An impressive social account can start a hiring conversation. It should not settle the decision. A candidate who presents a million-view video has shown you a result worth discussing, not yet explained who made it happen or which skills it demonstrates.

The useful question is not simply, “How big were the numbers?” It is, “What did you do, under what conditions, and what supports your interpretation?” Follow that thread from the brief through the work, distribution, outcome and subsequent changes.

Define the job before opening the portfolio

Decide what the person must deliver in the role. Producing short videos, managing customer conversations and running paid campaigns require different evidence. Unless the job explicitly includes being an influencer, do not make a large personal following an informal entry requirement.

CIPD’s guidance on selection methods recommends structured interviews with common questions and pre-agreed scoring criteria. Apply that principle to portfolio reviews: agree the job-relevant criteria before seeing applicants, then ask the same core questions while allowing follow-ups about each person’s work.

Request two relevant samples rather than an extensive new presentation. Tell applicants what you will assess, allow accessible formats and invite requests for adjustments. A modestly presented sample can still contain strong evidence.

Separate personal contribution from the team’s result

For each campaign, ask the candidate to identify their involvement in research, scripting, filming, editing, publishing, community management and analysis. Distinguish work they owned, work they shared, work they approved and work completed by colleagues or external suppliers.

Then ask for an artefact that makes the contribution understandable: a script revision, an editing decision, a publishing schedule, a redacted moderation example or an annotated report. Where templates or AI tools were used, ask what the candidate changed, checked and took responsibility for.

Teamwork is not a weakness. The problem is an unclear claim. “I edited the cut-downs from footage supplied by the production team” is more useful than “I delivered the whole campaign” when the evidence supports only the former. Equally, a manager may demonstrate execution through briefing, quality control and coordination rather than personally filming every frame.

Put every result in context

Ask for a short context note covering the objective, starting point, timeframe, resources, distribution method and relevant outcome. Identify the intended audience and what success meant before publication. Record the account’s position when the candidate became involved, not only its current size.

Use exact campaign and reporting dates. Ask about available media spend, production support, creator fees and staff time, as well as what the candidate controlled. “No advertising spend” does not mean the work had no cost. Where commercial confidentiality applies, agreed budget bands may provide enough context.

Match the outcome to the brief. An awareness campaign should not automatically fail because it lacks sales data; ask what evidence supports exposure to the intended audience. An enquiry campaign needs more than views. A community-management sample may be best assessed through the handling of questions and escalations.

Keep observation separate from causation. Enquiries recorded during a campaign do not, by timing alone, establish that the campaign caused them. Ask what tracking was available, what other activity ran simultaneously and which conclusions remain uncertain.

Check what the numbers actually measure

A follower total does not show how much of the audience pre-dated the candidate’s involvement. A view count does not, by itself, establish that people watched to the end, understood the message or took the intended next step.

Use the platform’s terminology rather than treating every number labelled “views” as interchangeable. YouTube’s official analytics guidance, for example, distinguishes unique viewers, an estimate for the selected reporting period, from engaged views, which count viewing beyond the opening seconds and exclude loops. It also describes a desktop “Organic & paid traffic” filter separating unpaid sources from paid advertisements.

That filter is useful context, not a substitute for asking about commercial arrangements. Do not assume another platform, account type or reporting screen exposes the same information. Ask candidates to retain the metric labels, dates and relevant filters in any permissioned, redacted evidence.

For a claimed engagement rate, ask for the numerator and denominator. Likes and comments divided by views cannot be compared directly with all interactions divided by followers. Compare like-for-like formats and reporting windows where possible; record differences rather than disguising them in a league table.

Separate four routes to exposure

Ask every applicant how the work reached its audience. A campaign can combine several routes, and the candidate may not have controlled all of them.

Organic exposure. Ask about unpaid discovery, search, recommendations, sharing and existing audiences. Establish what the candidate did to support discovery and what distribution came from others. Do not confuse unpaid placement with cost-free production.

Official platform advertising. Ask whether the content ran through a platform’s advertising product, directly or via an agency. Request the objective, spend, reporting period and the candidate’s responsibilities. Paid distribution is not inherently weaker evidence; assess the relevant planning, creative or optimisation work.

Collaborations. Ask which partner published or shared the work, whose audience was involved and whether the arrangement included fees, gifts or reciprocal promotion. Separate the candidate’s contribution from the partner’s production and distribution. A collaboration may also receive platform advertising support.

Purchased engagement-count services. A2G Store lists services for increasing TikTok followers, likes and views. Such services are a reason to ask how reported activity was generated, not evidence of a candidate’s competence, an interested audience or business results. Where used, record what was purchased separately from official advertising, without assuming the origin or quality of a particular service’s activity.

