Installed base and product analytics for MedTech, built on an honest separation between the devices you can observe and the devices you are assuming are still where you left them.
A device ships to a distributor, who sells it to a hospital group, which installs it at one site and later moves it to another, sells it on, lends it to a partner facility, puts it in storage during a refurbishment or quietly decommissions it. None of those events reliably reach the manufacturer. The shipment record is accurate and progressively less informative, and the installed base figure derived from it describes where devices went rather than where they are.
A connected device reporting daily is an observation. A unit shipped four years ago with no service contract, no connectivity and no contact since is an assumption that it still exists, is still installed, and is still at the address on the shipping record. Both appear in the installed base figure as one unit, and almost no organization reports the split. The proportion that is assumed rather than observed is the single most useful thing to know about your installed base number, and it is rarely calculated.
The second problem is that the events which would correct the record happen at the customer and nobody has an incentive to tell you.
The record decays from the moment of shipment
Every transfer, relocation, resale and decommissioning happens without notifying the manufacturer.
Distributors sit between you and the truth
Where an indirect channel exists, the shipping record ends at the distributor and the rest is inference.
Â
Connectivity divides the fleet into two populations
Connected devices are observed continuously and unconnected ones are assumed, and analysis rarely distinguishes them.
Â
Decommissioning is almost never reported
Devices leave service quietly, which inflates the installed base and deflates every rate calculated from it.
Telemetry coverage introduces a bias nobody states
Performance analytics built on connected devices is frequently interpreted as representative of the whole fleet, and the connected population is not a random sample.
Â
Utilization is the most valuable and least available signal
How much a device is actually used drives reliability, commercial and clinical questions, and most fleets cannot report it.
Installed base intelligence is built from evidence of different strengths: continuous telemetry, a service visit last month, a warranty registration from six years ago, and a shipment to a distributor who does not report onward. Treating those as equivalent produces a confident number nobody can defend. Grading them produces a number with stated uncertainty, which is considerably more useful.
Step 1
Define eight rules: installed, active, distributor inventory, loaners, replacements, transfers, retirement and unknown status.
Step 2
Grade the evidence behind each device, since telemetry, service contact and shipment record are not equivalent.
Â
Step 3
Set a recency window per status, because a device last confirmed three years ago is not confirmed now.
Â
Step 4
Resolve device identity across systems, which is covered in depth on our data platform page.
Â
Step 5
Capture the correcting events, particularly service visits, which are the main route by which the record improves.
Â
Step 6
Model location and ownership as time-bound, since both change and historical questions depend on knowing when.
Step 7
Establish utilization where connectivity allows, and state clearly where it does not.
Step 8
Report rates against a defined denominator, naming which population the rate is calculated on.
Step 9
Design the distributor data position, since indirect channels determine how much of the base is knowable at all.
Step 10
Monitor the observed proportion over time, because it degrades unless something actively maintains it.
Your installed base accuracy tracks your service coverage.
The record improves when somebody from the company touches the device, and the main mechanism for that is a service visit. Devices under contract are confirmed regularly and are the best-understood part of the fleet. Devices outside service coverage drift, and the drift is invisible because nothing reports it. That relationship means installed base quality is partly a commercial question about service penetration rather than purely a data engineering one, and improving the data frequently means improving contact rather than improving the pipeline.
Device identity resolution and cross-system integration are covered on our MedTech Data Platform Engineering page. These are the capabilities that sit on top of that foundation.
Establish and Qualify
Device Status Modelling
A defined lifecycle from shipped through installed, active, dormant and retired, with the transitions and the evidence each one requires.
Evidence Grading
Each device carrying the strength and recency of what is known about it, so analysis can restrict to observed units where it matters.
Location and Ownership History
Time-bound records of where a device has been and who held it, since both change and historical questions depend on it.
Channel and Distributor Modelling
What is knowable through indirect channels, and the explicit gap where it is not, rather than inference presented as fact.
Analyse
Reliability and Failure Analytics
Failure rates calculated against a stated denominator and a stated population, covered further on our surveillance page.
Utilization Analytics
How much devices are actually used, where connectivity supports it, which is the most commercially valuable signal most fleets lack.
Service Performance Analytics
Intervention frequency, parts consumption, repeat visits and time to resolution, joined to device configuration and age.
Cohort and Lifecycle Analysis
Performance by build period, generation, configuration and age, which is where product improvement signals actually appear.
Maintain and Serve
Record Correction Workflows
Capturing what service, support and commercial teams learn, since they encounter the truth and usually have nowhere to put it.
Coverage Monitoring
Tracking the observed proportion over time, because it decays and nothing announces the decay.
Analytics Delivery
Governed datasets for quality, service, commercial and engineering consumers at the rigour each one requires.
Field Action Support
Identifying affected units and their current locations, which is the installed base question with the sharpest consequence.
What CaliberFocus does, and does not do?
Device identity establishes the foundation, while time adds historical context to understand failures and complaints. We distinguish observed from assumed devices, making data confidence and its impact on analytics visible. When gaps stem from service coverage or channel structure, we identify the commercial fix. Quality and regulatory teams retain responsibility for field action scope, reportability and affected populations.
