WF-19 — Research Impact Capture

Ten impact units scores 100, and that is a low bar for an institute of learning inside a health system with policy reach. KPI 4.5 is not a ratio. It is a count. Each project can earn at most one unit in each of four dimensions, so ten units means as few as three well-evidenced projects, or ten single- dimension claims spread across the portfolio, accumulated over five academic years. The units are not lost because the impact does not happen. They are lost because nobody records them at the moment they occur, and by the time an annual data call goes out, the invitation email has been archived, the award is remembered but not evidenced, and the policy document that cited the work is no longer findable. The whole design problem of this workflow is that the capture must be continuous and the reporting is annual.

1. Purpose

To ensure that every instance in which IoL research produces an economic or social effect in the world (a subsequent grant, revenue or a licence, public recognition or reach, a citation in policy or a briefing to officials) is recorded against the project that produced it at the time it occurs, with dated evidence attached, so that the institution can state what its research has changed and can evidence that statement without notice.

2. Scope

Scope statement. This process manages research impact from the opening of an impact record at project registration, through continuous capture of impact events against the four OBEF dimensions, to the annual roll-up, de-duplication across the five-year window and submission of the institutional count.

Applies to. All IoL research and scholarly projects registered under WF-15, including educational research, health professions education research, curriculum and assessment development studies, and student dissertation and capstone projects that produce an external effect. Impact arising in the past five academic years. Institution-level reporting only.

Does not apply to. The publication of the research itself, which is WF-18 and a different KPI. The granting of an IP asset, which is WF-20 and KPI 4.6; note that an IP grant may nonetheless evidence a commercialisation unit here, and the same underlying event legitimately appears in both workflows serving two different KPIs. Industry contribution revenue as a financial total, which is WF-24 and KPI 3.4. Programme-level reporting of any kind: KPI 4.5 has no programme-level equivalent, and no PGDipHPE or MScHPE row will ever carry it.

Applicable requirements. OBEF guide v11.5 KPI 4.5; CHEDS Research Impact.xlsx template; CHEDS Master API specification 18 February 2026, fields OBF4.5.N, OBF4.5.D, OBF4.5.R, OBF4.5M; MBRU research policy [IoL to confirm]; data protection where impact records name individuals.

The four dimensions that govern everything below. One unit each, maximum four units per project.

The scoring mechanic, stated explicitly because it decides how the register must be built.

1. Meeting more than one criterion within the same dimension earns no extra unit. A project that won a prize, generated wide media coverage and whose authors spoke at eight public events earns one public reach unit, not three.
2. Two projects each achieving the same dimension earn two units. Breadth across projects is what accumulates, not depth within one.
3. Units are counted once, not in every year of the five-year window. A unit earned in 2023 is not re-earned in 2024 because the impact continued.
4. The time frame is a single value computed over the past five academic years, not an annual figure and not a rolling average of annual figures.

The consequence for register design is direct: the register is keyed on project by dimension, each cell is a boolean that can be set once, and the evidence hangs off the cell. Anything organised by year, or by event, will double count or will lose the once-only rule.

3. Trigger, boundary and endpoint

4. SIPOC

5. Accountability

Process owner. IoL Research Lead, or the role able to require project teams to report impact events and to commission the quarterly review. The owner must be senior enough that a fifteen-minute standing item on a research meeting agenda survives a busy quarter.

Escalation.

6. Process steps

1. Open the impact record when the project is registered, not when the impact happens. Every project admitted by WF-15 receives an impact record keyed on its Project ID, with four dimension slots, all empty. An empty record is the thing that makes the later capture possible, because it gives the project team somewhere to put an event on the day it occurs. [CAPTURE] Project ID, Project Title, Department, College, Type of Scholarly Activity, project start date and status. Serves 4.5 and the C6 once-only assertion.

2. Publish the standing capture triggers, and attach them to processes that already run. Impact events arrive through a small number of predictable channels, and each of those channels is already somebody's job. The trigger list makes reporting a by-product rather than a request. [CAPTURE] the trigger list as a controlled document, with a named owner per trigger. Serves 4.5.

