Audience Quality Estimator

Start with a decision, not a verdict about accounts

Audience quality is useful only when it means something specific. A regional language school needs a different audience from a camera repair community. Start with a decision: choosing a campaign topic, reviewing community participation, or considering a creator partnership. Define relevance using evidence you can observe rather than assumptions about the people behind profiles.

This Audience Quality Estimator organizes a manual sample review. You inspect your own chosen sample separately and enter aggregate counts for size, relevant accounts, unclear fit, and optional participation and review flags. The tool calculates separate percentages and compares them with thresholds you choose. It does not inspect individual accounts, retrieve profiles, connect to a platform, import audience records, or decide which accounts belong in any category.

Choose Instagram, TikTok, YouTube, Facebook, LinkedIn, Threads, X (Twitter), Snapchat, or Reddit as the context. This label does not change the formulas into platform-specific detection. The review goal also labels your purpose without supplying a classification rule. Keep that rule in your own notes.

The result describes the sampled accounts and observations you supplied. It is not an authenticity percentage, a fake follower detector, or a statement about everyone who follows or views your content. A bounded review can support a practical next step without presenting selected observations as access to an entire audience.

Prepare one sample and a clear relevance rule

N is the number of distinct accounts in the sample you reviewed. Count an account once even if it appears under several posts or contributes repeatedly. State where the sample came from, how you selected it, and which dates it describes. A visible discussion sample is a sample of those contributors, not everyone who watched. A prepared follower roster has a different scope and may still omit accounts your source did not contain.

R counts accounts meeting your relevance rule. For an introductory illustration class, demonstrated interest in learning drawing may fit, while a professional print supplier may not. Make the rule concrete enough for another reviewer to apply. General popularity is not a substitute for goal fit. The same account could suit another project under a different rule.

U is the count whose fit remains unclear. Use it when the available evidence cannot support either a relevant or not relevant classification. A private or sparse profile can leave fit unclear without saying anything about authenticity. The remaining N minus R minus U accounts are classified as not relevant to the stated goal. Unclear fit is not a category for unknown participation, and it must not be used to hide unfinished optional checks.

Enter whole account counts: N can range from one to one million, R and U must each be within N, and R plus U cannot exceed N. Check the totals against your working notes before interpreting a result. If you discover a duplicate, correct all affected counts rather than changing N alone. The audience sampling guide explains how selection, missing evidence, and observation periods affect a repeatable manual review.

Read relevance together with fit classification coverage

The classified-fit count is N minus U. Relevance is R divided by that classified count, multiplied by 100. This answers a narrow question: among the sampled accounts whose fit you could classify, what share met the relevance rule? It does not include unclear accounts in the relevance denominator. Keep the formula in mind when comparing results, because a high percentage based on few classified accounts can conceal a substantial amount of missing fit evidence.

Fit classification coverage is the classified count divided by N, multiplied by 100. It describes completeness of fit classification, not another quality score or the proportion of a platform audience inspected. Full classification coverage can still describe a narrow visible source. Read coverage beside relevance to see how much of your selected sample supports that percentage.

Suppose you review 50 accounts and classify 18 as relevant while leaving 20 unclear. There are 30 classified-fit accounts, including 12 classified as not relevant. Relevance is 60%, while coverage is also 60% for a different reason. The matching percentages are a coincidence. The first describes relevant accounts within the classified group; the second describes the classified group within the full sample. Report the denominator counts to keep that distinction visible.

If U equals N, there are no classified-fit accounts and relevance is unmeasured. Do not interpret that as zero relevance or change uncertain classifications to force an answer. Check whether the rule requires evidence your source cannot provide. You may need a better source or a different review question. If you revise the goal or rule, record the revision and treat the resulting totals as a new review rather than quietly replacing the original conclusion.

