The most common question in an annual review with a care provider is not how many people visited the website, but whether it delivers anything at all. That can only be answered with figures that match the operation: enquiries from relatives and applications from care professionals. At the end of 2023, 1,100,672 (Statistisches Bundesamt) people in need of care in Germany were supported with the involvement of home care services, and the federal projection expects a shortfall of 280,000 to 690,000 (Statistisches Bundesamt) care workers by 2049 depending on the scenario. Both are markets in which a website can measurably contribute. This article shows which six to eight metrics a care service actually needs, how to count phone calls, what Section 25 TDDDG requires when measuring, and what a quarterly rhythm looks like that a team can keep up alongside its rounds.
Key takeaways
- Visitor counts measure the wrong crowd for a home care service: the catchment area spans a few kilometres while a guide article is read nationwide. Only separate figures for relatives' enquiries and applications from care staff carry meaning.
- Six to eight numbers suffice: submitted contact forms, taps on the phone number, started and completed applications, booked appointments, the landing pages behind enquiries, and the mobile share. Each gets a written definition, an owner and one possible action.
- First contact in home care usually arrives by phone. It becomes countable through the phone number as a clickable link plus one single origin question in the first call, sorted into fixed categories: website, referral, hospital or social services, advert, other.
- Under section 25(1) of the German TDDDG, storing or reading information on the end device requires consent regardless of any personal reference (Datenschutzkonferenz). A counter, logfiles without personal data or a simple counting pixel avoid that access entirely.
- No blanket clearance exists for audience measurement without consent; guidelines 2/2023 of the European Data Protection Board also cover tracking based on the IP address alone. Diagnoses, care levels and names never belong in URLs or page titles.
- Four one-hour reviews a year and exactly one change per quarter keep the analysis reliable. Cost per enquiry comes from quarterly spend on maintenance, hosting and content divided by counted enquiries, compared across four quarters against your own prior year.
Why visitor numbers say little about success
Almost every care service website reports some kind of visitor number. It is the most readily available and at the same time the least meaningful figure in the whole report. A home care service works within a catchment area of a few kilometres, whereas a guide article about applying for a care level is read across the entire country. If visitor numbers rise sharply while the number of first consultations stays flat, nothing has been gained but traffic. Conversely, a site with stable visitor numbers can produce noticeably more enquiries as soon as the phone number, the form and the service area sit in the right places.
The second reason is the underlying population. As of 31 December 2023, 5,688,473 (Statistisches Bundesamt) people in Germany were in need of care, 4,888,882 (Statistisches Bundesamt) of them were cared for at home and 1,100,672 (Statistisches Bundesamt) with the involvement of home care services. Around 15,500 (Statistisches Bundesamt) home care services operate nationwide. For a single provider, only a small, geographically limited slice of that is relevant. A metric that does not reflect that slice simply measures the wrong group of people.
The third reason is the trigger. Enquiries in care follow events: a hospital discharge, a fall, a diagnosis, the holiday of the daughter who normally provides care. These events are not evenly spread across the year. Comparing two months without a time reference easily turns coincidence into apparent effect. That is why every metric needs a fixed comparison period, ideally the same quarter of the previous year, plus a note about special events such as a newspaper advert or an open day.
Two currencies, two funnels
The most expensive measurement mistake
The six to eight metrics a care service really needs
A care service does not need a dashboard with forty values. It needs six to eight figures that one person can collect and explain in twenty minutes per quarter. The selection below has proven itself in the ongoing care of care websites because every single figure is tied to an action. Adding more metrics does not add insight, it adds debate. How these figures translate into concrete improvements is described in detail in our article on winning more enquiries through the care website.
Submitted contact enquiries
Not views of the form, but messages actually sent. The gap between the two numbers shows how many people give up along the way and is the fastest indicator of a form that is too long.
Clicks on the phone number
On smartphones, the tap on a linked number is the most common first contact. It can be counted without recording a conversation, and it is the key bridge between the website and the phone list.
Applications started and finished
Two separate values for the same process. Only the comparison shows whether the career page attracts too little interest or whether the application form loses interested people along the way.
Appointment bookings
Booked consultations are the most reliable precursor to a contract. They also show whether the time slots on offer fit the daily routine of relatives who work full time.
Landing pages of enquiries
On which page did the visit begin that turned into an enquiry? This attribution decides which content is expanded and which merely absorbs attention.
Share of mobile visits
Around 59 per cent (Statista) of global web traffic runs on mobile devices. Your own share decides where testing and improvement have to start first.
