scarcer, fragmenting them across many ad sets is more damaging than fragmenting form fills. Pooling signal into fewer campaigns is the same logic we walk through in how many Facebook ad campaigns you should run, and it becomes close to mandatory once you optimise higher up the ladder.

Making the higher rungs technically possible

You cannot optimise for an event Meta cannot see. The default tracking setup on most service-business websites reports a form submission and nothing else, which is why most accounts are stuck on rung 1 by default rather than by decision.

Getting to rung 2 and 3 usually means some combination of:

• Firing distinct events for distinct outcomes. A generic “Lead” event on every form is not enough. A booking confirmation page or calendar-booking event should fire its own conversion, separate from the raw enquiry, so you can point an ad set at it.
• Running the Conversions API alongside the Pixel. Server-side events close the gaps that browsers and ad blockers open, and they are how events that happen after the click, like a booking confirmed in your calendar system, get back to Meta reliably.
• Feeding outcomes back from your CRM. When your team marks a lead as qualified, or a job as won, that can be sent to Meta as an offline or CRM conversion. Even where the volume is too low to optimise on directly, it gives the model examples of what a genuinely good outcome looks like. Wiring this up is a big part of the CRM integration work we do for clients, precisely because the ad account is only ever as smart as the outcomes it gets told about.

One thing to be clear-eyed about: none of this is a substitute for diagnosing quality problems properly. If your leads are poor, the optimisation event is one suspect among several, alongside creative, offer and follow-up speed. Our lead quality diagnostic guide walks the full list. But of all the suspects, the optimisation event is the one that silently shapes every auction you enter, which is why it is the first thing we check in a Meta Ads account review.

The decision, condensed

If you take one thing from this post, take the framing: choosing your optimisation event is choosing the question Meta asks about every person it considers showing your ad to.

Optimise for form fills and the question is “how likely is this person to fill in a form?” Optimise for booked calls and the question becomes “how likely is this person to put a meeting in a calendar and answer the phone?” Those questions have different answers for the same person, and the auction maths, your bid weighted by estimated action rate and ad quality, converts those different answers directly into different delivery and different costs.

Work up the ladder as far as your event volume allows. Track the technical work needed to make each rung visible to Meta. Judge changes on cost per qualified lead, not cost per lead. And revisit the decision as the account grows, because the right rung at $1,500 a month of spend is often the wrong rung at $6,000.

Frequently asked questions

What is estimated action rate in Facebook ads?

It is Meta’s prediction of how likely a specific person is to take the action your campaign is optimising for, calculated for every person in every ad auction. Meta combines it with your bid and your ad quality to produce total value, and the ad with the highest total value wins the auction. A higher estimated action rate means you can win auctions with a lower bid, so it directly affects what you pay.

Can I see my estimated action rate anywhere in Ads Manager?

No. It is calculated per person, per auction, and Meta does not expose it as a metric. You influence it rather than read it: through the optimisation event you choose, the conversion data you feed back, and creative that attracts the behaviour you optimised for. Falling costs and improving lead quality at stable budgets are the visible evidence it is working in your favour.

Should I optimise for leads or for a deeper event like booked calls?

Optimise for the deepest event that can still generate somewhere near 50 events per week, which is Meta’s published guidance for exiting the learning phase. For many NZ service businesses that is a qualified enquiry or a booked call rather than a raw form fill. If your budget only produces a handful of deep events a week, stay one rung shallower and tighten your form qualification instead, then move up as volume grows.

Why did my cost per lead go up when I changed my optimisation event?

Because you asked for a rarer behaviour. A booked call will always cost more per event than a form fill, and the ad set also re-enters the learning phase after the change, which makes delivery less efficient for a period. Judge the change on cost per qualified lead and cost per customer over a few weeks. In our experience the shallow CPL rises while the numbers that reach your bank account improve.

Does estimated action rate affect lead quality or just cost?

Both, through the same mechanism. The estimated action rate is a prediction about your chosen event, so it steers delivery towards people likely to trigger that event. If the event is loosely connected to buying, delivery drifts towards people who complete it without buying, and quality falls. If the event is a genuine qualifying step, delivery is steered towards people who behave like buyers. Cost and quality are two outputs of the one targeting decision.

Point the machine at the right target

Meta’s delivery system is very good at finding more of whatever you tell it success looks like. The estimated action rate is the mechanism that does the finding, auction by auction, and your optimisation event is the instruction it follows. Most underperforming lead-gen accounts we open have world-class machinery pointed at a target nobody consciously chose.

If you would rather have it pointed at qualified enquiries and booked calls, that is the exact wiring, event structure, Conversions API, CRM feedback and campaign consolidation, that we set up for New Zealand service businesses. Book a call with Lucid Leads and we will look at what your account is currently optimising for and what it should be.