   Channel	Who already sees it first	Dimension it may serve	What they must do
   Grant or funding award notification	Finance and grants administration	Funding attraction	Flag any award of AED 50,000 or more that follows an existing registered project, with the award letter
   Revenue receipt against a project output	Finance	Commercialisation	Flag cumulative receipts reaching AED 50,000, with the revenue record
   Licence, sub-agreement or trademark concluded	Contracts and technology transfer	Commercialisation	Flag the executed agreement or the trademark certificate
   Prize, award or competition result	Principal investigator, communications	Public reach	Flag the citation or award notice
   Invitation to present at a public event	The invited author	Public reach	Log the invitation against the project, and increment the counter
   Educational material published or publicly referenced	Principal investigator, programme office	Public reach	Flag the publication or the reference, with the link and date
   Media or social media coverage	Communications office	Public reach	Flag the coverage record
   Research named in a policy document	Principal investigator, research office	Public-policy citation	Flag the document, page and date
   Request to brief government or international officials	The invited author	Public-policy citation	Flag the request and the briefing record

3. Capture the event at occurrence, within 20 working days, with the evidence attached. The report is short: Project ID, what happened, the date, the dimension it may serve, and a link or file reference to the artefact. It is not an assessment; the assessment is step 4. [CAPTURE] event date, event description, proposed dimension, evidence reference, reporter. Serves 4.5. The 20 working days is the whole control. An event reported at the time carries its own evidence, because the email, the letter or the certificate is still in front of the person reporting it. The same event reconstructed eleven months later is a recollection, and a recollection cannot be submitted.

4. Test the event against the dimension criteria, and record the test. The Research Lead confirms which dimension, if any, the event satisfies, against the criteria in section 2. Two tests are counted rather than asserted: the AED 50,000 thresholds, which are tested against documented amounts and never against estimates, and the at least five public events test for public reach, which requires five evidenced invitations, not an impression that the author speaks often. [CAPTURE] the dimension determination, the criterion relied on, the tester and the date. Serves 4.5. Exception route. An event that satisfies no dimension is recorded in the register as a non-qualifying impact event and left unset. It is still worth holding: it is material for WF-04, for annual reporting and for the impact narrative, and it prevents the same event being re-proposed every year.

5. Set the dimension once. If the dimension is already set for that project, the register records the new event as additional evidence and does not create a second unit. If it is not set, the boolean is set with its date and evidence reference. [CAPTURE] the dimension boolean, first-qualifying-event date, evidence reference. Serves 4.5 and validation rules 1 and 2.

6. Maintain the running counters. Two dimensions have thresholds that accumulate and therefore need a counter rather than an event flag: the invitation counter for public reach (which must reach five evidenced public events for one project's authors) and the cumulative revenue counter for commercialisation (which must reach AED 50,000). Both are kept per project. [CAPTURE] the counter value and the evidence list behind it. Serves 4.5. Nobody keeps an invitation counter unless asked to. This is the most commonly available and most commonly unclaimed of the four dimensions for an academic unit whose staff are invited to speak.

7. Monitor for policy citations and briefing requests. Set standing alerts on the names of the principal investigators and the project titles, and ask project leads once a quarter the direct question: has your work been cited in a policy document, or have you been asked to brief officials. [CAPTURE] the policy document reference with page and date, or the briefing request and record. Serves 4.5. For an institute of learning inside a health system, this dimension is more available than it looks. Health professions education research feeds workforce and training policy, and briefings to health authority officials are ordinary professional activity that nobody thinks to record as research impact.

8. Review quarterly, for fifteen minutes. A standing item on the IoL research meeting: which projects have new impact events, which counters have moved, which dimensions are one evidenced event short. This is the mechanism that converts continuous capture from an instruction into a habit. [CAPTURE] review date, register changes since the last review. Serves 4.5. Decision point. If a project is one evidenced invitation short of the five-event test, the review decides whether to seek and evidence the fifth. That is legitimate: the invitations either happen or they do not, and asking whether they have happened is not the same as inventing them.

9. Roll up annually across the five-year window and de-duplicate. Sum the set dimensions across all projects with any impact in the window. Confirm no project contributes more than four units, no unit is counted in more than one year, and no event is attributed to two projects. [CAPTURE] the unit count, the window used, the de-duplication result. Serves 4.5 and assertion C6.

10. Reconcile with KPI 4.6 and KPI 3.4 before release. The same underlying event can properly serve more than one KPI while being counted once in each: a granted trademark is a KPI 4.6 asset and evidence for a commercialisation unit here; licensing revenue is a KPI 3.4 industry contribution and evidence for the same unit. What must not happen is the same event being counted twice within KPI 4.5. [CAPTURE] the cross-KPI reconciliation record. Serves 4.5, 4.6 and 3.4 consistency.