Two further values are useful but optional. The first is visibility in the catchment area: how often does the service appear in local search for its own towns, and how many directions requests or calls result from it? How to build that visibility systematically is covered in our article on local SEO for care services. The second is the loading time of the three most important pages, because it affects every other metric. Both belong in the quarterly report, but they steer a decision less often than the six core figures.
- Every metric gets a written definition so that the same thing is counted next quarter
- Every metric gets an owner inside the team, not only a service provider
- Every metric gets a comparison period, preferably the same quarter of the previous year
- Every metric gets a possible action: what would we do if this value dropped by a quarter?
- Metrics without a possible action are removed, however interesting they look
- Enquiries and applications are reported separately and not added into one total
- Events such as adverts, fairs or press coverage are noted with a date and flagged in the report
| Metric | What it answers | Common misreading |
|---|---|---|
| Visits | How much attention the site receives overall | Read as demand within the catchment area |
| Submitted enquiries | How many relatives take the written route | Taken as the total number of enquiries |
| Phone clicks | How often the site triggers a conversation | Equated with the number of calls actually held |
| Applications started | How appealing the job advert is | Counted as applications received |
| Applications finished | How well the form fits everyday life | Attributed to the quality of applicants |
| Appointment bookings | How committed the interest is | Viewed without the no-show rate |
| Landing pages | Which content triggers enquiries | Confused with the most popular page |
| Mobile share | Where improvements take effect first | Dismissed as a purely technical figure |
The counting problem in care: most enquiries arrive by phone
In hardly any other sector is the gap between website and conversion as telephone-shaped as in care. Relatives who have to organise support after a hospital discharge rarely write an email. They call, often in the evening, often from a hospital corridor, often with a smartphone in hand. That is exactly why pure form measurement breaks down here: the success of the website ends up in the phone rather than in the statistics.
The solution is unspectacular and has worked for years. First, the phone number is marked up as a clickable link and the click on it is counted. That is an event on your own page, not an intrusion into the conversation. Second, one single additional question is asked during the first conversation: how did you hear about us? The answer is sorted into four or five fixed categories. Third, both figures are placed side by side each quarter. The comparison is not exact, but it is robust enough to support decisions.
- Present the phone number as a clickable link on every page, visible in the header and at the end of each service page
- Count clicks on the number as an event, without recording phone numbers or conversation content
- Ask exactly one source question during the first conversation and sort the answer into fixed categories
- Keep the categories lean: website, referral, hospital or social services, advert, other
- Record answers in the same list used for enquiries anyway, without any health details
- At the end of the quarter, place phone clicks, form enquiries and source answers side by side in one table
- Note discrepancies instead of explaining them away: they show which channel is being underestimated
The source question belongs in the conversation, not in the form
A side effect of this routine is worth more than the number itself. Anyone who asks about the source during the first conversation regularly learns which phrasing on the website tipped the balance and which question remained open. Those sentences are the best template for the next content revision, because they come from the target group rather than from an editorial meeting.
Keep recruiting metrics separate
Relatives and applicants arrive with different questions, via different pages and at a different pace. Relatives decide under time pressure and look for reassurance; care professionals compare over weeks, often while still employed elsewhere, and look for reliability in rosters, routes and onboarding. Merging both groups into a single conversion figure destroys exactly the differences that make improvement possible. Which content carries a career page is described in our article on staff recruiting through the care website.
| Aspect | Relatives funnel | Applicant funnel |
|---|---|---|
| Trigger | An event, frequently at short notice | Dissatisfaction or relocation, medium term |
| Typical landing page | Service page or guide article | Job advert or career page |
| Preferred contact route | Phone call, often in the evening | Form or short application, often at night |
| Decision time | Hours to a few days | Weeks to months |
| Core metric | Enquiries and phone clicks | Applications started and finished |
| Second metric | Appointment bookings | Drop-off point in the application form |
| Typical bottleneck | Unclear service area | Too many mandatory fields and missing roster details |
The applicant funnel deserves one extra refinement: the drop-off point. If the form is split into sections, you can count how many people reach the second section and how many submit. In practice the break regularly happens in the same place, namely where a cover letter or a CV file is demanded. Removing that hurdle changes the figure visibly within a single quarter. A suitable structure for job adverts and short applications is outlined on our page about career pages and recruiting.
