Meta Value Rules vs. Age Restrictions: We Tested Both

If you want to age-target on Meta, you have two levers: value rules, which bid more or less for specific segments while delivery still sees everyone, and detailed targeting restrictions, which wall segments off entirely. We ran both head-to-head against an unconstrained control for three weeks, same creatives, same budgets, and the answer was not close.

The one-line result: value rules doubled delivery to our target age at zero efficiency cost. The hard age restriction bought the exact same audiences at roughly 2x the price, and got worse every week it ran.

TL;DR

• We ran a 3-arm bidding test for a consumer iOS subscription app whose audience skews women 45+: an unconstrained control, a value rules arm (bid +5% on 65+, -20% on 45 to 54), and a hard age restriction arm (women 55+ only). Same 5 creatives, equal budgets, subscribe-optimized, AEM 7-day click attribution.
• Value rules tied the control exactly: $19.06 vs $19.27 per trial on identical $1,696 spend, while doubling delivery to the target 65+ age group ($448 to $938 of spend) at no efficiency cost.
• The hard restriction paid roughly double for the same people: $30 to $39 per trial vs the $18 to $23 both other arms paid for 55+ inventory. It deteriorated as it spent ($23.60 to $51.09 per trial week over week) and was shut down after two weeks.
• The rule: steer with value rules, never restrict. Restrictions forfeit the auction efficiency Meta's delivery system finds on its own.
• We then consolidated into the control anyway. The decision framework for shipping a control on a null result is the second half of this post.

The setup

• Advertiser: consumer iOS subscription app, audience skews women 45+. The paid goal is trial starts (in-app subscribe event), with roughly 1 in 3 trials converting to paid downstream.
• Why the test existed: historical account data suggested 65+ audiences started trials far cheaper than 45 to 54 (roughly $10 vs $21 per trial). If that held, age steering should cut acquisition cost. Meta gives you two mechanisms: value rules (bid multipliers by segment) or detailed targeting restriction (exclude ages outright). Nobody should assume they behave the same, so we tested both.
• Structure: one campaign, three ad sets cloned from the same AEM source with 7-day click attribution preserved, the same 5 creatives in each, equal budgets of about $95/day per arm, subscribe-optimized, iOS only, three consecutive weeks in mid-2026.
  • C, control: no rules, ages 35+ open.
  • T1, value rules: +5% bid on 65+, -20% bid on 45 to 54, attached as a value rule set on the ad set.
  • T2, restriction: targeting narrowed to women 55+.

Results

Lifetime, by arm

Arm	Spend	Trials	Cost per trial
C, control	$1,696	88	$19.27
T1, value rules	$1,696	89	$19.06
T2, restriction (shut down week 2)	$767	23	$33.33

Weekly cost per trial (trials in parentheses)

Week	Control	Value rules	Restriction
Week 1	$19.82 (15)	$17.13 (18)	$23.60 (13)
Week 2	$17.74 (41)	$18.94 (38)	$51.09 (9)
Week 3	$20.97 (32)	$20.24 (33)	shut down

The restriction did not just lose. It deteriorated as the system was forced deeper into a narrowed audience: cost per trial more than doubled from week 1 to week 2 while spending more money for fewer trials.

The money table: age group by arm (lifetime spend, cost per trial)

Age group	Control	Value rules	Restriction (55+ only)
Under 45	$109 at $27	$87 at $22	excluded
45 to 54	$496 at $17	$57 at $57	excluded
55 to 64	$642 at $23	$613 at $19	$457 at $30
65+	$448 at $17	$938 at $18	$310 at $39
Total	$1,696 at $19.27	$1,696 at $19.06	$767 at $33.33

Three things to read off this table:

1. Value rules moved the mix, not the cost. 65+ spend doubled from $448 to $938 while holding $18 per trial. Spend on 45 to 54 collapsed about 90%, from $496 to $57. Total efficiency was unchanged.
2. The restriction bought the same ages at about 2x. In the 55 to 64 group: $30 per trial vs $19 in the value rules arm, a 58% premium. In 65+: $39 vs $18, a 117% premium. Identical audiences, identical creatives. The only difference was how the audience was reached: steered vs walled.
3. The original premise died in the data. In the unconstrained control, 65+ and 45 to 54 both ran about $17 per trial. The "65+ is much cheaper" gap from historical data did not reproduce. The value rules were amplifying an assumption the test itself disproved.

Why restrictions cost double

Why do restricted audiences cost more on Facebook? The mechanism matters because it generalizes beyond age targeting:

• Auction freedom is the algorithm's edge. An unconstrained ad set finds the cheapest convertible impressions across the whole audience. A restricted ad set must fill its budget from a fixed pool regardless of price. You are forcing the buyer to shop in one aisle.
• It was not saturation. Frequency ran 1.1 to 1.3 across all arms, barely one impression per person over three weeks, with reach of 55k to 60k per arm against a national audience.
• It was not inventory price alone. CPMs for 65+ run only about 15 to 25% above 45 to 54 ($24 to $26 vs $20). The restriction's penalty of 58 to 117% is far larger than the CPM gap. The loss is conversion efficiency, not media price.

This matches what Meta's own delivery documentation has implied for years and what broad targeting advocates have argued: constraints on delivery are a tax, and the tax compounds as the system exhausts the cheap corners of a small pool.

