One Sponsor Is a Single Point of Failure: Sustaining Executive Support for an MDM Program (original)

A sponsor and a funder are different objects. Telling them apart takes two meetings; making sure the program outlives the one you have takes the rest of the first year.

The money lands and the room exhales. I have learned to pay attention right at that moment, because it is the moment the program stops being an argument and becomes a structure — and nobody has yet asked what is holding it up.

Last Monday I spent the whole piece on how that argument gets built and funded. This one is about the part after. You have an approval, a line in somebody's budget, and an executive who said yes. The questions now are different in kind: how long does that yes last, what is it carrying, and what breaks it?

I am going to read sponsorship the way you would read a system — components, behavior under load, edge cases. The vocabulary is borrowed and I want to say so; sponsorship is people, and people are not a load path. But it earns its keep, because it moves the conversation off whether an executive likes your program, which nobody can act on, and onto whether the program survives a bad quarter, which you can check on a Tuesday.

Components: a sponsor and a funder are not the same object

Start with what the field agrees on, because it is unusually unanimous here. The Project Management Institute's in-depth study of executive sponsorship found that actively engaged sponsors were, for the second year running, the top driver of projects meeting their original goals and business intent [1]. Prosci reports the same over a longer run: active and visible sponsorship has been number one on its list of top contributors to change success in every benchmarking report since 1998 [2]. Different instruments, different decades, same answer.

Read the adjectives, though, because they are doing the work. Not assigned. Not named on the charter. Actively engaged. The same research found that across 2012, 2013 and 2014 fewer than two in three projects — 63 percent — had actively engaged sponsors, and that one in three unsuccessful projects failed to meet its goals because of poorly engaged ones [1].

That gap has a name in practice, and my co-author — who has been on both sides of that table more than I have — puts it flatly: Funders just write a check and make sure it clears.

Unpack that and you have the most useful distinction I know in this area. A funder authorizes the money and confirms it cleared; the transaction is the whole of the relationship. A sponsor buys into the process the whole way through to the thing you deliver — they have something at stake in whether it lands. Neither role is illegitimate, and on a well-run program the same person is often both. What is dangerous is mistaking one for the other. Both sign. Both appear on the steering deck. Only one shows up when a department head refuses to change a process.

Both practitioner frameworks describe the sponsor half in operating terms, and every item is something a funder never does. Prosci's three parts: visible participation throughout, building a coalition of sponsorship, and communicating support directly to the people the change lands on — with the warning that sponsors cannot disappear after the kickoff [2]. PMI's five most important actions, page 9: removing roadblocks, aligning the work to strategy, championing the program, adding resources, resolving issues quickly [1]. Neither body is disinterested — one sells the method, the other credentials the profession — so it matters that an auditor with no stake in either found the same verbs. Of seven federal technology programs that hit their cost, schedule and scope targets, six named senior executive support as critical: procuring funding, supplying information at critical moments, intervening when another department would not cooperate, defining the vision [8].

The test is engagement, and it runs in the first two meetings

The good news is that engagement is legible early and cheaply: attendance, and the shape of the questions. Someone who is there to authorize a purchase order asks about the number and the date. Someone who has something at stake asks what happens to the data when two systems disagree, and who is going to be unhappy about the answer.

It also has a price you can hold a calendar against. PMI's sponsors reported carrying three projects at once and giving about thirteen hours a week to each, on top of their day jobs (page 8) [1]; the UK government's project delivery standard turns the same quantity into a rule, requiring the senior responsible owner of a major project to give it at least half their time until the full business case is approved [10]. Engagement is not a mood. It is hours, and somebody either pays them or does not.

There is a second question underneath about that same person, and my co-author puts it more sharply than I would: The important point is determining whether that person is the Decider or merely an Advisor. Someone will always fill the funding role, or the program would not have reached a meeting at all. What you are establishing is whether that person can also decide — and engagement does not answer it. An advisor with real conviction attends everything and asks the sharpest question in the room; a decider with signing authority can be bored by the whole business and sign anyway.

For that, the indicators are structural, and page 9 of PMI's report shows why. It measured how often project managers said sponsors demonstrated a given skill against how often sponsors said they did, and found gaps of 45 to 48 points on motivating, active listening, communication and managing change. Exactly one item drew agreement — whether the sponsor had authority over the affected business unit — and the report explains it: that one is confirmable on the organizational chart rather than a matter of subjective judgment [1].