- Evaluate applications per job advert, not only as a total for the career page
- Track applications started and finished as two values; the difference is the actual message
- Count timing as well: applications in care frequently arrive outside office hours
- Track response time as its own metric, measured from receipt to the first personal reply
- Keep health or origin details out of any evaluation, including free-text fields
- Note application sources outside the website separately so the site is neither praised nor blamed for them
Privacy-friendly measurement: what Section 25 TDDDG requires
The legal starting point is clearer than it is often portrayed. Under Section 25 (1) of the German Telecommunications and Digital Services Data Protection Act, storing information on a user's terminal equipment and accessing information already stored there is only permitted with prior consent based on clear and comprehensive information (Telekommunikation-Digitale-Dienste-Datenschutz-Gesetz). Section 25 (2) leaves only two cases without consent: carrying out the transmission of a message, and strict necessity for a digital service expressly requested by the user (Telekommunikation-Digitale-Dienste-Datenschutz-Gesetz). Crucially, according to the German data protection supervisory authorities, it does not matter whether the information relates to an identifiable person (Datenschutzkonferenz).
That shifts the real question. It is not whether measurement is anonymous, but whether the terminal equipment is accessed at all. For a pure count of page views, the guidance issued by the German data protection authorities for providers of digital services explicitly names three routes: incrementing a counter on each request, determining the number of page views from log files without personal data, or embedding a simple counting pixel of the directly requested service that captures no further user data (Datenschutzkonferenz). That is precisely the technical basis of cookie-free measurement in your own hosting.
At the same time, the authorities deliberately refuse to issue a convenient blanket clearance. The guidance makes no general statement as to whether reach measurement may be deployed without consent (Datenschutzkonferenz). The reasoning is sound: the usual labels are too vague, and a ruling would at best be possible for a precisely defined configuration and purpose; it would no longer hold as soon as further information about users or an additional evaluation result is added (Datenschutzkonferenz). Anyone running a lean count today and enabling an extra module tomorrow is in a different legal position.
Even the simple measurement of visitor numbers is therefore not per se to be classified as part of the basic service, but depends on the specific purpose pursued in each case.
The supervisory test criteria in brief
From these criteria a selection rule can be derived that works without legal training. It is not maximum data depth that decides, but the question of which figure triggers an action. Data minimisation and purpose limitation are not brakes but selection criteria: what is not collected needs neither explanation nor protection nor deletion. The same standard applies to forms and application journeys, as described in our article on data protection on care websites and secure forms.
- Run measurement in your own hosting so that no data flows to third parties
- Work without cookies and without recognition identifiers so that the terminal device is not accessed
- Build no profiles of individuals and follow no user journeys across several services
- Collect only the six to eight agreed metrics and record the purpose of each metric in writing
- Keep the retention period for raw data short and document deletion periods in the record of processing activities
- Describe the measurement in the privacy notice in plain language instead of hiding it behind jargon
- Coordinate with the data protection officer before going live and document the decision
Where supervisory practice and the European Data Protection Board diverge
Anyone relying solely on German supervisory practice misses a second layer. The European Data Protection Board adopted version 2.0 of its Guidelines 2/2023 on the technical scope of Article 5(3) of the ePrivacy Directive on 7 October 2024 (Europäischer Datenschutzausschuss). They cover five use cases: URL and pixel tracking, local processing, tracking based on IP only, intermittent and mediated reporting from connected devices, and unique identifiers (Europäischer Datenschutzausschuss). The German guidance explicitly presents itself as a supplement to these guidelines (Datenschutzkonferenz).
The third use case carries the most weight. According to the guidelines, Article 5(3) can already apply where the IP address alone is used to track navigation; what matters is whether that information originates from the terminal equipment. Unless a provider can rule that out, it has to take all the steps required by Article 5(3) (Europäischer Datenschutzausschuss). At the same time, the Board makes clear that applicability of the provision does not automatically mean consent has to be collected; in each case it must be assessed whether an exemption applies (Europäischer Datenschutzausschuss).
The divergence is therefore less dramatic than some summaries suggest, but it is real: German supervision describes log-file based counting without personal data as a viable route for a pure reach count (Datenschutzkonferenz), while the European view frames the technical scope more broadly and includes access to an IP address originating from terminal equipment (Europäischer Datenschutzausschuss). For a care service this is not an academic question but a design requirement: anyone who does not store IP addresses permanently, but uses them only for delivery and discards them immediately, reduces the exposure under both readings.
Anonymisation is a procedure, not a label
Sensitive contexts: what does not belong in a URL
Care is a health context, and health data falls under Article 9 of the General Data Protection Regulation as a special category of personal data with a general prohibition on processing and narrowly framed exceptions (Datenschutz-Grundverordnung). In web analysis the risk rarely comes from the counting itself, but from side routes: descriptive page names, parameters in the address bar, form fields whose content accidentally ends up in a report. A page title that links a diagnosis or a form of care to a name is already a problem before any measurement takes place.
