How to read paid social results
One measurement standard for effectiveness, performance and efficiency.
Platform definitions checked September 2026.
Why a standard
Three vocabularies collide every time paid social results are presented. Platform vocabulary talks in ThruPlays, hook rates and focused views. Media vocabulary talks in CPMs, reach and viewability. Brand vocabulary talks in recall, association and penetration. Most decks use all three on the same slide, and most readers hold only one of them fluently.
The result is predictable. A hold rate gets read as brand impact. A CPM reduction gets read as effectiveness. A "view" gets quoted without anyone saying which of the seven things a view can mean it actually meant. And the word "viewability" gets used for a metric that has nothing to do with the MRC standard, which is the fastest way to lose a media auditor.
This piece is the standard we hold ourselves to, on the website, in case studies, in sales material and in every customer report. Results, the word for the whole, break into three tiers. Effectiveness measures whether the final outcome achieved the intended goal. Performance measures how well the process intended to deliver that outcome was executed. Efficiency measures how well resources were used to deliver it. Every number we present sits in one of the three. Every metric is stated three ways, the technical term, the plain-English name, and what it actually tells you. And every tier lines up against one chain, the one that runs from an impression being bought to a brand being remembered.
The chain everything measures against
Amplified describes the path from exposure to memory in four steps: exposure, attention, encoding, retrieval. An ad is served. Some share of the people it is served to give it attention. Some share of those hold attention long enough for the message to be encoded. And encoding, repeated, is what builds the memory that gets retrieved in a buying moment, which is mental availability, the thing that drives brand growth. Their line for it is that attention is the gateway to mental availability, and without attention there is no encoding, and without memory there is no recall when the purchase decision happens.
That chain is useful because each step has an owner and a metric.
Exposure is what media buys. Impressions, reach, frequency. The media team's half of the job, and the necessary condition for everything downstream. Zero impressions guarantees zero everything.
Attention is what the opening of the ad earns. On a feed, the observable proxy is whether the person stopped scrolling, and the platform's unit for that is a three-second play.
Encoding is what the body of the ad earns. The observable proxy is whether the person who stopped kept watching long enough for a message to land, and the platform's unit is a ThruPlay on Meta or a focused view on TikTok.
Retrieval is mental availability. It is not observable in the ad account at all. It is measured by asking people, in a brand lift study, a forced-exposure pretest, or a brand tracker.
Play seconds, attentive seconds, and why this is a probability argument
Attention research measures eyes on the ad, with eye tracking. Platforms measure video running in a viewport. They are different units, and Lumen's cross-media panel puts the population average at roughly two play seconds for every attentive second on Meta and TikTok feeds. So we do not call stop rate or watched rate attention measurement, because they are not.
What they are is stronger than a proxy, and the reason is the medium. On a skippable, mobile, thumb-driven feed the default behaviour is to leave. Every second an ad keeps playing is a second the viewer chose not to scroll, and Dentsu's work with Kantar and Lumen found that once an ad is watched past two seconds the viewer has effectively chosen not to skip, and that voluntary attention works harder than forced attention. That two-to-one ratio is an average across all impressions, including the ones scrolled past in half a second. Conditional on a three-second play, the attention density is materially higher, and conditional on a fifteen-second play, higher again.
So the argument is statistical. If five times as many people play the ad for three seconds and three times as many watch it to the end, the odds that any given impression received active attention have risen, and the expected active attention delivered per thousand impressions has risen with them. We cannot tell you which impressions were attended to. We can tell you the probability went up, by how much the share of watched impressions went up, on a medium where the default is to give none. That is the claim, and it is the claim the performance tier makes.
The second thing to carry is that attention is a necessary condition, not a sufficient one. The Advertising Research Foundation's 2026 validation work found no one-to-one relationship between attention and upper-funnel brand lift, and that media weight, reach and frequency are needed for attention to convert into brand impact. A quadrupling of watched impressions is evidence that the ad did its job and that brand building became four times as possible. It is not, on its own, evidence that the brand moved. That distinction is the reason the tiers below are kept separate.
