TL;DR
Digital friction is the time that disappears between tools rather than inside them. Waiting for access, finding the right document, repeating something a system already knows. It is largely invisible on any dashboard, and the cheapest way to find it is to ask rather than to instrument.
- It lives between systems, which is why no single tool reports it.
- Waiting is the biggest category, and it is the easiest to fix.
- Ask, do not instrument. A survey of twenty people beats a platform.
- Not all of it is worth fixing. Some friction is a control working.
- The same complaint from three different teams is the signal.
- Day one and the first week concentrate more of it than any other period.
Where digital friction actually accumulates
Four categories, and they are not equally expensive.
| Category | What it looks like | Usually worth fixing? |
|---|---|---|
| Waiting for access | Cannot start until somebody approves | Yes, almost always |
| Finding things | Three versions, none dated | Yes, and it is cheap |
| Re-entering data | Typing what a system already holds | Sometimes, check the volume |
| Switching context | Six tools for one task | Rarely, and the fix is usually worse |
Waiting dominates. It is also the one people under-report, because waiting does not feel like work being lost, it feels like a normal part of the day.
Best tools for HR Operations
Why dashboards miss it
Because every tool reports on time spent inside itself. The gap between closing one and being able to open the next belongs to nothing, so it appears in no report, and the aggregate of those gaps is the bulk of the problem.
This is the practical reason an instrumentation-first approach underperforms here. It measures the parts that were already visible and leaves the expensive part unmeasured, which produces a confident report about the wrong thing.
How to find it without instrumenting anybody
Twenty conversations beats any platform, and it takes a week.
Pick twenty people across three or four teams and ask one question: what did you wait for this week, and how long?
Not what is frustrating, which produces opinions. What did you wait for, which produces instances with durations attached. Write each one down with the team and the wait.
What the list tells you
Sort it and three things fall out immediately.
The same wait named by three different teams. That is a system problem and it is the one to fix first, because the fix divides across all of them.
A long wait named by one person. Usually an edge case. Worth noting, rarely worth a project.
A short wait named by everybody. The sleeper. Four minutes, forty times a week, across two hundred people is a large number that nobody has ever felt as large.
Measuring the process rather than the person is also the position that keeps this straightforward. the UK Information Commissioner guidance on monitoring workers sets out expectations around monitoring at work, and a conversation about what somebody waited for sits comfortably outside that, where keystroke-level telemetry does not.
Repeat the exercise every six months with different people. The list changes, and the changes are more informative than the list.
One question to add on the second pass
What did you give up on. Not what you waited for, what you stopped trying to do because the route was too slow to be worth it.
That answer never appears in any system, because an abandoned request leaves no record anywhere. It is also where the most expensive friction hides: the report nobody runs, the tool nobody asks for, the process improvement nobody proposes because the last three went nowhere.
Some friction is a control doing its job
Disclosure: RemoAsset, named below, is owned by the same people who publish HROpsLab. It is described with its limitations stated, on the same terms as every other supplier. This site takes no vendor money.
The part that gets skipped in the enthusiasm. Not every delay is waste, and a programme that treats all friction as the enemy will remove something that was holding a line.
Approval steps exist because somebody decided a human should look. Access requests take time because granting access instantly to anyone who asks is the failure mode the request was built to prevent. Frameworks such as ISO/IEC 27001, the information security management standard expect access to be granted on a basis somebody can defend, and the deliberate pause is frequently the control itself.
So sort the list into two columns before fixing anything: friction that protects something, and friction that is just nobody having looked at it.
The useful middle path
Most protective friction can be made faster without being removed. A same-day commitment, a named deputy for every approver, a standing pre-approval for anything under a threshold. The control stays and the wait shrinks.
That distinction is also the one that gets a programme supported rather than resisted, because the security and finance people stop reading it as an attempt to route around them.
The physical exception
One category cannot be sped up by better process, and it is worth separating out early: anything waiting on hardware. A laptop that has not arrived is not an approval problem, and no amount of workflow redesign shortens a shipment.