- Avoid diagnoses, care levels or forms of care as parameters in the address bar
- Avoid names, phone numbers or addresses in page titles, file names or form URLs
- Name confirmation pages neutrally, for example enquiry received rather than enquiry ventilated intensive care
- Keep free-text form fields out of reports as a matter of principle
- Store application attachments separately from the analysis area and give them a deletion period
- Evaluate internal search terms only in aggregate, because search queries can be highly personal
- Avoid screenshots and screen recordings in care contexts, because they capture content unintentionally
The quiet route through error pages and search parameters
A realistic quarterly rhythm instead of a permanent dashboard
A dashboard that would have to be opened daily stops being opened after two weeks in a care service. Daily life consists of rounds, rosters and sick leave, not curve analysis. That is why a quarterly rhythm has proven itself: four appointments a year, about an hour each, with a fixed agenda and exactly one change as the outcome. It sounds slow, but it is faster than the usual cycle of hectic activity followed by standstill.
The reason lies in the statistics of small numbers. A care service with thirty enquiries per quarter cannot draw conclusions from a single week, because two extra calls double the curve. Only across a quarter does a value emerge that is robust enough to justify a change. And only if exactly one change is implemented per quarter can its effect be attributed at all. Two simultaneous changes produce a figure that nobody can explain afterwards.
- Week 1: compile the quarter's figures, six to eight values, the same period as the previous year
- Week 1: note special events such as an advert, a press article or a website outage
- Week 2: formulate a single hypothesis, for example that applications fail at the file upload
- Week 2: test the hypothesis against the source answers from first conversations
- Weeks 3 to 4: implement exactly one change and document it with a date
- Weeks 5 to 12: do not fine-tune further, so that the comparison period stays clean
- End of the quarter: measure again, record the result in one sentence and derive the next hypothesis
- Once a year: review the metric definitions and delete values that have become obsolete
One hour per quarter
Fixed dates in the calendar, one responsible person, one agenda. If the appointment is not scheduled, it moves permanently into the realm of good intentions.
One change per quarter
Only one change allows a clean attribution of the effect. All other ideas go on a list and are reviewed in the following quarter.
Same counting, same period
Metric definitions stay stable across the year. If a definition changes, it is flagged in the report, otherwise an apparent effect is created.
In practice, this rhythm belongs in ongoing website maintenance for care services, because updates, security patches and content work already converge there. If content is reviewed in the same session, no extra effort arises, only an agenda with two items. There is no shortage of reasons for content work, for example presenting inspection results, which we cover in our article on the 2026 care quality inspection on your website.
Cost per enquiry and per hire, calculated simply
In the end comes the question every provider cares about: what does a won enquiry cost, and what does a hire cost? The calculation is deliberately coarse, because a falsely precise figure does more harm than good. It reads: the cost of website maintenance and hosting in one quarter divided by the number of enquiries counted in that quarter. The same principle applies to applications and hires. Benchmarks from other sectors do not help here; what matters is how your own figure develops across four quarters.
- Costs include maintenance, hosting and content work, not the one-off cost of the build
- Enquiries include submitted forms plus the calls attributed to the website during first conversations
- One-off events such as a trade fair are noted in the report rather than silently included
- For recruiting, the same calculation uses completed applications and actual hires
- The value is viewed across four quarters, not judged as a snapshot
- The comparison runs against your own previous year, not against figures from other sectors
| Question in the board meeting | Weak answer | Robust answer |
|---|---|---|
| Does the website deliver? | We had more visitors than last year | We had more submitted enquiries and more phone clicks than last quarter |
| Is the career page worth it? | The job advert was read often | More of the started applications were completed than before |
| Why are enquiries falling? | Our ranking dropped | The landing pages of enquiries have shifted, a piece of content is missing |
| What does an enquiry cost us? | That cannot be said | Quarterly cost divided by counted enquiries, tracked across four quarters |
| Is the measurement lawful? | We have a banner | The terminal device is not accessed and the assessment is documented |
| Who takes care of it? | The agency will get in touch | A named person in the team, four dates a year, one sentence of minutes per quarter |
Once this calculation is in place, the next step is editorial rather than technical. A care website with a clear structure produces better figures than any additional metric, because it shortens the paths relatives and applicants have to travel. On mobile devices that applies twice over; which details decide there is shown in our article on mobile optimisation for relatives. Billing and documentation topics also feed back into the website, as our article on the electronic proof of service shows. If you want to set the figures up properly for the first time, our overview of services for care websites provides the framework, and the scope can be clarified in an initial conversation about measurement and reporting.
Sources and studies