There is a counter-argument fourteen pages later in the same document. Two Boston Consulting Group authors, writing on page 15, argue that being the "right" sponsor often has little to do with the actual authority a sponsor holds over the program team, and that the ability to build and leverage networks can trump one's place in the organization chart — particularly for initiatives that cut across the business [1]. Master data management (MDM) is about as cross-cutting as work gets. It is also measurably how our own function operates: of 27 US state chief data officers surveyed in 2025, 15 percent reported no established authority at all over data management policy, and the report's summary of the role is that effective ones lead through influence rather than authority [13].

I do not think the findings compete, but I am not leaving my own instruction standing as it was. Authority tells you what can be compelled; influence tells you what can be persuaded. They fail differently — compelled cooperation degrades into the letter of the instruction, persuaded cooperation evaporates when the persuader gets busy — and what gets programs into trouble is not choosing wrong, it is not knowing which one you are relying on this quarter.

So make it two checks. The org chart gives you the boundary of what can be compelled. The second costs one conversation: list the functions whose processes have to change for your program to land, strike the ones inside your sponsor's authority, and for each of the rest ask them who they would call. The names are the reach of the influence you are relying on; the silences are scope nobody has agreed to. A sponsor who can name somebody for two functions out of seven is not weak — their reach stops short of your program, and the remedy is the rest of this piece.

What sponsorship carries, and the load it will not take

What sponsorship actually buys, stated narrowly enough to be useful: budgeting priority. A sponsor worth the name has money they personally allocated and will defend when the portfolio gets squeezed. It is the part of the role that survives a bad quarter.

What it does not buy — and this is where a lot of otherwise careful plans have a hole in them: enterprise adoption. An excited executive is not a changed process. The departments whose ways of working have to change are a separate problem with a separate solution. The auditors sort it the same way: on the GAO's list of nine success factors, senior executive support is one item and the involvement of end users in requirements and in testing are two others, counted separately, because they are separately obtainable and separately missable [8].

The survey evidence wants reading carefully. The 2026 benchmark survey of artificial intelligence (AI) and data leadership — senior AI and data executives at nearly 110 large enterprises — asks which is the greater impediment to adoption, culture and change management or technology. Ninety-three percent pick culture [4]. A forced choice ranks rather than measures. What it says is that when the people who own these programs are made to choose, the human obstacle out-ranks the technical one — every year since 2021, never below about 78 percent [4].

The research literature supplies a mechanism for how enthusiasm at the top turns into obstruction below. Liette Lapointe and Suzanne Rivard, writing in MIS Quarterly, built a multilevel model in which resistance has an object that migrates: groups first resist the system; if the work shifts the balance of power between them and other user groups, the object moves to the system's significance; and if what is at stake is their power relative to the system's advocates, it moves again — to the advocates themselves [6]. I am reading that from the published abstract; I could not get the paper. Kotter's I could, and he describes the same thing from the executive floor: a division officer who "paid lip service to the process but did not change his behavior," colleagues who did nothing about it, and a renewal effort that collapsed underneath him [9].

MDM is unusually exposed to this, because our deliverable is a redistribution of authority. Riikka Vilminko-Heikkinen, Paul Brous and Samuli Pekkola studied an MDM development process directly and reported that managing master data as an organization-wide function enforces changes in responsibilities and established ways of working, that those changes create tensions which can become conflicts, and that thirteen MDM-specific paradoxes arose [5] — again the abstract, so take the count as reported rather than verified. The practitioner data agrees: organizational resistance was the third-ranked challenge those state chief data officers named, above authority and below staffing [13]. We do not deliver a system. We deliver an answer to who decides, and somebody was already answering it.

Depth is the specification, and it is not a committee

So: the single engaged sponsor is necessary, insufficient, and — this is the part that gets skipped — a single point of failure. The mitigation is not a better sponsor. It is more of them, which is not a novel idea: building a coalition of sponsors is the second of Prosci's three sponsor responsibilities and, by their measurement, the one sponsors struggle with most [2], and it is the second of John Kotter's eight steps for leading change [3]. The consensus has said this for thirty years.

