Audience · people analytics · HR data · measurement · privacy limits · programme evaluation

HR and People Analytics: What Merch Usage Can and Cannot Show

Merchandise programmes generate more data than most teams realise, and most of that data is either operational or uninterpretable. There is a distribution record, a size curve, an issue log and a leftover count, and none of those are engagement metrics. The temptation is to bridge that gap by instrumenting the items — a code per person, a scan at an event, a serial linked to a payroll record — and that route runs straight into the part of the job where proportionality and purpose limitation are not abstract. This page is written for the analysts and data owners asked to evaluate a merchandise programme rather than to admire it. It covers what the existing records actually support as a conclusion, why a giveaway count is not a measure of anything that matters, where the privacy boundary sits when an item becomes linked to a person, and how to phrase a question so that the answer is a number instead of an opinion.

Discuss measurement design Other audiences

What a merchandise programme can legitimately measure

Start from the records that already exist, because they are more useful than teams assume and considerably less sensitive than the alternatives. Four data elements are generated by any running programme: what was ordered, what was issued, what was exchanged or returned, and what remained in stock. Everything else is derived from those four by counting rather than by tracking.

From those four, a defensible set of conclusions is available. Which role groups and which locations take the highest proportion of what is issued. Which sizes run out first and which accumulate. How much of the original order is still present a year later, which is the only genuine durability signal available and the closest thing in this category to a usage measure. How the exchange rate varies by garment type, which tells you something about sizing accuracy rather than about people. Each of those is an aggregate over a defined group, produced by a counting process that does not require knowing who took what.

What those records do not support is anything about sentiment, engagement, belonging or perception. There is no quantity of distribution data that establishes whether people feel identified with the organisation, because no part of the distribution process asks them. A programme with perfect records and a high perceived belonging score tells you nothing about the link, and a programme with poor records and strong sentiment tells you even less. The honest position is that this category generates operational data only, and that treating a distribution figure as a proxy for a cultural outcome is the most common analytical error made in it.

The distinction that decides everything else

The line between an operational record and a people record is whether the record identifies a person. A spreadsheet showing that forty garments in size large were issued at one location on one date is an inventory record. The same spreadsheet with names in a column is a people record, subject to entirely different handling requirements, retention rules and access controls. The physical act is identical. The classification is not, and it is decided by what is written in a column rather than by any intention behind the exercise.

Most HR data environments make this distinction explicitly and most operational workflows make it implicitly and carelessly. A distribution sheet maintained by an office manager is usually operational. The same sheet emailed to a wider group, or imported into a reporting tool alongside other workforce data, can become a people record without anybody deciding that it should. This is worth a governance conversation before the first order rather than after the first export.

Why a giveaway count is not an engagement measure

A giveaway count answers one question accurately: how many items were handed over. Every step from there towards a meaningful conclusion requires an assumption that the data itself does not support. It assumes the recipient kept the item. It assumes keeping the item implies a positive view. It assumes a positive view implies behaviour. Each of those steps is plausible and none of them is measured.

The standard failure mode is a report that moves from units distributed to people reached to engagement rate, with no step in the argument acknowledged. Units distributed and people reached are frequently the same number, because one item goes to one person in most giveaway programmes, which means the two figures carry identical information and are presented as if they were independent. That is not a measurement error so much as a presentation error, and it is the reason these numbers tend to survive scrutiny better than they should.

There is a second problem specific to the way gifts are handed out at events. Distribution at a booth is socially expensive to refuse, particularly when the item is handed over by a person the recipient has just met. The count that results is close to a count of attendance, which means it measures who came, not who was persuaded. Where attendance is known independently, the giveaway figure adds nothing.

None of this is an argument against measuring. It is an argument for measuring the thing that is actually available, which in this category means adoption, retention and coverage rather than reach or reaction. A programme that reports the share of a role group wearing issued workwear after a fixed interval, derived from a voluntary and aggregate return, is reporting something real, and it is honest about how it was collected.

Instrumentation, and where the privacy line sits

The question arrives because somebody wants a per-item answer. Serialise the garments and you can count exactly how many are in circulation. Add a scannable code and you can count how often they are scanned. Link a code to a person and you can count how often a specific individual interacts with a specific object. The first two are logistics. The third is a materially different activity and it is where proportionality has to be assessed.