Installed base data serves functions whose questions look similar and whose correct denominators differ. The third column is what each one actually needs counted, and using the wrong population is how two teams produce different rates from the same data and both are right.
| Consumer | What they ask | The denominator they need |
|---|---|---|
| Quality and post-market | How often does this fail in the field | Devices confirmed in service during the period, not units ever shipped |
| Service operations | What will we need to support next year | Devices under contract plus those likely to generate demand |
| Commercial | What is our footprint and replacement opportunity | Devices present at customers, by age and configuration |
| Engineering | Is the design change working | Devices built to each revision, with comparable exposure time |
| Regulatory | Which units are affected by this action | Units matching the criteria, with current location, regardless of status |
Shipped, units transferred through distribution. Installed, units with evidence of installation. Active, units meeting an agreed operational definition. Connected, units communicating within a defined period. Service-covered, units with valid warranty or contract. Retired, units confirmed removed from service. These overlap, each serves a real question, and the platform should preserve the definition and the evidence supporting every classification rather than forcing one number.
Each system knows something about the device and none knows everything. The third column is the strength of what it establishes, which is the property that should travel with the data and almost never does.
| Source | Relative volume | What it implies architecturally |
|---|---|---|
| Device telemetry | Dominant, and growing with the fleet | Tiering, summarization, late arrival handling and retention policy |
| Manufacturing and historian | Large, and growing with production | Industrial boundary, time alignment and deliberate resolution reduction |
| Quality systems | Small, and the most consequential | Separate governance, controlled access and traceability to the source |
| ERP and supply chain | Moderate and transactional | Reconciliation to the system of record rather than a parallel truth |
| Service and field systems | Moderate, frequently semi-structured | Extraction and device association, covered on our document AI page |
| PLM and engineering | Small and slow changing | Revision applicability mapped to units actually built to it |
A device that stopped reporting is telling you something. Decide what.
When a connected unit goes quiet it may have been decommissioned, moved behind a firewall, had its connectivity disabled by IT, failed, or simply been switched off for a refurbishment. Those have entirely different implications for the installed base, for reliability analysis and for service. Most platforms treat a silent device as an absence of data rather than as an event worth investigating, which means the single richest source of installed base correction is discarded as noise. Classifying dormancy, and following up on it, improves the record more than any integration project.
Installed base data is used commercially most of the time and regulatorily when it matters most. A field action requires identifying affected units and reaching whoever holds them, and the quality of that answer is determined by work done years earlier. An installed base maintained adequately for forecasting may be inadequate for notification, and the difference becomes visible under conditions where there is no time to fix it.
Traceability
- Affected units identifiable from build characteristics, with current holder and location where known
- Evidence strength and recency attached, so a notification list distinguishes confirmed from presumed
- Historical location and ownership retained, since the question may concern a past period
- Lineage to the source system and record for every status claim
Governance
- Device status definitions owned by a named function and used consistently across quality, service and commercial
- Denominator definitions documented per analysis type, since the correct population differs by question
- Engineering controls capable of supporting the frameworks that apply, which may include FDA regulation, ISO 13485 and electronic record requirements
- Applicability, interpretation and validation strategy determined by your quality and regulatory functions rather than by us
Security and privacy
- Customer location and configuration data treated as commercially sensitive under the relevant agreements
- Patient-associated device data handled to the obligations your counsel defines, covered on our privacy page
- Distributor data used within the terms of the channel agreement, which frequently restrict onward use
- Access governed by purpose, since commercial targeting and quality investigation are different grants
Quality of the record
- Observed proportion monitored and reported rather than assumed stable
- Correction events captured from service and support, with a route for field staff to record what they find
- Decommissioning actively sought rather than waited for, since it is the least reported event and the most distorting
- Reconciliation against customer asset records where the relationship permits it
Two teams calculating the same failure rate from the same data will produce different numbers if one counts shipped units and the other counts devices in service, and both will defend their figure. The disagreement is not about the data, it is about the population, and it persists because the denominator definition is rarely stated alongside the result.
Observed proportion
What share of the base is confirmed within a recency window, which bounds the reliability of everything else.
Status accuracy
Whether devices recorded as active are active, testable by sampling against service or customer contact.
Location currency
How old the location evidence is, since a field action depends on it and it decays silently.
Denominator consistency
Whether the same population is used for the same metric across functions and periods.
Exposure calculation
Whether rates account for time in service rather than treating all devices as equally exposed.
Correction throughput
Whether record corrections are being captured at a rate that keeps pace with real-world change.
Installed base programmes are usually reported as records integrated and dashboards delivered. The outcomes that matter are whether the observed proportion is known, whether rates use a stated denominator, and whether a field action could actually be executed.
| Category | What we measure | Why it matters |
|---|---|---|
| Can every department agree | For one product family, whether manufactured, shipped, installed, active, with distributors, transferred, replaced, retired, connected and service-covered counts reconcile, current and historical locations are identifiable, firmware at a past date is retrievable, failures connect to the correct population, and discrepancies between departments can be explained | Inconsistent answers point to identity, lifecycle and definition gaps rather than to reporting |
| Observed proportion | Devices confirmed within a recency window against the total | Bounds the reliability of every rate and is almost never calculated |
| Field action readiness | Units identifiable and reachable in a rehearsed scenario | The honest test of installed base quality, under the conditions that matter |
| Denominator consistency | Whether functions use the same population for the same metric | Where two correct answers disagree and nobody can resolve it |
| Exposure-adjusted reporting | Rates using device-years rather than device counts | Determines whether generation comparisons mean anything |
| Correction throughput | Record corrections captured per period against expected real-world change | Whether the record is keeping pace or quietly decaying |
We will calculate your observed proportion, separate confirmed devices from assumed ones, review the denominators in use across functions, rehearse a field action against current data, and assess whether dormancy is being investigated or discarded.
Bring us one product family, your current installed base figures and a challenge your teams cannot answer consistently.
Security & Compliance