The lever there is ordering earlier and holding stock, which is a procurement decision rather than a friction one. Configured device procurement and delivery handles the mechanics, and the honest limitation is that lead times still vary by country, so the trigger has to move rather than the supplier.
The decision table
| Situation | Scale | Setup | Primary pain | Recommended starting point |
|---|---|---|---|---|
| Vague sense things are slow | Any | No data | Cannot name the problem | Twenty conversations, one question |
| Same wait across three teams | Any | Shared approver | One bottleneck | Named deputy and a time commitment |
| Short wait, very frequent | 200 plus | Daily task | Invisible in aggregate | Count it before dismissing it |
| Being sold telemetry | Any | No baseline | Measures the visible half | Ask first, instrument later if at all |
| Waiting on hardware | Any | Remote hiring | Physical lead time | Move the order trigger earlier |
The first week concentrates most of it
If the list is long and the appetite is short, work on the first week and ignore everything else for a quarter.
A new joiner hits every piece of friction the company has, in five days, with no accumulated workarounds and no idea who to ask. It is the densest sample available and the one where the cost is highest, because a slow first week sets an expectation that is expensive to reset.
Three measurements, taken from the last twenty joiners rather than from memory.
- Days from start date to being able to do the actual job. Not to having accounts. To doing the work.
- Number of separate people they had to ask to get there.
- How many of the five days were lost entirely.
What usually comes back
That the blockers are a small set repeating, and that most of them were triggered late rather than handled badly. The access request raised on the start date, the hardware ordered when the contract was signed rather than when the offer was accepted, the system owner who was never told a new person was coming.
None of those costs money to fix. Each is a timing change rather than a tooling one, which is why the first week repays attention out of proportion to the effort.
Each is a timing fix rather than a tooling one, which is cheap. Designing around how people actually experience a sequence rather than how the org chart imagines it is the substance of ISO 9241-210, the standard on human-centred design for interactive systems, and the first week is where that difference is most visible.
One thing worth checking in the same pass: whether the tools a new joiner must use are usable by everyone. the W3C Web Accessibility Initiative introduction to accessibility covers the ground, and the first week is when somebody will hit a barrier and quietly decide not to mention it. For the wider ownership question, see what HR actually owns in digital employee experience.
See how the hardware half of the first week is handled at configured device procurement and delivery, or read what HR actually owns in digital employee experience.
Frequently asked questions
What is digital friction?
The time that disappears between tools rather than inside them. Waiting for access, hunting for the right document, re-entering something a system already holds. It is largely invisible on dashboards because every tool reports on time spent inside itself, and the gap between closing one and opening the next belongs to nothing.
How do we find it?
Ask rather than instrument. Pick twenty people across three or four teams and ask what they waited for this week and for how long. Not what is frustrating, which produces opinions, but what they waited for, which produces instances with durations. A week of conversations beats a platform here.
What does the resulting list show?
Three patterns. The same wait named by three teams, which is a system problem and the first to fix because the fix divides across all of them. A long wait named by one person, usually an edge case. And a short wait named by everyone, which is the sleeper: four minutes, forty times a week, across two hundred people.
Is all friction worth removing?
No, and this is the part that gets skipped. Approval steps exist because somebody decided a human should look, and access requests take time precisely because instant access to anyone who asks is the failure they prevent. Sort the list into friction that protects something and friction nobody has looked at.
How do you speed up a control without removing it?
A same-day commitment, a named deputy for every approver, and a standing pre-approval below a threshold. The control stays and the wait shrinks. This also gets a programme supported rather than resisted, because security and finance stop reading it as an attempt to route around them.
Is this the same as monitoring staff?
It should not be. Asking somebody what they waited for is a conversation about a process. Keystroke-level telemetry is something else, with its own expectations around transparency and necessity set out in published guidance on monitoring at work. Keeping the measurement on queues and systems rather than individuals keeps the question simple.
Where should we start if we only do one thing?
The first week. A new joiner hits every piece of friction you have, in five days, with no workarounds and nobody to ask. Measure days from start to doing the actual job, how many separate people they had to ask, and how many days were lost entirely. The blockers are usually a small set, triggered late rather than handled badly.