The useful test is whether the programme needs to know who, or only needs to know how many. Almost every merchandising question is a count question: how many were issued, how many were exchanged, how many are still in use, which size failed. A count does not require identity. When identity genuinely is needed — for example to run an exchange process where a person returns a specific item — the link can exist for that narrow purpose and should expire when the exchange window closes rather than being retained indefinitely.

Three specific patterns cause most of the trouble. The first is per-person serialisation applied at scale without a defined purpose, which produces a tracking capability that will eventually be used for something nobody approved at the outset. The second is joining merchandise data to payroll, performance or engagement records, which converts an operational dataset into a workforce profile and puts it under a completely different regime. The third is retaining a distribution log indefinitely because it might be useful later, which is easy to do and difficult to justify retrospectively.

Operational record: Quantity, size, garment type, location and date. Identifies nobody. Appropriate retention is a business decision, usually short
Exchange record: Links a person to an item for the length of an exchange window. Necessary to operate the process, and should end when the window ends
Issue acknowledgement: Confirms a named person received a named item on a date. Proportionate for onboarding workwear, unnecessary for giveaways
Aggregated report: Counts by team, location or role group with a minimum group size, so small groups cannot be identified indirectly
Serialised inventory: Identifies an item rather than a person. Useful for stock accuracy, rarely justifies personal tracking
Joined workforce data: Merchandise linked to performance, payroll or engagement records. Raises the classification to workforce data and requires an explicit decision, not a default

The minimum group size point is easy to overlook and matters whenever reporting reaches a small team. A breakdown by department is uninformative when one department has four people, and a table that shows size preferences for a team of four is both useless to the programme and a small privacy exposure for the individuals in it. Suppressing small groups is a reporting convention that costs nothing and removes the problem.

Asking a question the data can actually answer

Most requests to this team arrive phrased as an aspiration rather than a question. Teams ask what people think, or whether the programme is working, or for an engagement figure. None of those can be answered from merchandise records, and answering them anyway with a proxy is how bad dashboards are built. The productive move is to translate the aspiration into a count that somebody actually recorded.

The question as usually asked	What the records can answer instead	Data needed
Is the programme working?	What share of the target role group has been issued an item, and what share is still in use after a fixed interval	Issue log plus a short voluntary return, reported as an aggregate
What do employees think of the brand internally?	Nothing from merchandise records. This needs a separate survey instrument	A survey with its own sampling, consent and reporting design
How many people did we reach?	How many items were distributed, and against what recorded attendance	Distribution count plus event attendance, reported as the same figure where it is the same figure
Which garment performs best?	Which size and role group combination had the lowest exchange and return rate	Issue and exchange records joined at item and size level
Should we run this again?	What share of the previous run remained in use, and what was the leftover position at the last reconciliation	Stock reconciliation records, which most programmes already keep informally
Is the sizing right?	What proportion of issued items were exchanged for a different size, by size and role group	Exchange records with the original and replacement size

The reformulation is usually straightforward and occasionally uncomfortable. The second row is the one worth pausing on: when a team asks for an internal sentiment view, the answer is that merchandise records cannot provide it, and the honest response is to say so rather than to construct a proxy that will be quoted later as a perception measure. A programme that has refused to invent a number has a more credible dashboard than one that has not.

Two structural habits make this sustainable. First, decide the reporting cadence before the programme launches and record only what that cadence requires. Second, decide at the outset what the programme will not claim to measure, and write it down. The exclusions are usually shorter than the inclusions, and they protect the programme from the most damaging kind of finding, which is a number that was trusted and turned out to mean nothing.

Four production lines, seen as a data owner

For a data owner, each production line is a different measurement problem before it is a production problem. What each one generates by default, and what it would take to measure properly.

Booth staff apparel

Pre-printed garments issued to defined roles, usually in a named size at a known location. This is the easiest line to measure well because the population is known and the distribution point is a single desk. Issue, size and date are routinely recorded without any extra effort, which makes it the right place to test any tracking idea before extending it. It is also the line where a wrong size has a real operational cost, so the exchange rate is a useful measure of sizing accuracy rather than of anything about the people involved.