Why we called control vs value rules a tie

Trials are Poisson counts on budget-balanced arms, so the significance of arm-vs-control is approximately

$z \approx \frac{N_a - N_c}{\sqrt{N_a + N_c}}$

Control vs value rules was 88 vs 89 trials, so $z \approx 0.1\sigma$. Nothing. Resolving a 1% CPA gap at this volume would take tens of thousands of trials, which means this test could never call that winner at this budget. The restriction, by contrast, cleared $2\sigma$ against the field within two weeks. Real losers announce themselves fast. Winners usually do not.

A note on method: we deliberately did not run a two-proportion test on subscribes per click. The denominator (impressions, clicks) is an endogenous auction outcome that the bid rules themselves change. Budget is what was held fixed, so Poisson-on-spend is the model that matches the objective.

The second half: shipping the null

With the restriction dead and control tied with value rules, the quieter decision was which arm to consolidate into. The point estimate favored value rules: ahead in 3 of 4 measurement windows, $0.21 per trial better lifetime. We consolidated into the control anyway, on a pre-registered decision rule:

1. A treatment must beat control to be adopted. A 1% edge at p around 0.3 is exactly the kind of directional winner that regression to the mean erases. Adopt those habitually and your account accumulates junk configuration.
2. The treatment's mechanism lacked its prerequisite. Value rules exist to encode known differential customer value. Our value assumption (65+ converts cheaper and is worth more) failed inside the test, and downstream trial-to-paid data by age did not exist yet.
3. The costs were asymmetric. Keeping the rules meant a live bid-down aimed at what the control revealed to be the cheapest large cell (45 to 54 at $17 per trial), 55% of spend concentrated in one age cell on an unproven bet, and configuration a future operator has to know exists.
4. Fragmentation had a concrete price. Both arms were running about 32 trials per week, under Meta's roughly 50 conversions per week learning threshold. Consolidated, the account runs about 65 per week and exits learning cleanly. If you are scaling Facebook campaigns, consolidation below the learning threshold is a win independent of everything else.

The flip condition was written down at decision time: if subscription data later shows certain ages convert trial-to-paid at a meaningfully higher rate, re-attach the saved value rule set. Meta now supports attaching and detaching value rule sets without a learning reset in most cases. More noisy CPA weeks do not reopen the decision.

Takeaways

1. Never age-restrict on Meta when you can age-steer. Restrictions bought identical audiences at about 2x in our test, and got worse as they spent.
2. Value rules are a steering wheel, not a discount. They redistributed spend dramatically at zero efficiency cost, but delivered no CPA gain, because they amplify your value assumption. Ours was wrong.
3. Check the premise inside the test. The unconstrained control is itself the measurement of your value assumption. Ours showed all ages converging to roughly the same cost per trial.
4. Losers separate fast. Winners usually do not. The restriction was $2\sigma$ bad in two weeks. The control-vs-rules winner was mathematically unresolvable. Budget your patience accordingly.
5. Ship the control on a null. Point estimates flatter treatments. Decision rules with pre-registered flip conditions beat re-litigating every week.
6. Consolidate below the learning threshold. Two ad sets at 32 conversions per week are worse than one at 65, independent of everything above.

FAQ

Should I restrict age targeting on Facebook ads?

Almost never, if your goal is efficiency. In our controlled test, a hard age restriction paid 58 to 117% more per conversion for the same age groups an unconstrained ad set reached on its own. If you need to shift delivery toward an age group, use value rules to bid up that segment instead of excluding everyone else. Legal or policy reasons to restrict (alcohol, gambling, housing and other special ad categories) are a different matter.

What are Meta value rules?

Value rules are bid multipliers you attach to an ad set that tell Meta's delivery system to bid a percentage more or less for specific segments, for example +5% for people 65+ or -20% for ages 45 to 54. Delivery still sees the whole audience, so the auction keeps its full freedom. Restrictions, by contrast, remove segments from delivery entirely.

Are value rules the same as value optimization on Facebook?

No. Value optimization (value-based bidding) changes what Meta optimizes for: it targets the highest purchase value rather than the most conversions, using the values you send with events. Value rules change how much Meta bids for specific segments you name (age, gender, region) while the optimization goal stays the same. You can run either without the other. Our test used value rules only.

Do value rules hurt performance?

In our test, no. The value rules arm matched the unconstrained control at $19.06 vs $19.27 per trial on identical spend, while doubling delivery to the target age group. But they also did not improve CPA, because value rules amplify your assumption about which segments are worth more. If that assumption is wrong, you are steering toward nothing.

Why did my restricted ad set get more expensive over time?

A restricted ad set must fill its budget from a fixed pool of people regardless of price. It buys the cheap conversions in that pool first, then is forced into progressively more expensive inventory. Ours went from $23.60 to $51.09 per trial in two weeks. An unconstrained ad set can keep hunting for cheap conversions across the entire audience instead.

How long should I run a Meta bidding test before calling a winner?

Losers separate quickly: our restriction arm was roughly $2\sigma$ worse within two weeks. Ties and small winners often cannot be resolved at small-account volume at all. A useful rule: compare conversion counts with a Poisson check, $z \approx (N_a - N_c) / \sqrt{N_a + N_c}$, and pre-register the decision rule (a treatment must clearly beat control to be adopted) before you start.

---

Running paid acquisition and want tests structured like this, with decision rules written before the money is spent? That is how Leapwave Ads runs accounts. We manage the auction, you keep the margin an agency would have taken.