The three tiers
Peter Drucker's line still does the work: efficiency is doing things right, effectiveness is doing the right thing. Effectiveness goes first because achieving the right goal matters more than saving money on the wrong one. Efficiency goes last because a cost is meaningless until you know what it bought. The middle tier is the one that connects them, and it is the one almost nobody in a client structure owns.
One word to be careful with. We say goal, or intended outcome, for what the brand is trying to achieve. We do not say objective, because on Meta and TikTok an objective is a campaign setting, awareness, traffic, sales, and the efficiency rules later in this piece need that word for that meaning.
1. Effectiveness. Did the brand move? Whether the final outcome achieved the intended goal. The retrieval end of the chain. Brand lift, message association, consideration, favourability, and where the measurement exists, penetration and sales. This tier is only reported when it has been measured. It is never proxied from the tiers below, because a watched impression is not a memory and we will not pretend it is. Accountability is shared: the brand sets the goal, media delivers the weight, we deliver the ads that turn weight into encoding.
2. Performance. What did people do with the ad? How well the process intended to deliver the outcome was executed. The attention and encoding steps, plus action where the campaign calls for it. Did they stop, did they stay, did they click or interact. Every metric in this tier is a share of the impressions media bought, so it isolates what the ad itself earned from what the media plan delivered. This is the tier we are accountable for, and it is the tier we can move without touching the media plan.
3. Efficiency. What did each one cost? How well resources were used to deliver the outcome. Cost per step down the chain. It splits in two, and the split matters commercially. Media efficiency is the cost of exposure, CPM, and it is the media team's number with a creative contribution that depends on the objective. Marketing efficiency is the all-in cost of a watched impression, media plus our fee, and it is the number that tells a CFO whether the whole arrangement returns more than it costs.
Why the tiers have to be read together
Each tier on its own can be made to look like success. The failures are specific and each has a name.
High performance that does not deliver the outcome is vanity. Stopped, watched, clicked, and the brand did not move. This is the risk of vanity metrics, and the chain is what separates a vanity metric from a leading indicator. Stop rate and watched rate sit upstream of encoding, so they are leading indicators of an outcome, and a failure downstream is a message problem or a media-weight problem, not a measurement problem. Interaction rate mostly does not sit on the chain, it selects for people who already like the brand, which is why it is the textbook vanity metric.
An efficient process that does not deliver the outcome is cost-effective waste. Cheap impressions nobody watched read as success on the media scorecard and as nothing at all on the brand tracker. This is the structural risk of splitting accountability, one team measured on efficiency, another on effectiveness, and nobody measured on the performance tier that connects them. It is also the most common shape of paid social underperformance we see, because the signals that would reveal it are in the account and nobody's job depends on reading them.
An effective result with no performance story is luck. The brand moved and nobody can say why. Without the middle tier, an effectiveness result cannot be diagnosed, repeated or scaled. It is a result, not a system.
The standard exists so that a result is only called a result when all three tiers have been read, and the sentence that describes it says which tier each number belongs to.
The glossary: three ways to say every metric
The table is the standard. The technical term is what the platform or the industry calls it and what an expert will look for. The plain-English name is what goes on a slide header or a website card. The third column is the one that matters for the broader marketer, because this is an industry where most people making budget decisions have never been told what these numbers actually mean.
