Your dashboard shows green availability, decent response times and few critical incidents. Yet down the corridor, accounting still keeps its reconciliations in a spreadsheet, buyers enter their orders in a shared file first, and the word “ERP” gets a collective sigh. Who is right, the dashboard or the corridor?
Both, but they are not measuring the same thing. Technical indicators tell you whether the system is running. Only a measure of ERP user satisfaction tells you whether the system is used, understood and accepted. This article lays out a complete method: build the survey, set its cadence, cross it with usage data, interpret the results without relying on invented benchmarks, then turn it all into a continuous improvement loop tied to the business case.
Why Technical KPIs Don’t Tell You Whether the ERP Is Adopted
Availability, latency and incident counts measure the health of the infrastructure. They answer “is the service being delivered?”, not “is the service useful to the people who rely on it?”. An ERP can be 100% available and still be bypassed by half the teams.
Three blind spots come up again and again:
- Forced usage is not buy-in. A user who logs in because they have no choice generates a login record, not value. They may enter the bare minimum, at the last minute, with approximate data.
- Workarounds are invisible to monitoring. The Excel export, the parallel tracking file, the email to the colleague who “knows how to do it” leave no trace in your supervision tools.
- Perceived effort shows up nowhere. Ten clicks for an operation that took three in the old tool trigger no alert, but they erode trust month after month.
This is the logic of silent rejection described in our article on the 8 post-go-live adoption KPIs: the appearance of compliance masks a slow disengagement. The satisfaction survey is the instrument that makes that disengagement measurable, provided it is well built. For the wider picture of IT leadership metrics, see also the 15 KPIs the CIO must track after go-live.
Building the ERP Satisfaction Survey
Core Questions: Internal NPS, Perceived Effort, Trust in the Data
A good survey takes a few minutes. Three dimensions cover most of the diagnosis.
1. Internal NPS. The Net Promoter Score rests on a single question, scored from 0 to 10: would you recommend this product or service? According to Bain & Company, which developed the method with Fred Reichheld, respondents are classified as promoters (9 to 10), passives (7 to 8) and detractors (0 to 6), and the score is the percentage of promoters minus the percentage of detractors. Adapted for internal use, the question becomes: “On a scale of 0 to 10, how likely are you to recommend the ERP to a colleague who has to do the same tasks as you?”
For an ERP, internal NPS does not carry the same weight as for a consumer product: users do not choose the tool. It remains useful as a comparative thermometer over time and across populations, as long as you never present it on its own.
2. Perceived effort. It measures friction: “It is easy for me to complete my main tasks in the ERP” (agreement from 1 to 5). It is often the best predictor of workarounds.
3. Trust in the data. “I rely on the figures the ERP gives me without cross-checking them elsewhere.” A negative answer explains most parallel spreadsheets: nobody leaves Excel until they believe the ERP.
Sidebar: a 10-Question Survey Template
| # | Question | Format |
|---|---|---|
| 1 | How likely are you to recommend the ERP to a colleague with the same tasks as you? | 0 to 10 (NPS) |
| 2 | Why did you give this score? | Free text |
| 3 | It is easy for me to complete my main tasks in the ERP. | 1 to 5 |
| 4 | I rely on ERP data without cross-checking it elsewhere. | 1 to 5 |
| 5 | How many of your recurring tasks do you still perform outside the ERP (spreadsheet, email, paper)? | None / a few / several / most |
| 6 | What is the most painful or slowest task in the ERP? | Free text |
| 7 | Did the training you received make you autonomous? | 1 to 5 |
| 8 | When you get stuck, do you get a useful answer within an acceptable time? | 1 to 5 |
| 9 | Since the last survey, have you noticed an improvement? | Yes / no / don’t know |
| 10 | Which single improvement would save you the most time? | Free text |
Questions 2, 6 and 10 are the most valuable part: they produce the verbatims that feed the backlog. Don’t drop them to “lighten” the survey.
Segmentation: Finance, Procurement, Sales Admin, Logistics, Management
A global average hides the gaps. Always segment by function:
- Finance: demanding on reliability and traceability, sensitive to period-end closes and reconciliations.
- Procurement: sensitive to how smoothly purchase requests and approvals flow.
- Sales administration: heavy users, very sensitive to data-entry speed.
- Logistics and warehouse: usage is often mobile or on handheld terminals, with almost no tolerance for slowness.
- Management: consumers of dashboards, rarely data-entry users, with a perception centred on reporting quality.
On sampling, two principles. For small teams (a few dozen people), survey everyone: a sample makes no sense. For larger populations, make sure each segment has enough respondents that the reading doesn’t hinge on two isolated opinions, and track the response rate by segment. A segment that doesn’t respond is a signal in itself. Guarantee anonymity or, failing that, aggregation confidentiality: without it, answers will be polite and useless.
Cadence: 30, 90 and 180 Days After Go-Live, Then Every Six Months
Timing matters as much as the questions:
- Day 30: the survey captures initial frustration, training gaps and configuration bugs. Keep it short (questions 1 to 4, 6 and 8).
- Day 90: the learning phase is over. This is the time for the full questionnaire, because workarounds are settling in.
- Day 180: it measures the effect of the fixes you have made. It is the first real “before/after fixes” comparison.
- Then every six months: a rhythm that tracks a trend without wearing teams out. Avoid more than two surveys a year on the same panel.