Add participation and flags only after complete checks

P is an optional count of distinct sampled accounts with meaningful participation during your stated observation period. Decide what meaningful means before counting: a specific product question, a useful troubleshooting answer, or a contribution that advances a learning discussion are possible rules. Count accounts, not their comments, likes, messages, or other interactions. An account contributing five qualifying replies adds one participant. Contributions from accounts outside this sample do not belong in its P count.

Participation is P divided by N, multiplied by 100. Enter P only after applying the same participation review across all N accounts. If you checked activity for only part of the sample, leave P blank until that check is complete. Blank means unmeasured; zero means a complete check found no qualifying participants. Unknown fit does not remove an account from the participation denominator, and it does not establish whether that account participated.

C is an optional count of sampled accounts with observed review flags under a stated checklist. Flag incidence is C divided by N, multiplied by 100. A flag records something needing investigation, such as an unresolved inconsistency in evidence you reviewed. It is not confirmation of a bot or a fabricated identity. Count a flagged account once even when several observations prompted the flag, and complete the same flag check across the full sample before supplying C.

The categories can overlap. A relevant account might contribute a useful answer and also have an observation requiring follow-up. Another account might have unclear fit while clearly meeting your participation rule. Keep those independent facts in the appropriate counts instead of forcing every account into one exclusive category. Neither P nor C needs to be smaller than R. Each optional count must be within N, and an incomplete optional review stays blank rather than becoming a reassuring zero.

Set thresholds as planning choices, not platform norms

The starting thresholds are 70% target relevance, 30% target participation, and a 10% flag review ceiling. They are illustrative planning settings that you can adjust. They are not published averages for Instagram, LinkedIn, or any other platform, and they do not define a universally good audience. Choose values that reflect the purpose and consequences of your review, then keep a record of them beside the sample definition.

Relevance and participation meet their chosen targets when the calculated percentage is at least the target. Flag incidence is within the chosen ceiling when it is at or below that ceiling. Equality counts as meeting the threshold in each case. These statuses compare your observations with your settings; they do not validate your sampling rule or confirm the truth of the underlying classifications. Read them as decision prompts supported by the recorded evidence.

Status comparisons use unrounded percentages, so a rounded display can appear to meet a boundary while the actual result remains below it. Inspect the numerator, denominator, and threshold together. Do not adjust counts to match the displayed impression. Before a costly decision based on a borderline result, improve or repeat the manual review instead of treating a tiny difference as certainty.

There is no composite score. Strong participation does not cancel weak relevance, and low observed flag incidence does not prove authenticity. An optional metric left blank does not meet its threshold; it was not measured. Changing a threshold changes the status comparison, not the observations. Set your criteria before reviewing a campaign when possible, and explain any later adjustment so a colleague does not mistake a changed target for improved audience evidence.

Work through a sample without blending the results

Imagine a fictional independent bookshop reviewing accounts selected for a reading group campaign. It records N equal to 100, R equal to 56, and U equal to 20. The other 24 accounts were classified as not relevant to this particular group. The selection note identifies the source and dates, and the relevance rule concerns demonstrated interest in the planned reading theme. These are manually assessed observations, not classifications supplied by the tool.

The classified-fit count is 100 minus 20, giving 80 accounts. Relevance is 56 divided by 80, or 70%. Classification coverage is 80 divided by 100, or 80%. With a 70% relevance target, the result is within target because the comparison includes equality. The conclusion is that the classified portion of this sample met the chosen fit target, while one fifth of the sample still lacked a clear fit classification.

A complete participation check finds 24 sampled accounts with qualifying contributions, so P is 24 and participation is 24%. A complete flag check finds eight accounts with observations needing review, so C is eight and flag incidence is 8%. Using the starter settings, participation needs review against the 30% target, while flags are within the 10% ceiling. Some of the eight flagged accounts may also belong to the 24 meaningful participants.

The bookshop can examine its invitation, investigate flags, and seek missing fit evidence. It cannot conclude that 70% of all followers will join, that 8% of its audience is fake, or that 24 comments occurred. These are sampled account counts with distinct follow-up tasks.