11. Release the count and the mechanism statement to WF-26. Populate OBF4.5.N, OBF4.5.D, OBF4.5.R and state in OBF4.5M the window used, the register from which the count is drawn, and the extract date. [CAPTURE] submitted values and the mechanism text. Serves 4.5. Note a genuine ambiguity in the submission contract. KPI 4.5 is a count, not a ratio, yet the submission carries a numerator, a denominator and a result. What belongs in OBF4.5.D for a non-ratio KPI is not stated anywhere in the guide. Do not guess. Ask CHEDS, record the answer, and apply it consistently; a plausible convention is denominator 1 with the count in both numerator and result, but a plausible convention adopted silently is a finding waiting to happen.

7. OBEF data generated

Capture rule. Four things cannot be reconstructed after the fact, and all four are ordinary correspondence that people delete. The speaker invitation (an email, held nowhere institutionally, and the only evidence that the five-event test was met). The policy document reference with its page and date (policy documents are revised and republished, and the version that named the research may not be the version online next year). The media coverage record (links rot, and a screenshot taken at the time is evidence while a memory of coverage is not). The dated award or prize citation. Amounts and agreements survive in finance and contracts systems; these four survive only if someone files them on the day.

Reproducibility test. A second analyst cannot today reproduce KPI 4.5 for MBRU from records alone, and that is the finding this workflow exists to fix. The four columns in Research Impact.xlsx map one to one onto the four OBEF dimensions, which is not a coincidence: the CHEDS template was built for this KPI. If those four columns are unpopulated, the KPI is unreportable, and that is the most likely explanation for MBRU's dashboard showing 5 of 6 KPIs reported in Pillar 4. Once the register exists and the columns are populated with a dated artefact behind each set cell, the test passes trivially, because the KPI is a sum of booleans.

8. Max-score design

Read the bottom of the anchor scale carefully, because it is where the argument sits. The distance from zero units to one unit is worth 25 points of a 2.0% KPI. The distance from eight to ten is worth another 25. A single evidenced speaker-invitation count, or a single policy citation, moves a KPI that is currently most likely unreported. That asymmetry is the reason this workflow is worth building even though its weight is small: the first unit is nearly free and the KPI may currently be null.

What this workflow must do to reach the target.

1. Populate the four dimension columns in Research Impact.xlsx. Everything else in this section is secondary to the fact that an empty template produces a null KPI, and a null KPI scores nothing.
2. Open an impact record for every project at registration, so that there is somewhere to put an event on the day it happens.
3. Attach the nine capture triggers to the people who already see the event first. Finance sees the award. Contracts sees the licence. Communications sees the coverage. The author sees the invitation. None of them currently has a reason to tell anybody.
4. Keep the invitation counter. The five-event public reach test is the most commonly satisfied and least commonly evidenced criterion available to an academic unit.
5. Ask the policy question directly, once a quarter. People do not think of a briefing to a health authority as research impact, so they will not volunteer it, and they will answer accurately if asked.
6. Hold the evidence, not the assertion. Every set dimension carries a dated artefact in the Evidence Register at the moment it is set, not at submission.
7. De-duplicate across the window before every submission, so no unit is counted in two years and no project exceeds four units.

[REDESIGN] actions.

Sequencing note. R1 before R2, because triggers need a destination. R6 last, and understood correctly: a backfill recovers only what left a paper trail, and its main value is to show how much did not. If the backfill finds two units where the team believed there were eight, that finding is the business case for R2, and it should be reported rather than tidied away.

9. Indirect strategy where data cannot be collected

Category 4 applies, and almost nothing else does. The activity exists and it is invisible.

Category 4, the activity exists but is invisible. IoL sits inside Dubai Health. Its staff research health professions education, advise on training and workforce questions, speak at professional meetings, publish educational materials that other institutions use, and are asked for their view by people who write policy. Every one of those is a candidate for a KPI 4.5 unit, and none of them is currently recorded as research impact, because research impact is not what any of them feels like at the moment it happens. It feels like a talk, an email, a committee, a course pack.

The response is capture, not tactics. The charter says so in general terms. This section says what it means operationally, because "capture at the moment it occurs" is easy to write and is the part that usually fails.

What "capture at the moment it occurs" actually requires. Six things, all of them concrete.