Align the first wave with feedback from the testing phase: the anomalies raised during UAT (see our user acceptance testing checklist before go-live) are an excellent baseline for the day-30 verbatims.
Cross-Checking Perception Against Usage Data
A survey measures what people say. Data measures what they do. The gap between the two is the most instructive part.
| Source | What it shows | Useful cross-check |
|---|---|---|
| Login and usage logs | Frequency, modules actually used | A segment that says it is satisfied but doesn’t use a deployed module |
| Support tickets | Volume and nature of requests | Ticket themes should match the verbatims |
| Excel workarounds | Parallel spreadsheets identified in field interviews | Answers to question 5 vs. observed reality |
| Data-entry time | Duration of a key process (order, invoice) | Perceived effort vs. measured time |
Three typical readings:
- Satisfied but low usage: be wary of response bias, or the module is judged useless. Dig deeper through interviews.
- Dissatisfied but high usage: forced usage. Risk of turnover and degraded data quality.
- Satisfied and high usage: this segment’s practices should be documented and spread through champions.
For collection, favour the ERP’s native analytics and exports from your ticketing tool. For workarounds, a few 30-minute interviews in each department will uncover more parallel files than any log.
Interpreting Results: Cautious Thresholds and Verbatims
There are few reliable public benchmarks for internal ERP NPS, and the figures circulating on the subject are generally not comparable (different populations, scales and contexts). Rather than importing a theoretical “good score”, build your own internal reference points.
Internal reference points, not norms. Internal NPS is above all a relative score:
- Take the day-90 wave as the organisation’s baseline.
- Compare each segment to that baseline and to itself from one wave to the next.
- Define alert thresholds in advance that suit your context, for example “a negative NPS in a segment triggers an interview within 15 days” or “a drop between two waves triggers a backlog review”. These are governance rules, not statistical truths.
- Look at the distribution (promoters, passives, detractors) rather than the net score alone: a zero NPS can hide two sharply opposed camps or a mass of indifference, two situations that call for opposite responses.
Verbatims are worth more than scores. Code the free-text comments by theme (slowness, ergonomics, data, training, support, processes) and count them. Three habits help:
- Quote verbatims as they are in your readout, without rephrasing: they persuade better than percentages.
- Separate what stems from the tool, the configuration, the process and the training. The same “it’s complicated” points to different remedies.
- Spot recurring low-cost requests (a pointless mandatory field, a screen to reorder): these are the quick wins that rebuild credibility.
The Improvement Loop: Backlog, Prioritisation, “You Said, We Did”
A survey with no follow-up is worse than no survey: it teaches teams that their opinion doesn’t count, and the response rate collapses in the next wave.
1. Feed a single backlog. Every point from the verbatims, tickets and perception/usage gaps becomes an ERP backlog entry, with an owner, an affected segment and a source.
2. Prioritise with a simple grid. Cross four criteria: number of users affected, severity of the friction, effort to fix, and link to a critical process (period-end close, invoicing, shipping). “Low effort, many users” fixes go first.
3. Choose the right lever. Depending on the cause, the answer is a configuration fix, an enhancement from the integrator, a targeted training session, a quick-reference sheet or a process change. Don’t treat everything with training: training someone on a poorly designed screen does not make it well designed.
4. Communicate: “you said, we did”. At each cycle, publish a short message: “You said: approving purchase requests takes too long. We did: removed two steps for standard amounts.” This simple sentence is the engine of participation in the next survey.
5. Close the loop with contributors. In line with the Net Promoter logic described by Bain (share the feedback with the relevant teams, then go back to the person who gave it), notify users directly when their point has been addressed, through champions or support.
Linking Satisfaction to the Business Case
Satisfaction is not a comfort indicator: it conditions return on investment. The benefits of an ERP business case (fewer redundant entries, faster closes, reliable data, quicker decisions) only materialise if processes are actually run in the tool.
So link the two levels explicitly:
- Business case assumption: “order entry will be 30% faster” (an example assumption, to be replaced with your own). Usage measure: observed entry time. Perception measure: perceived effort among sales administration.
- Assumption: “no more reconciliations outside the ERP”. Usage measure: number of parallel spreadsheets identified. Perception: trust in the data on the finance side.
Then present the steering committee with a single table: expected benefit, usage indicator, perception indicator, gap, decided action. A segment whose satisfaction is slipping is a risk to a quantified benefit in the business case, and should be treated as such in post-go-live reviews. If adoption weakens, post-go-live ROI deteriorates before the financial figures show it: the survey is a leading indicator.
5 Common Mistakes and How to Counter Them
- Measuring only once, right after go-live. At day 15, everyone is still learning. Countermeasure: day 30, 90, 180 cadence, then every six months.
- Publishing a global score with no segmentation. The average hides struggling sales admin behind a satisfied finance team. Countermeasure: report by function, with response rates.
- Not cross-checking with usage data. Perception without facts invites debates of opinion. Countermeasure: a perception/usage/tickets table for each segment.
- Importing an external benchmark or setting an arbitrary target. It pushes people to optimise the score instead of solving the problems. Countermeasure: internal reference points compared from one wave to the next.
- Changing nothing, or not saying so. The response rate drops and distrust sets in. Countermeasure: a visible backlog, announced timelines, and “you said, we did” communication at every cycle.
Going Further
To move from method to implementation, download our post-go-live adoption and KPI checklist: it brings together the usage indicators, the survey template and the backlog tracking grid to use from the first wave at day 30.