Keep the platform context honest

For Instagram and TikTok, name the content behind the sample. An Instagram repair demonstration and a TikTok comedy response can draw different visible contributors, even for the same business. On Facebook, an event discussion may suit an attendance question better than a general page conversation. State whether you reviewed followers, selected contributors, or another accessible group. Do not assume a profile relationship tells you who saw a particular post or why they responded.

On YouTube, viewers are not simply subscribers, and a manual comment sample is narrower than either category. YouTube defines monthly audience through a rolling 28-day viewing window, so keep that aggregate separate from a fixed comment review. On LinkedIn, impressions count times a post was shown, while members reached estimates distinct members and Pages. Neither native number gives this estimator a ready-made roster with classified relevance. Use them only as separately scoped context.

Threads and X can produce conversations involving accounts beyond your usual recurring contributors. If you choose reply authors as the source, say that the sample concerns those authors rather than all followers. X view counts can include repeated viewing, so a view total is not a distinct account sample. On Snapchat, a manually observed reply or a known contact can support a bounded review, but the tool cannot reveal unseen viewers or infer their interests from aggregate activity.

For Reddit, distinguish discussion participants from profile followers and community members. A technical thread can support a topic review without describing the whole community. Use evidence actually available to you; native analytics and visibility vary by account and context. The YouTube automation guide covers separate operational work, not a source of audience fit data.

Choose the next evidence or action from the pattern

When relevance meets your target but participation is low, examine the invitation. A specific question or a clearer event explanation may create a reason to contribute. That is a proposed test, not a finding from the current sample. Low participation does not reveal whether accounts saw the content, lacked time, or preferred to learn without responding.

When participation is strong but relevance is weak, check whether the topic serves a broader group than the offer. A popular discussion can still be valuable, but it may not support a narrow campaign decision. The audience quality versus engagement guide explains that trade-off. Use the separate Engagement Rate Calculator for response-event calculations rather than putting likes or comments into this tool’s account fields.

When fit coverage is low, identify the evidence gap. You may need a clearer relevance rule or a more suitable source, not a more confident interpretation. When flag incidence exceeds your ceiling, inspect the specific observations and the consistency of the checklist. Neither pattern supports automatic account removal. If you separately maintain comparable prepared rosters, the Follower & Connection List Checker can compare their membership; it does not infer relevance or feed records into this estimator.

Connect commercial conclusions to separately verified outcomes. A relevant sampled account is not automatically a qualified lead, and a meaningful contribution is not a sale. Track inquiries, qualification, costs, and confirmed outcomes in their proper records. The Social Media ROI Calculator is useful once you have supported financial inputs. Do not multiply sample fit by total followers to invent a revenue forecast or an expected number of customers.

Save the review and keep operations separate

Copy the plain-text review for a concise project note, download CSV for aggregate values, or save the two-page PDF for a chart and practical interpretation. Keep the source definition, selection rule, observation dates, and classification checklist alongside the report in your own storage. The report’s footer points to this exact estimator. If you revise the input counts or thresholds, generate a fresh export so the saved interpretation matches the values being discussed.

The browser tool does not send or persist inputs, require usernames, or maintain a review history. Save a report when you need a baseline. The optional period label identifies the observation without filtering activity or fetching history. A later comparison needs a new manual review with compatible definitions and an explanation of any changed source or window.

SMTasker supports workflows across nine platforms through Android phones and a local Windows or macOS client. Skye can help with supported workspace tasks using your own OpenAI or Anthropic key, with separate provider billing; external AI can connect through MCP with the permissions you choose. These capabilities are not audience authenticity analytics. This free estimator does not connect its results to SMTasker, transfer individual account records, or apply operational changes automatically.

Use the five-day free trial to test a specific supported account routine separately from this evidence review. A useful handoff states what the sample showed, what remains uncertain, and which next step you chose. Preserve those distinctions so a colleague can understand and repeat the review without relying on a score that hides unfinished observations.