1. A destination that already exists before the event does. The impact record is opened at project registration with four empty cells. A person who has just been invited to speak must have somewhere to put that fact in under two minutes. If the destination has to be created first, the event is not recorded.
2. The report is short and is not an assessment. Five fields: project, what happened, date, dimension it might serve, evidence attached. Whether it qualifies is decided later by the Research Lead. Asking a busy clinician-academic to judge a dimension criterion is asking them not to report.
3. The reporter is the person who sees it first, not the person who owns the KPI. Finance sees the award letter. Contracts sees the executed licence. Communications sees the coverage. The author sees the invitation. Nine channels, nine named owners, each already handling the artefact for another reason. The marginal cost of one flag is close to zero; the cost of a research office trying to discover the same nine things a year later is high and mostly fails.
4. The evidence travels with the report. Not a promise to find it later. The invitation email forwarded, the certificate scanned, the coverage screenshotted with its date, the policy document page saved as a PDF. This is the single operational difference between a register that can be submitted and one that cannot.
5. A standing counter for the criteria that accumulate. Five public events and AED 50,000 of revenue are both thresholds crossed by accumulation, and neither is visible at the moment of any single event. A counter kept per project is the only way anyone knows the threshold has been crossed.
6. A short, regular review that is not the annual data call. Fifteen minutes a quarter, on an agenda that already exists. Its purpose is not to collect; collection has already happened. Its purpose is to notice near misses while the fifth invitation can still be sought and evidenced.

Category 2 does not apply. KPI 4.5 is a count, not a survey, so there is no population and no small-sample problem. Appendix C grouping is irrelevant here.

Category 3, activity does not exist, applies to at most one dimension and must be classified honestly. IoL may genuinely have no commercialisation activity: no revenue from project outputs, no licences, no trademarks. If that is true, say so, and put the effort into the two social dimensions, which cost nothing but attention. There is no weight redistribution available on KPI 4.5. The guide provides redistribution in exactly four programme-level cases and this is not one of them, and KPI 4.5 has no programme level at all. An absent dimension is simply an unearned unit, and classifying it as absent gains nothing. That removes any incentive to misclassify, which is worth stating.

Category 5, performance genuinely low, is a real possibility and should be tested before it is accepted. If the register runs for a year with the triggers live and produces one or two units, the honest reading is that IoL's research is not yet producing external effect at scale, which for a young institute is unsurprising and is a legitimate finding for WF-04 and a corrective action plan. What must not happen is concluding that from an empty register, because an empty register measures the absence of capture, not the absence of impact. Give R1 and R2 one full year before drawing any conclusion about performance.

A note on the relationship to KPI 4.6, since the two look similar and are not. KPI 4.6 counts granted IP assets, and it is a property test: was something granted, is MBRU's share at least 20%, in which year. KPI 4.5 counts effects in the world, and it is an evidence test. A trademark registration is a 4.6 asset and separately satisfies a 4.5 commercialisation criterion, and counting it in both places is correct because they are different KPIs. Counting it twice inside 4.5, once as a trademark and once as a licence, is not.

Boundary check. This workflow must never:

• set a dimension without a dated, retrievable evidence artefact held at the time the dimension was set;
• count more than one unit per dimension per project, however many criteria within that dimension are met;
• count more than four units from a single project;
• count the same unit in more than one year of the five-year window;
• attribute the same event to two projects in order to earn two units;
• treat an estimated or approximate amount as meeting the AED 50,000 thresholds, which are tested against documented amounts;
• assert the "at least five public events" test without five separately evidenced invitations, since the criterion is counted, not judged;
• count a speaking invitation that was declined, or an event that did not take place;
• count internal MBRU or Dubai Health seminars as public events;
• count a policy document that cites the field generally rather than naming the research;
• count a briefing that was offered but not requested;
• report a unit at programme level, since KPI 4.5 has no programme-level application;
• reconstruct an impact claim at submission time from recollection, which is the specific failure this workflow exists to prevent.

10. Service standards

11. Records and evidence

Appendix B readiness. The Ministry may demand proof of academic, economic or social research impact including policy impact, outreach, funding attraction, publication, IP and commercialisation, giving registration of research spin-off companies and speaker invitations as examples. Today, IoL could most likely produce the funding award letters and any licensing agreements within 15 working days, because finance and contracts hold them for their own reasons. It could not produce a speaker-invitation list, a media coverage record with dates, or a policy citation reference [IoL to confirm], because nothing currently asks anyone to keep them. Those three are precisely the evidence for the two dimensions most available to IoL, which is why R2, R3 and R5 matter more than their modelled point value suggests.

12. Risks and controls

13. Performance measures

Watch the second outcome measure rather than the first. Because each project yields at most four units and most yield one, the count is driven by how many projects have any impact recorded, not by how much impact the best project has. A portfolio with eight projects at one unit each outscores a portfolio with two projects at four units each, and it is also the more honest description of a functioning research unit.

14. Change control