These routes are not always separately measurable. Mark an unavailable breakdown as unknown; do not manufacture an “organic” total by subtracting a purchased quantity without supporting records. An unusual spike or ratio is not proof of misconduct. Ask a neutral follow-up and give the candidate an opportunity to explain.

Use a compact portfolio-review rubric

The framework below is illustrative, not a validated assessment instrument or a universal hiring threshold. Adapt the expectations to the responsibilities and seniority of the vacancy before reviewing applicants.

Dimension	Evidence to request	What a strong explanation shows
Judgement	The brief, audience need, alternatives and risks.	Choices fit the objective; trade-offs and approval boundaries are explained.
Execution	Work samples, revisions and the candidate’s responsibilities.	The applicant can explain delivery, quality checks and collaboration under the stated constraints.
Measurement	Baseline, dates, definitions, resources, distribution and outcomes.	Claims match the evidence; calculations are clear and attribution limits are acknowledged.
Learning	Feedback, a revision or a proposed next test.	The applicant identifies a specific change and explains what evidence would justify it.

For each dimension, record “clear evidence”, “partial evidence” or “not yet evidenced”, together with the sample and reason. Missing access to historical data is not the same as poor ability. Agree a proportionate way to address important gaps. Have reviewers record their initial assessments independently before discussing them.

Two hypothetical portfolio comparisons

The following applicants and figures are invented illustrations, not case studies, performance benchmarks or claims about real campaigns.

Example one: one million views versus 18,000

Candidate A presents a national launch video with one million views. They edited the shorter versions, while other teams developed the concept and handled distribution. The portfolio contains no campaign dates, spending breakdown or outcome beyond the public view count. This supports a discussion of editing, not yet a claim of end-to-end campaign management.

Candidate B presents a four-week campaign with 18,000 views for an account that started with 1,600 followers. They show the brief, scripts, edits and publishing plan, identify a £400 advertising budget and separate the reported paid and unpaid activity. The client approved the claims and handled subsequent customer calls.

A redacted report records 12 completed enquiry forms associated with campaign-tagged links. Candidate B explains that the tracking does not establish additional sales or prove those enquiries would never otherwise have happened. They also show how recurring questions informed a revised video.

For a hands-on content-management vacancy, Candidate B currently supplies more assessable evidence across the rubric. Candidate A should receive the same opportunity to provide context. The smaller number is not inherently better; the clearer account of the work is.

Example two: 90,000 followers versus 2,400

Candidate C shows three polished posts from a brand account with 90,000 followers. The submission does not say how many followers the account had when they joined, or whether they managed its conversations. There is no basis for alleging anything improper; there is simply not enough information to judge community-management skill.

Candidate D presents six weeks of work on a 2,400-follower account, with no paid distribution during that period. Their anonymised samples show response guidance, an escalation log and an FAQ revised after a customer misunderstood its wording. They distinguish their own moderation decisions from issues resolved by the support team and state that they cannot verify subsequent purchases.

For a support-heavy social role, Candidate D gives the panel specific decisions to examine. Candidate C may have equally strong experience, but the follower total cannot supply the missing evidence. Assess both against the same requirements rather than awarding points for account size.

Ask questions that reveal decisions and limits

Use a small set of common prompts, then explore the supporting sample. These questions turn a presentation into a discussion of the work:

“Which part was your responsibility, and which decisions belonged to someone else?”

“What approach did you reject, and why did your chosen approach suit the audience and brief?”

“Which result would you report to a manager, how was it calculated, and what caveat belongs beside it?”

“What changed after feedback or disappointing results, and what would you test next?”

“Which claim would you be unwilling to make from these records, and what evidence would you need?”

Close evidence gaps without crossing privacy boundaries

Accept anonymised briefs, redacted exports and work samples the candidate has permission to share. Avoid collecting customer identities or unrelated information. Do not request passwords or access to private accounts. Former employees may no longer have dashboard access; offer another route to demonstrate the skill.

A clearly labelled reconstruction can explain a process, but should not be presented as an original historical record. Where necessary, use a short exercise with fictional material: ask the candidate to revise a brief, assess three sample posts or recommend a next test. Explain the task, time allowance, permitted tools and adjustment options in advance. Do not turn assessment into an unpaid live campaign.

Finish the review with a written hiring reason tied to the role: what the candidate demonstrated, what remains uncertain and why the evidence is sufficient for the decision. Hire for the work they can explain and perform, not merely the largest number attached to their name.