Exposure (media's tier, reported for context)
| Technical term | Plain English | What it actually tells you | Definition and notes |
|---|---|---|---|
| Impressions | Times shown | How many times the ad was served. Not how many times it was seen | Meta and TikTok count an impression when the ad enters the screen. Meta video counts at first frame |
| Reach | People reached | How many different people were served the ad at least once | Ad-level reach inside a shared ad set does not sum, the same person can see several ads |
| Frequency | Times per person | How often the average person was served it | Impressions divided by reach. Under most auction objectives it cannot be managed |
| MRC viewable impression | Viewable | The ad was at least half on screen for two continuous seconds. A technical delivery floor, not a measure of attention | The MRC industry standard, reported via DoubleVerify, IAS or Moat. This is the only thing "viewability" means. Never use the word for anything in the tiers below |
Performance (our tier)
| Technical term | Plain English | What it actually tells you | Definition and notes |
|---|---|---|---|
| 3-second video plays ÷ impressions (hook rate, thumbstop rate) | Stop rate | The share of people who paused for at least three seconds instead of scrolling past. The opening did its job or it did not | Meta counts a 3-second play as three seconds or nearly the total length if the video is shorter. Not a native metric, it is calculated. TikTok's nearest native unit is the 2-second view, which is a lower bar and not equivalent |
| ThruPlays ÷ 3-second plays (hold rate) | Hold rate | Of the people who stopped, the share who stayed long enough to take in the message. The story held or it did not | Calculated, not native. Isolates the body of the ad from the opening. Length-sensitive: a six-second ad will always show a higher hold rate than a thirty-second one |
| ThruPlays ÷ impressions (ThruPlay rate) | Watched rate | The share of everyone served the ad who watched enough of it for a message to land | Meta's ThruPlay counts completion for videos under 15 seconds, or 15 seconds for longer ones. This is stop rate multiplied by hold rate |
| Focused views ÷ impressions (TikTok) | Watched rate (TikTok) | The TikTok equivalent of the above, on a different threshold | A focused view counts at six seconds, or completion for shorter ads, or a positive interaction within the first six seconds. A 15-second variant exists. Not the same currency as ThruPlay, never blend them into one number |
| Watched impressions | Watched impressions | The absolute number of times the message was actually consumed. The number that mental availability is built from | Our umbrella term. ThruPlays on Meta, focused views on TikTok, always reported per platform with the threshold stated |
| Video plays at 100% ÷ impressions (VCR) | Full-watch rate | The share who watched to the end | Available on both platforms. Only meaningful for short assets, feed drop-off makes it look artificially low on anything over 15 seconds |
| Video average play time | Seconds watched | How long the average viewer stayed | Strong secondary proof point. Compare like-for-like on asset length |
| 2-second continuous video plays | Two-second views | The video appeared and was not instantly scrolled past | Weakly discriminating between creatives. We report it only when a client's plan is optimised on it |
| Link click-through rate (CTR), landing page view rate | Click rate | The share who went somewhere because of the ad | Report when the objective is traffic or conversion. Landing page views are the more honest of the two, they price in the drop-off between click and arrival |
| Post engagements ÷ impressions (engagement rate) | Interaction rate | The share who reacted, commented, shared or saved | On Meta, clicks no longer count as interactions. Report with care: interaction selects for people who already like you, and engagement without branding is entertainment |
Efficiency
| Technical term | Plain English | What it actually tells you | Definition and notes |
|---|---|---|---|
| CPM | Cost per thousand impressions | What the media plan paid to be shown. Media efficiency | The media team's number. Creative contributes to it, and the direction depends on the objective: under reach and awareness, more relevant creative lowers CPM; under conversion, better creative raises CPM and lowers cost per outcome. We report CPM changes with the objective stated |
| Cost per ThruPlay (CPTP), cost per focused view (CPFV) | Cost per watched impression | What each consumed message cost, on media dollars alone | CPM divided by (1,000 × watched rate). This is the bridge between the tiers: at a given CPM, the ad determines what share of impressions get watched, so this is where performance shows up as money |
| All-in cost per watched impression | Cost per watched impression, fees included. Marketing efficiency | Whether the whole arrangement, media plus creative fee, delivers a consumed message for less than the media-only baseline did | (Media spend + creative fee) ÷ watched impressions, against the brand's prior creative on media spend alone. The working/non-working question answered in one number |
