What is purchase frequency (F) in a target group profile?

‘This audience purchases regularly’ leaves many questions unanswered. Does that mean three orders per month, an annual renewal or several products in one order? Purchase frequency (F) specifies how often a defined group buys within a stated period. This makes purchase behaviour comparable, provided you also record what counts and which data is available.

Purchase frequency (F) describes in the target group profile how many relevant purchases per defined unit have been recorded or reported within an explicit observation period, stating the counting rule, source and data coverage.

This is Neurofactor’s practical profile definition. The unit may be a person, household, customer account or organisation. A recorded number is a count. A statement such as ‘roughly monthly’ is a reported purchase rhythm. Keep the provenance and precision of each visible.

For an entire audience, describe a distribution or a clearly defined average. State whether it covers all accounts studied or only accounts that purchased during the period. Without that denominator, average purchase frequency is difficult to interpret.

When counting orders, one order containing five products is one purchase. Five valid orders may represent five purchases. If you count purchase days, several orders on the same day may instead be combined. Choose the rule beforehand and label the measure accordingly.

Do not confuse ‘repeat’ with ‘every purchase after the first row in my export’. An existing customer may have purchased before the observation window. In that case, even the first order within the window can be a repeat purchase.

F stands for frequency in RFM: recency, frequency and monetary value. Counting purchases is not a separate psychological theory with a single established inventor. The relevant research question concerns how purchase history is defined and interpreted.

In their RFM research, Fader, Hardie and Lee use the number of repeat purchases within a specified period. Check which purchases a frequency measure includes. Fader, Hardie & Lee, 2005.

Dick and Basu approach customer loyalty as the relationship between relative attitude and repeat patronage. Their conceptual framework therefore extends beyond a transaction count alone. Dick & Basu, 1994.

Our practical application is to measure the count carefully and investigate preference or motivation separately. The documentation rules below are a profile workflow, not a validated loyalty scale or a universal formula for future purchasing.

Start with the question the profile needs to answer. Are you describing purchases from one supplier, purchases across a category or the replacement cycle of a particular product? These questions require different data.

For subscriptions, specify whether you count payments, renewals or new contractual decisions. Twelve automatic payments are not necessarily twelve separate deliberate choices. Similarly, one framework agreement may generate several call-off orders: the counting rule determines which question you answer.

1. Define the population. Determine which units belong in the analysis, including those that purchase nothing in the window.
2. Define the period and event. State which purchases count and how the start and end boundaries work.
3. Check coverage. Examine missing channels, duplicate accounts, order lines and changes in recording.
4. Count per unit. Count unique valid purchase events within the window. Keep the count and observation duration separate.
5. Choose an appropriate summary. For example, show account-level counts or a distribution across frequency bands. Include the denominator with an average.
6. Handle unequal observation durations explicitly. Compare suitable cohorts or report a descriptive count per unit of time. Make limited comparability visible.
7. Separate zero from unknown. Zero means no purchases observed in a defined window with sufficient coverage; unknown means a complete count cannot be established.
8. Check interpretation. Consider frequency alongside recency, product context and purchase rhythm before recommending a next step.

For consistently defined units with complete observation, this workflow uses:

Average purchases in the window = total valid purchases ÷ number of units in the chosen population.

A descriptive pace per 30 days can be calculated as purchase count ÷ observation days × 30. This standardises the time unit, but not automatically season, customer stage or purchasing opportunity. Extending it to a year requires further assumptions and is an extrapolation, not a measured annual frequency.

Add an F score only with documented boundaries, reference group and direction. ‘High’ may mean high within one customer database without the group purchasing more often than other market groups.

Ask, for example: ‘How many separate purchases did your organisation make in this category between these dates?’ Then ask what the person counts as a purchase and what supports the answer. A budget round is not necessarily an order; an order may also occur outside the normal cycle.

A product-independent audience description does not provide a universal purchase frequency. Define the category or decision first. If it is unspecified, a reported decision cycle may be useful context while F remains unknown.

Fictional example — no customer data or sector benchmark. We count unique paid, non-cancelled orders in one product category from one supplier. The window runs from 1 January 2026 to 1 April 2026, excluding 1 April: 90 days.

The three fully observed accounts A, B and C generated nine orders. Their average is 9 ÷ 3 = 3 orders per account in 90 days. If the question concerns only the two purchasing accounts, it becomes 9 ÷ 2 = 4.5 orders per purchasing account in 90 days. Both calculations are correct, but answer different questions.

B and D each have three orders, but D was observed for less time. Their descriptive pace per 30 days is respectively 1 and 3 orders. That difference does not establish that D will purchase three times as much over the coming year.

A possible profile sentence reads: ‘In the fictional window, three fully observed accounts averaged three orders each, including one account with no purchase; one account with shorter observation and one with missing data are reported separately.’

An associative target group is a bounded group of people whose similar associations guide decisions within a specific choice context.

Purchase frequency helps describe behaviour around that context. The count alone does not explain why someone purchases repeatedly. A recurring purchase might, for example, relate to a working agreement, availability or a deliberate preference. These are possible explanations to investigate, not findings about this audience.

Ask which associations mattered at concrete purchase moments and how they influenced the choice. People who buy equally often do not thereby constitute an associative target group. Conversely, people with similar decision-related associations may have different purchasing opportunities.

Do not infer BIS, BAS or delay-discounting scores from frequent or infrequent purchasing. Keep the behavioural measure alongside separately investigated motivation, associations and time preference.

• Leaving out the period. ‘Six times’ becomes interpretable only with an observation window.
• Including buyers only in the denominator without saying so. State explicitly whom the average describes.
• Assessing new accounts as though they were observed equally long. Retain entry date and observation duration.
• Counting order lines or payments as decisions. Choose the event that matches the research question.
• Automatically subtracting the first export row. Repeat purchases need a substantive definition of the initial purchase.
• Converting a reported rhythm into measured counts. Preserve self-report and uncertainty.
• Equating high frequency with loyalty or profit. Investigate preference and economic value with suitable additional data.

With transaction data:

‘Within [purchase definition and scope], [defined population] makes [distribution or average with denominator] purchases in [period], based on [source and coverage]; [shorter observation or missing information] is reported separately.’

Without transaction data:

‘The audience reports [purchase rhythm or count with period] for [category and context]; this is [self-report/estimate], while recorded purchase frequency is [unknown/unavailable].’

Store count, period, observation duration and score version separately. Make the denominator visible in every language. For a new research round, update the period and provenance and preserve the earlier position as a revision. More precise-sounding wording does not make up for a missing source.

Purchase frequency becomes useful when readers can see what was counted, for whom, over which period and with what coverage. Separate measured counts, reported purchase rhythm and forecasts. F can then support the profile without filling gaps in data or motives.