Where the consensus is less clear is the arithmetic, and page 8 of PMI's report is where it gets quoted for more than it says. Nearly a third of projects at organizations using executive sponsors have multiple sponsors, and the report states plainly that multiple sponsors are not necessarily an indicator of greater success [1]. That is a failure to find an effect across all projects, not a finding against them, and it stops anyone — me included — claiming more names is automatically better. The same page then qualifies it in a way that lands directly on our discipline: a majority of organizations report that work driving significant change (79 percent), work of high complexity (71 percent) and work with relatively high budgets (60 percent) is significantly more likely to succeed with multiple sponsors [1]. An enterprise MDM program is all three at once.

Prosci's structure resolves the tension, and it is the structure I would build to. There is a primary sponsor — the leader who authorizes the change and is ultimately responsible for it realizing its intended benefits — and that person enlists the others [2]. Accountability sits in one seat, and a government that has to defend the arrangement in front of a parliamentary committee reaches the same answer in the same words: there can be only one accountable person, and the accountability cannot be delegated or shared [10]. Support sits in as many seats as you can get — and my co-author sets that target higher than most programs dare to ask for: the entire executive team, as the goal rather than the fallback.

Kotter gets read as the brake on this, and he is not one. The eight steps came out of a decade watching more than a hundred companies attempt transformation, and every number in that account is a floor: three to five people in the first year, growing to twenty or fifty in a large firm before much progress is possible at all, a coalition that "grows and grows over time" [9]. He does say it never included all of a company's most senior executives — and in the same breath the reason, which is the half that gets dropped: because some people just will not buy in, at least not at first. That describes where a coalition starts, not where it ends. The names you do not have are a map of buy-in you have not won yet; the names you have are worth what the person behind them will spend.

Which raises what the consensus is thinnest on, and it is not the count. "Build a coalition" is a noun with a verb parked in front of it. What is the ask? Not your name on my steering deck — ask for that and that is precisely what you will get. In the US House of Representatives a bill has exactly one sponsor and an unlimited number of cosponsors; cosponsors do not sign the bill; a name goes on by a form the sponsor's office files, and can come off again [11]. Support that costs nothing to give costs nothing to withdraw — and everybody reads the list as evidence anyway, which is why supporters collect it.

The House has priced the difference between a list and a constituency, and the price is steep. Under its Consensus Calendar rule an unreported bill earns an alternative route to the floor only once it has held at least 290 cosponsors — roughly two-thirds of the chamber — for twenty-five legislative days [12]. Duration, not signatures. And even that buys a hearing rather than an outcome: in the 117th Congress eight such motions were filed, four bills reached the Calendar, and none of the four became law. One came off it when the committee that actually held jurisdiction marked the bill up and reported it without recommendation, which under the rule removes it [12]. Two-thirds of the chamber on the cover, and the few whose remit genuinely covered it ended the thing in an afternoon.

So a co-sponsor is specified by what they will spend, and Prosci states the job: enlisted to legitimize the change inside their own part of the organization [2]. That has a scope, so the ask is made one function at a time and in that function's terms. You already have the instrument for reading the answer: the engagement test from the first two meetings, re-run once per co-sponsor. Do they turn up, and what do they ask about? Someone who asks when the migration lands has joined a list. Someone who asks which of their own processes changes, and who in their function will hate it, has read the thing. Kotter puts the same test as a currency: the coalition has to be powerful in titles, expertise, reputations and relationships, and efforts that hand it to a staff executive rather than a line manager "never achieve the power that is required" [9].

Then, after the yes, when both kinds look identical for a quarter, two things separate them: is there something of theirs on your roadmap, and did any of their money move. A coalition of people who fail both is not depth. It is a longer list of people to notify when the program is canceled. How the conversation runs, and what to do with a refusal, is below the line. It is not mine.

The failure mode is absence, not change

Now the edge cases. The first thing to get right is which event you are bracing for. Programs prepare for the sponsor changing; what ends them is the sponsor being gone. Confusing the two costs you a contingency plan pointed at the wrong risk.

Change. A new occupant of the seat is close to a non-event, under two conditions you can check rather than hope for. Both come from below the line, and both belong on the risk register in place of the vague entry usually there. Emphasis shifts. You re-run the orientation from month one. The requirements do not move.

Hold that next to a body of research that reads the other way. The escalation literature in information systems treats a change of the responsible executive as a classic trigger for a failing program to be reassessed or stopped: a new decision-maker carries no personal responsibility for the original commitment and is free to kill it. Magnus Mähring, Mark Keil, Lars Mathiassen and Jan Pries-Heje identify seven roles that shape whether de-escalation happens, two of them named the exit sponsor and the exit champion [7]. Abstract again; the deposited paper would not open for me either. Kotter's eighth error is the version I could read in full, and it is quieter: where the change champion was the retiring executive and the successor was not a resistor but was not a champion either, the signs of renewal had gone within two years [9]. Nobody canceled anything.