| CPC, CPA, ROAS | Cost per click, cost per action, return on ad spend | What each click, lead or sale cost | Demand-capture metrics. Report when the objective is conversion. Not proxies for brand effect |
| Cost per lift point | Cost per point of brand lift | What each percentage point of measured brand movement cost | Only where a lift study exists. Media spend divided by lift in points. The truest efficiency number and the rarest |
Effectiveness
| Technical term | Plain English | What it actually tells you | Definition and notes |
|---|---|---|---|
| Brand lift study (Meta, TikTok) | Brand lift | Whether people who saw the ads remember, associate or consider the brand more than a matched group who did not | Survey-based, exposed against holdout. Measures ad recall, awareness, message association, favourability, intent. Minimum spend thresholds apply. Platform-run, so the platform grades its own homework, worth saying |
| Forced-exposure pretest (Swayable, System1, Kantar Link) | Brand impact in a controlled test | Whether the ad moves favourability, association or intent when everyone watches it in full | Controlled and randomised, so it isolates the ad from the media. It removes the attention step entirely, so a strong result here plus a weak watched rate in-feed is a message problem solved and an attention problem not |
| Brand tracker movement | Brand health | Whether awareness, consideration or penetration moved over the period | Kantar, Nielsen, System1 and similar. Slow, expensive, and the only measure that includes everything the brand did, not just this campaign |
| Estimated ad recall lift (Meta) | Predicted recall | Meta's model of how many people would remember the ad if asked | A prediction, not a measurement. Useful as a directional signal, never reported in this tier as if it were a lift study |
| Sales lift, MMM contribution | Sales effect | Whether the campaign moved sales, and how much of the movement it can claim | Where the client runs it. We report it, we do not produce it |
The tile
Every metric, wherever it appears, on a chart, a one-pager, a sales deck or the website, is presented as the same tile. Three lines, a fourth when there is room. An expert and a general marketer read the same tile and each gets what they need.
Line 1, tier · metric. The tier first, then the technical name of the metric. Performance · ThruPlay rate. Efficiency · CPM. Effectiveness · Brand lift. The tier tells any reader what kind of number they are looking at, so the framework travels with every number and never has to be explained on the page. The metric is the technical term, so an expert can check it. Always this order, always this format, always one line. Metric names that run past about 28 characters take a short form: completion rate for 100% view-through rate, cost per full view for cost per completed view.
Line 2, the number. A multiple for gains, a percentage for cost reductions. Teal when the tier is effectiveness, blackberry otherwise, because the outcome is what the eye should land on first.
Line 3, what it means to the brand. One sentence, plain English, written for someone who has never opened an ads manager. Two lines maximum, about 60 characters, because a third line breaks the rhythm of a row. It has to be a true sentence about people or money, not the metric restated: five times as many people watched the whole ad passes, 5x completed view rate does not. This is the line most result designs leave out and the one the buyer actually reads.
Line 4, the definition, optional. Formula, threshold, baseline. Two lines maximum or dropped. Printed in decks and one-pagers, a hover or footnote on the website, dropped on charts.
The tile at three sizes is the same tile. Full, with the definition, for decks and one-pagers. Compact, three lines, for the website. Inline, number first with the eyebrow and meaning beside it, for chart annotations and table rows.
An illustrative set, with numbers that are internally consistent and labelled as an example rather than a result:
Performance · Stop rate 2.5x Two and a half times as many people stopped scrolling. 3-second plays ÷ impressions, vs previous creative, same account.
Performance · ThruPlay rate 4.0x Four times as many people watched the whole message. ThruPlays ÷ impressions. Full ad, or 15 seconds for longer ads.
Efficiency · CPM −20% Every thousand times shown cost a fifth less than before. Awareness objective, vs previous creative. Media efficiency.
Efficiency · All-in cost per watched impression −75% With our fee included, each watched message cost a quarter of before. (Media + fee) ÷ ThruPlays, vs media-only baseline. Marketing efficiency.
Effectiveness · Message association +4.7 points Nearly five more people in every hundred now link the brand with the claim. Forced-exposure test, n=9,800.