High-volume giveaways

Multiplied items handed out at events or across a campaign. This is the line that generates the largest volume of numbers and the smallest amount of information per number. Distribution counts are easy to produce and easy to misread, because they track attendance more closely than most reporting assumes. What is genuinely useful and rarely recorded is the leftover count at the end of an event, because it is the only figure that distinguishes a well-bought giveaway from an over-bought one, and it requires somebody to open the boxes at close.

Lead-capture gifts

Items attached to a sign-up, a membership or a donation. The measurement opportunity is real because the event itself generates a comparable count, and the ratio between items handed over and sign-ups completed is a figure that actually informs the next event. What must not happen is a join between the individual who took the item and the individual record behind the sign-up. Aggregate counts from the same event are sufficient for the question and carry none of that exposure.

VIP and speaker kits

Per-person packages for hosted guests. The numbers here are small, which means any breakdown is likely to identify individuals, so reporting should stay at the level of count, purpose and budget line. The useful measure is not sentiment and not reach; it is whether the programme produced the intended number of prepared kits, how many were actually delivered, and whether the spend sat where the file says it did. A report that carries a name attached to an item description in this category is a disproportionate record.

Setting the measurement design before the order

The most useful moment to decide what will be recorded is before the first order, because the delivery format determines what is easy to capture. A supplier who is told at specification stage that distribution will be recorded by size, role group and location will pack accordingly, deliver labelled cartons by location, and supply a size breakdown per line. Told nothing, they will deliver in mixed cartons and the team will reconstruct the breakdown from memory a month later.

Three decisions belong in that specification. What fields the delivery will carry, so the counting happens at handover rather than afterwards. What the exchange window is and where exchange records are held. And what the programme will not measure, which is the field that most often gets left blank and most often causes trouble later.

The last point is worth stating plainly to stakeholders. A merchandise programme is an operational and creative programme. It is not an engagement instrument, it was not designed as one, and retrofitting measurement to it after the fact produces numbers that look like evidence and are not. Being able to say that clearly, with the real measures attached to it, is a more useful answer to an analyst than a chart.

Questions people analytics and HR data teams ask

Can we serialise the garments so we know exactly how many are in circulation?

Yes, at item level, and it is useful for stock accuracy because it makes a loss visible instead of invisible. The point at which the answer changes is when the serial is joined to a named person or to a workforce record. That join is a different activity with a different classification, and it should be an explicit decision with a stated purpose rather than a convenience that appears in a spreadsheet. If the purpose is operating an exchange process, the join is proportionate for the length of the window and should end with the window.

How many people did the programme reach?

The only honest answer is the number of items distributed, reported alongside whatever attendance or registration figure exists for the same event, and labelled as a count of distribution rather than of people. In most giveaway programmes one item goes to one person, so the two numbers are the same number and presenting them as independent overstates the evidence. If the programme distributed some items in multiples, or some were collected by colleagues on behalf of others, the two figures genuinely differ and both should be given. Anything beyond that requires asking people, which is a survey and not a count.

Is a scan code on every item acceptable under our data policy?

It depends on what the code resolves to. A code on an item that identifies the item and not the person is an inventory mechanism and usually causes no issue. A code that resolves to a named individual is personal data about a workforce member, and the decision to deploy it belongs with the people who own data governance rather than with the programme team. The practical test is whether the programme could operate without the link. If it could, the link is optional and should be left off, and that single question resolves most of these cases.

Is Corp-Merch a partner or an official supplier to HR and people analytics teams?

No. Corp-Merch is an independent B2B merchandise supplier. It has no relationship with, endorsement from or contracted status with any employer, HR function, analytics team or organisation referred to on this page, and no company described here is presented as a client. Everything on this page is general practice for operating and evaluating a corporate merchandise programme, and nothing in it is legal advice on data protection, employment law or works council matters in any jurisdiction.

This page is written for people analytics, HR data and workforce insights teams evaluating corporate merchandise programmes in Armenia and the wider EAEU. Corp-Merch is an independent B2B merchandise supplier and is not affiliated with, endorsed by or contracted by any employer, HR function or organisation named or implied on this page. The classification of operational records, people records and workforce data described here is general practice and is not legal advice on data protection, employment law, consent or works council obligations in any jurisdiction; those decisions belong with the organisation’s own data governance and legal owners. No monitoring recommendation on this page is a substitute for that review, and nothing here is a binding price or a commitment on any specific outcome.