Absence. The two accounts stop competing once you notice they name different events. An executive who arrives looking for the off switch is not a new sponsor; they are the absence of one wearing the title, and absence does not require a vacancy. Either way the number of people who will defend your line item at the next portfolio review is zero, while the org chart still shows a name above your program. Sorted like that, the escalation research is a study of absence.

The mechanism is not mysterious. Everything sponsorship buys is personal. Page 10 of PMI's report lists where sponsors most help — rallying senior management, intervening on escalated issues, removing roadblocks, stakeholder management, championing the program — and every one of those stops the day the person stops [1]. Nothing on it is institutional. So the program does not decay gradually; it loses all of it at once.

Plan for the handover either way. The 2026 benchmark survey puts about half of chief data officer tenures under three years [4]; a survey of state government chief data officers, run the year before by a different body over a different population, put the median at 28 months and found 56 percent of respondents were not the first to hold the seat [13]. Two small samples agreeing is not a measured turnover rate, but it is enough to stop you treating the occupant as a fixture. The UK standard simply assumes the event: the appointment letter records tenure and any succession plan up front, a departing owner must hand over, and a month is the stated minimum [10]. Treat the change of seat as scheduled with an unknown date, and the transition plan as the deliverable that decides which case you are in when it arrives.

And a third case, which is neither. It happens nowhere near your sponsor's level: the strategy above the program changes. Your sponsor can be engaged, present and entirely sincere and still be holding a line item now weighed by people who have never met you, against priorities that did not exist when it was funded. My co-author's advice there is the least sentimental thing in this piece: Then you had best finish the project before the analysis is complete or you're going to need to fight for your funding. A strategic review is a clock. Against all three cases the coalition does one job: it makes absence degrade into change.

Where your next sponsor is coming from

One more thing about the present moment, because it changes the supply side. Master data work has spent most of its existence competing for executive attention it had to manufacture. Not today. The number I would act on from that 2026 benchmark survey is 93 percent of respondents saying interest in AI has led to a greater focus on data inside their own firm [4] — a change they watched happen where they work, rather than a rating of their own priorities, which is what the headline 99 percent figure really is. Take it, knowing what kind of sponsorship an enthusiasm produces.

A sponsor generated by a wave has a mandate defined by the wave. Their commitment is to the AI outcome, and master data is what stands between them and it — excellent, right up until the roadmap changes and it isn't. Worse, the seat itself is unsettled: about four in ten of those organizations have appointed a chief AI officer and there is no agreement about where the role belongs, the reporting line splitting four ways [4]. A seat the organization has not finished designing is a seat it may yet design away. Spend the window converting one enthusiastic backer into several durable ones — and note what those same executives name as the obstacle, with the money flowing and the tooling in production. Culture. Our lesson, arriving in somebody else's vocabulary with a much larger budget attached.

What I would be doing in month four

Sponsorship looks like a status — you have it or you don't — and treating it that way is why programs are surprised by their endings.

So, from month four: confirm on the org chart where the primary sponsor's authority ends, and for the functions outside that boundary ask who they would call — the silences are scope nobody has claimed. Find out which budget line the money came out of and whether it was theirs. Assume the budget will hold and the adoption will not. Write down what a clean handover of that seat would require, in the quarter when nobody needs it. And treat co-sponsorship across the executive team, and a line in the enterprise data strategy, as deliverables with the standing of anything you are building — one function at a time, each pitched on something of theirs that is broken today.

I keep bees, and the part that maps is not the honey. A colony with one queen and no capacity to raise another is not a strong colony having a good summer. It is a colony one bad afternoon from being over, looking exactly like the healthy one beside it until the afternoon.

Tomorrow, Isabel takes the same question into a platform — Profisee — and asks whether rapid time-to-value can hold a sponsor's attention past the first release. Standing disclosure, unchanged from last Monday: the practice behind this publication is a certified Profisee implementation partner, paid by clients for deployment work rather than by the vendor for the sale, and our services revenue is downstream of this category of software being chosen at all. Weigh her criticisms above her praise regardless.

Master data isn't a project you finish. It's a hive you keep — and a hive that cannot raise another queen is a hive with a countdown.