The arithmetic behind that set, so the reader can see that the tiles agree with each other: the baseline creative stopped 12% of people and held 25% of those, so 3% of impressions were watched, at a $10 CPM, which is 33 cents per watched impression. The new creative stopped 30% and held 40%, so 12% were watched, at an $8 CPM, which is 6.7 cents per watched impression on media alone and 8.3 cents once a 25% fee is added to the media cost. The all-in figure is a quarter of the baseline, hence 75% lower. If the fee had not been recovered, that tile would say so, and it is the tile that keeps us honest.
What the tile does to a row of results is worth noticing. Three performance tiles in a row with no effectiveness tile is a row that has not proven the outcome, and a reader who knows the framework sees that at a glance. That is the intended effect. The tier prefix is how the vanity-metric check happens on the page rather than in a footnote.
The results sentence, for a case study opening or a QBR headline, takes the tiles in chain order and ends on the outcome: Four times as many people watched the message, at a quarter of the all-in cost per watched impression, and message association rose 4.7 points. Where there is no lift study, the sentence stops after the cost, and does not borrow an effectiveness claim it has not earned.
The rules that travel with the standard
Multiples for gains, percentages for reductions. "+408%" is 5.1 times, not 4 times, and half the room will read it as 4. Write 5.1x. Write "five times as many", never "five times more", which some readers hear as 6x.
Name the baseline. Every multiple is against something: the brand's prior creative in the same account, the account average, or a published benchmark. Say which. Benchmark figures circulating for hook rate, hold rate and ThruPlay rate now largely cite each other rather than any primary sample, so the brand's own baseline is the only one we lead with.
State the objective on every efficiency card. A CPM reduction under reach optimisation and a CPM increase under conversion optimisation can both be the mechanism working. Without the objective, the number is uninterpretable.
State the platform and the threshold. ThruPlay on Meta and focused view on TikTok are different currencies. Report them side by side, never summed.
Match asset lengths. Meta counts a 3-second play as nearly the full length if the video is shorter, and a ThruPlay as completion for anything under 15 seconds. A six-second asset will inflate every rate in the performance tier against a thirty-second one. Compare like with like, or say that you have not.
Never say viewability for anything but the MRC standard. The word has one industry meaning, at least half the pixels on screen for two continuous seconds. Watched rate is not viewability. Stop rate is not viewability. Using the word loosely tells a procurement team you do not know the difference.
Never proxy effectiveness. A hold rate is not brand lift. A watched rate is not mental availability. If the effectiveness tier has not been measured, the case study says what was measured and stops there.
Name the interest. Where a number comes from a platform, a vendor selling the metric, or a dataset drawn from performance accounts, say so. Amplified and Lumen sell attention measurement. Meta runs its own brand lift studies. Motion's benchmarks come from DTC-heavy accounts. None of that makes the numbers wrong, all of it belongs next to them.
Modelled or measured. Estimated ad recall lift is a model. Estimated action rate is a model. A brand lift study is a measurement with a holdout. A forced-exposure pretest is a measurement without the media. Knowing which is which changes what a reader is allowed to conclude.
What we claim, and what we do not
We are accountable for the performance tier. At whatever media plan the client and their media team have chosen, the share of impressions that get stopped, watched and acted on is what our ads earned, and we report it against the brand's own baseline in the brand's own account.
We claim a contribution to media efficiency where we have measured it, with the objective stated. Under awareness and reach objectives, better-matched creative clears at lower CPMs, and we have seen that in the accounts we run. We do not claim a universal multiplier, because the direction flips under conversion objectives and the magnitude has never been cleanly quantified for commercial advertising.
We claim marketing efficiency only when the all-in arithmetic supports it. The fee sits in the numerator and the watched impression in the denominator, against the media-only baseline, and the number is what it is. This is the one card that can come out against us, which is exactly why it belongs in the standard.
We report effectiveness when it has been measured and we do not manufacture it when it has not. Attention is the necessary condition for mental availability, not a guarantee of it. Zero watched impressions guarantees zero brand building. Four times as many watched impressions, on a medium where the default is to skip, makes active attention four times as probable and brand building four times as possible. Whether it happened is a question a lift study answers, not a dashboard, and we would rather run the study than borrow the claim.
