Practice software: why a scattered stack costs more than it looks

The practice software stack nobody assembled

Ask a practising therapist which services they use, and you will almost certainly get a history rather than a list. The calendar appeared when there were more than five clients. The notes stayed wherever they started. Worksheets sit in a folder on a laptop, payments land on a card, conversation happens in a messenger, and questionnaire results live in a third place or nowhere at all. None of these was a decision. Each was a hole plugged on a particular day — sometimes with a service, sometimes with the first practice app in the search results, sometimes with nothing at all.

Two typical stacks come out of this. The first is paid: four or five subscriptions, each inexpensive, each covering its own piece. The second is improvised: a notebook, a spreadsheet, a messenger and memory. Usually both coexist in the same practice, and that is normal — it is how most private practice actually runs.

The problem is not that one of them is bad. The problem is that both carry a price, and neither arrives as an invoice. In the first it is smeared across subscriptions that stopped being noticed long ago. In the second it sits outside money altogether.

The subscriptions you stop seeing

A subscription is designed to stop being noticed: the charge is automatic, the amount is small, the decision was made once. Individually each looks reasonable. Together they form a standing cost of the practice that is rarely counted as a whole — unlike room rent or supervision, which are visible and get discussed.

It is worth noticing what the money goes to. The best-evidenced function in the whole set is an ordinary appointment reminder. A systematic review of 29 studies of hospital outpatient appointments (Hasvold & Wootton, 2011) shows a relative reduction in non-attendance of 34%; a manual phone call beats an automated message (39% against 29%), and the cost per reminder came to about €0.41 across roughly half of the included studies — the review made no formal cost-effectiveness assessment, and these are 2011 prices. What works is the fact of the reminder, not the ingenuity: a cluster-randomised trial across 38,945 appointments (Teo et al., 2023) tested letters with behavioural nudges and found no effect (OR = 1.20; 95% CI 0.90–1.60; p = 0.21).

Next to it sits the most common sales argument — "a therapist spends N hours a week on paperwork". A scoping review of 35 studies of documentation burden among physicians and nurses working in electronic health records (Moy et al., 2021) found that no standard, validated measure of that burden exists: fifteen different measurement characteristics and no consensus. Any specific figure is the result of one survey, not a property of the profession, and there is nothing to check it against.

"Standard and validated measures of documentation burden are lacking… there appears to be no consensus on the best approach nor degree of rigor to study documentation burden."Moy et al., JAMIA, 2021

Notebook, messenger, spreadsheet

The second stack looks free, and in the expense column it genuinely is zero. But improvised tools do not stop being processing of personal data because nobody pays for them. Information about a client's mental state is a special category of personal data, and a messenger thread, phone notes and a laptop spreadsheet fall under the same requirements as any platform.

The duties sit with the practitioner as the operator, not with the service, and they do not disappear when there is no service. The storage location is a separate requirement: the primary database holding Russian citizens' data must sit in Russia — recording, storing and retrieving it outside the country is prohibited (152-FZ, art. 18(5); the localisation requirement itself has applied since 1 September 2015, this wording since 1 July 2025). Cross-border transfer remains possible afterwards, but requires notifying the regulator in advance — in practice that means "almost never". What exactly needs to exist on paper — contract, consents, intake form — is covered in the article on the paperwork of private practice.

This layer has no effectiveness metric: no study will show that careful record-keeping improved a therapy outcome. It does have a price, and it is not measured in effect sizes. The database of the Finnish psychotherapy provider Vastaamo was copied back in 2018; the story became public in 2020 when the extortion began, and the company itself announced ≈36,000 patients. The extortionists went not only to the clinic but to the clients themselves. The hacker's sentence covers 9,231 counts of disseminating private-life information and 20,745 counts of attempted blackmail. The company was declared bankrupt in February 2021, and the supervisory authority's €608,000 GDPR fine arrived that December, by which point it no longer existed.

The cost of the seams

The main cost of a scattered stack is neither the subscriptions nor the risk, but the seams between the tools. The same details get entered twice: the client's name in the calendar, again in the notes, again in the payments sheet. No tool knows what its neighbour knows, so the only place where the stack becomes a whole is the practitioner's head.

• The client's trajectory never assembles: seeing what happened over six months means opening three places and matching them by hand
• Questionnaire results live apart from notes, so progress exists only as an impression
• When a client is referred on (with their consent), or when you take leave, the context is handed over as a retelling — there is nowhere to export it from
• A withdrawn consent or a deletion request has to be honoured in each tool separately, and forgetting one is easy

It should be said plainly: this is reasoning, not a research finding. Studies measuring the cost of fragmentation specifically in private psychological practice could not be found, and anyone quoting you a number here invented it. But every item on that list can be checked against your own stack in five minutes, and the check is usually uncomfortable.

What has to live in one place

Of all the layers, only one has a demonstrated effect on outcome rather than on convenience. One of the largest meta-analyses of progress feedback by patient count (de Jong et al., 2021) pooled 58 studies and 110 effect sizes across 21,699 patients: d = 0.15 (95% CI 0.10–0.20), d = 0.17 for off-track cases, plus a small favourable effect on dropout (OR = 1.19; 95% CI 1.03–1.38). The effect is small and deserves an honest name: systematic measurement helps, it does not transform therapy.

It has limits too. A meta-analysis of PCOMS found an overall effect of 0.27, present in counselling (0.45) but not in psychiatric settings (0.10; p = 0.144), and its authors point to possible allegiance bias (Østergård, Randa & Hougaard, 2018). In adolescents aged 11–18, a Cochrane review of six RCTs across 1,097 participants concludes the evidence is insufficient and its certainty "very low" (Bergman et al., 2018).

From this follows a practical criterion for assembling a stack. Measuring progress only means anything as a continuous line: one point is not a trajectory. So questionnaire results, session notes and working documents have to share one client history, or the single layer with a demonstrated effect on outcome simply does not function. A calendar, video calls and taking payment, by contrast, have nothing to do with that line and live perfectly well apart — their separateness costs nothing.

Soveria is built around that criterion: assessments with automatic scoring, between-session progress, clinical notes and a library of working documents sit on one client timeline. What it does not do is bring clients: it is not a marketplace or a matching service, and if the current problem is finding your first ten people, it will not be solved here.

Choosing a service, not a collection

Rebuilding a stack in a single decision is neither possible nor necessary. But taking it apart once is worth doing — if only to see what it consists of and what it costs.

• List every tool and add up a year of subscriptions. It is the only figure here you can know exactly
• Mark where client data is stored. Messengers and personal notes count
• Check how many places you would have to stop processing in if consent were withdrawn — and delete from, where keeping the data is no longer needed (152-FZ, art. 21(5): up to 30 days unless the contract says otherwise). More than one place is a seam
• Ask yourself where a given client's last six months live. If the answer is "in three places and in my head", that is the layer to consolidate first

A scattered stack is not a sign of poor organisation. It is the natural result of a practice growing faster than decisions about tools were made. But it has a price, and it is not in the subscriptions: it is paid in time spent reconciling the same details by hand, and in the absence of the continuous line that makes measurement worth anything. The best-evidenced function in the whole set costs about €0.41 in 2011 prices. The rest is decided not by price, but by whether your tools add up to one client history or four different ones.