The Tow Center guide by Hansen and Goligoski (2018) separates three models. Subscription is a transaction: payment buys access to a product. Donation is a gift to a cause, with nothing expected in return. Membership sits between them. Members give money, and often time, knowledge or contacts, because they believe in the organisation's purpose, and the organisation in turn opens itself to them. The Membership Puzzle Project developed this idea into practical guidance and stressed that the relationship must be designed before the payment page (Membership Puzzle Project, 2020). The underlying notion is close to what Lewis et al. (2014) call reciprocal journalism, in which journalists and audiences exchange benefits over time and trust accumulates through the exchange.

Comparative survey work finds that only a minority of people in most countries pay for online news, that payment is associated with age, interest in news and an existing habit of paying for print, and that those who do not pay often say free alternatives are sufficient (Fletcher and Nielsen, 2017). A systematic review of the drivers of willingness to pay reaches similar conclusions and notes how much of the evidence rests on stated intention and not on observed behaviour (O'Brien et al., 2020). Studies of regional newspapers add two cautions. Introducing a paywall can reduce traffic and with it the title's civic reach (Olsen et al., 2020), and stated intent to pay for a digital edition is weak when the same content has long been free (Chyi, 2012). A newsroom planning membership should therefore expect a small core of supporters, should measure what people do and not what they say, and should watch for effects on readership as well as revenue.

A regional title has one market and one launch. It cannot randomise towns. Two traditions help. Within a website, visitors can be allocated at random to alternative versions of a message or page, and the conditions under which such online controlled experiments give trustworthy answers are well described (Kohavi et al., 2020). Where randomisation is impossible, quasi-experimental designs such as interrupted time series and staggered introduction offer weaker but still disciplined inference, provided the threats to validity are named in advance (Shadish et al., 2002). Qualitative evidence from interviews and documents is best organised as a case study with a stated protocol (Yin, 2018).

Before anything is offered, the newsroom records where it stands. The audit covers revenue by source for at least the preceding two years, so that seasonal patterns are visible; audience measures, separating occasional visitors from the regular local readers who are the realistic market for membership; existing direct relationships such as newsletter lists and print subscribers; and the cost of the journalism the scheme is meant to sustain. Picard (2011) gives the standard account of how media revenues and cost structures should be analysed. The audit fixes the comparison against which later change is judged, and it often shows that the newsroom does not know how many loyal readers it has.

Published guidance is consistent that membership fails when it is designed from the publisher's needs alone (Membership Puzzle Project, 2020). The listening phase combines a short survey of newsletter recipients and regular visitors, semi-structured interviews with readers chosen to differ in age, location and attachment to the title, and one or more open meetings. The questions concern what readers value in the coverage, what they miss, how they would want to be involved and what they would regard as a fair request. Stated willingness to pay is recorded and treated with reserve, for the reasons given above. The output is a written summary agreed by editorial and commercial staff.

The offer is now written down: the appeal, the price points, what members receive and what remains open to all. Alongside it the newsroom writes its hypotheses in plain terms, for example that an appeal based on the title's civic role will convert more regular readers than one based on member benefits, or that members who attend an event will renew at a higher rate than those who do not. It also writes the rules by which it will judge the trial: the level of recurring income by a stated date that would justify continuing, the level below which the scheme would be redesigned or closed, and the editorial time it is prepared to spend. Fixing these before launch is the single most important discipline in the framework, because it prevents the result from being reinterpreted to suit the hopes of those who ran it.

The launch is staged so that comparisons are possible. On the website, regular visitors can be allocated at random to alternative appeal messages, and the allocation and outcome logged. Guidance on online experiments warns against stopping a test as soon as a difference appears, against running many variants on little traffic, and against judging on clicks when the outcome of interest is payment (Kohavi et al., 2020). A regional site has limited traffic, so the newsroom should test few alternatives, each expressing a real difference in approach, and accept that small effects will not be detectable. Channels that cannot be randomised, such as a print appeal or an editor's letter, are introduced at different times so that their effects can be distinguished in the daily series of joins.

A launch produces a burst of goodwill. What sustains a newsroom is renewal. Members are therefore grouped by the month in which they joined and each group is followed separately, since an overall retention figure hides the difference between early enthusiasts and later joiners. Fader and Hardie (2007) show why retention observed in the first periods of a contractual relationship should not be projected forward as a constant rate. Reasons for leaving are collected at cancellation, and participation is recorded alongside payment: attendance at events, responses to call-outs, contributions of knowledge to reporting.

At the date fixed in Phase 3 the newsroom compares outcomes with its rules and records a decision to continue, change or stop. The review draws on the quantitative record and on a second round of interviews with members, lapsed members and staff, analysed against the case study protocol (Yin, 2018). It should state what cannot be concluded as plainly as what can.