How to measure attention on social with data you already have takes the performance tier into an account's own data, and is the natural next read.
Media creates the opportunity to see your ads. Creative determines whether they work.
Questions
Why do you read effectiveness first if you are only accountable for performance?
Because achieving the right goal matters more than saving money on the wrong one, and a performance number only means something once you know what it was for. Effectiveness is whether the brand moved, performance is whether the ads earned attention on the way there, efficiency is what each step cost, and reading them in that order is what stops a hold rate being quoted as brand impact or a CPM reduction being quoted as effectiveness.
Accountability is shared along the chain. The brand sets the goal, media delivers the weight, and the ads turn weight into watched impressions, which is the tier we are answerable for and the tier we can move without touching the media plan. We report effectiveness when it has been measured, and when it has not, we say what was measured and stop.
What is the difference between media efficiency and marketing efficiency?
Media efficiency is CPM, what the media plan paid to be shown, and it is the media team's number with a creative contribution whose direction depends on the objective. Under reach and awareness objectives more relevant creative clears at a lower CPM, under conversion objectives better creative raises CPM and lowers the cost per outcome, which is why every efficiency card states the objective.
Marketing efficiency is the all-in cost per watched impression, media spend plus the creative fee divided by watched impressions, against the brand's prior creative on media spend alone. It is the one number that answers the working versus non-working question, and it is the one card that can come out against the fee, which is exactly why it belongs in the standard.
What does "a view" actually mean on Meta and TikTok?
Seven different things, which is why the standard never uses the word on its own. On Meta, a 3-second video play counts three seconds or nearly the full length of a shorter video, a ThruPlay counts completion for videos under fifteen seconds or fifteen seconds for longer ones, and a 2-second continuous play is the floor below which the ad was scrolled past at once. On TikTok, a 2-second view is the nearest native unit to Meta's 3-second play and a lower bar, and a focused view counts six seconds, completion, or a positive interaction inside the first six seconds.
None of these is viewability, which has one meaning, the MRC standard of at least half the pixels on screen for two continuous seconds. Every result we present names the platform and the threshold, and ThruPlay on Meta and focused view on TikTok are reported side by side, never summed into one number.
Sources
- Amplified Intelligence, "How attention fuels mental availability". The exposure, attention, encoding, retrieval chain, and the active, passive and non-attention classification. Amplified sells attention measurement
- Lumen Research, cross-platform attention benchmarks presented to the Media Research Group, panel late 2022 to early 2023. Attentive seconds per thousand impressions and average view time by platform. Lumen sells attention measurement
- VCCP Media with Karen Nelson-Field and Amplified, distinctive assets and attention, 72 ads across 8 brands. Source of the 1.5-second active attention finding, which is a retrieval threshold established on brands with existing distinctive assets
- Dentsu with Kantar and Lumen, attention economy research. Source of the finding that a view past two seconds is a choice not to skip, that voluntary attention outperforms forced attention, and that brand-building effect plateaus around twenty seconds of active viewing. Dentsu and Lumen both sell attention measurement
- Advertising Research Foundation, Attention Measurement Validation Initiative, Phase 3 (June 2026). No one-to-one relationship between attention and upper-funnel lift; media weight, reach and frequency as necessary conditions
- Media Rating Council, viewable impression guidelines for video: 50% of pixels in view for two continuous seconds
- Meta Business Help Centre, video metrics definitions (3-second video plays, ThruPlay, 2-second continuous video plays, video plays at 25/50/75/95/100%, video average play time), and brand lift study methodology. Verify wording against the live Help Centre
- TikTok Ads Manager Help Centre, video view metrics (2-second views, 6-second focused views, 15-second focused views, video views at 25/50/75/100%, average watch time per video view) and brand lift study methodology
- Creatively Squared, "What creative can and cannot change, by campaign type" (August 2026), for the objective-dependence of creative's effect on CPM and the play-seconds to attentive-seconds conversion
- Peter Drucker, The Effective Executive (1967), for the efficiency and effectiveness distinction