The Federal Data Strategy’s hardest actions were about people

The 2020 Action Plan, released on December 23, 2019, read like a technology program. Its foundation was organizational: authority for a data governance body, staff hours to run it and measured data skills. The people milestones lagged while most shared tools arrived on schedule.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

The 2020 Action Plan, released on December 23, 2019, read like a technology program. Its foundation was organizational: authority for a data governance body, staff hours to run it and measured data skills. The people milestones lagged while most shared tools arrived on schedule.

  1. The week On December 23, 2019, the Administration released the Federal Data Strategy 2020 Action Plan.
  2. How the plan was built, and how it was scored The plan’s structure was unusually explicit.
  3. Authority: who decides about data The plan’s second action asked each agency to constitute an inclusive and empowered data governance body, chaired by the Chief Data Officer.
  4. Hours: staff the governance body can use The plan asked agencies, after assessing staff skills, to dedicate staff to support the governance body.
  5. Skills: assess, find the gap, plan to close it The fourth Agency Action asked the large agencies to assess the data literacy and data skills of their staff.
  6. The strongest objections The most obvious objection is timing.
  7. What state and local agencies can take from it The Federal Data Strategy bound federal agencies only.
  8. With hindsight The Chief Data Officers Council met for the first time on January 31, 2020, completing milestone 7.1 of the plan.

The week

On December 23, 2019, the Administration released the Federal Data Strategy 2020 Action Plan. The announcement on the strategy’s website called it a flagship product of the President’s Management Agenda. It said the plan came out of Cross-Agency Priority Goal 2, on data as a strategic asset. That goal was led by the Office of Management and Budget, the Office of Science and Technology Policy, the Department of Commerce and the Small Business Administration. Federal agencies, the announcement said, were reviewing the plan as they prepared to carry out its actions in calendar year 2020.

The plan sat on two earlier foundations. In June 2019, OMB memorandum M-19-18 had set out the Federal Data Strategy itself: a mission statement, principles and 40 practices, with a vision running to 2030. It called for annual action plans to pick the steps for each year. Separately, the Foundations for Evidence-Based Policymaking Act of 2018, known as the Evidence Act, had become law in January 2019. Among other things it required each agency to designate a Chief Data Officer and created a council of those officers. As the action plan recorded, OMB’s July 2019 guidance on the Act, M-19-23, required all agencies to establish a data governance body chaired by the Chief Data Officer by September 30, 2019.

A draft action plan had been published earlier for comment, and the strategy’s site posted both the draft and the revisions made in response. The final plan contained 20 actions in three groups. Agency Actions were steps each agency had to take. The other two groups, Community of Practice Actions and Shared Solution Actions, covered work by groups of agencies, pilot projects and government-wide efforts to support all agencies through shared tools and resources.

Six Agency Actions came first, described as foundational. Agencies were to identify the data needed to answer their priority questions, building on the learning agendas the Evidence Act required. They were to constitute a diverse data governance body and publish its membership, charter and meeting cadence by January 31, 2020. They were to assess the maturity of their data and data infrastructure. They were to identify opportunities to increase staff data skills. And they were to identify priority data assets for open data plans and to publish and update data inventories.

The plan explained what it meant by some of these terms. A maturity assessment, it said, lets an organization evaluate itself against documented practices, find gaps and set priorities. It was to cover governance, data management, culture, systems and tools, analytics, staff skills, resources and legal compliance. The governance body was to choose the maturity model. The results were meant to feed investment decisions and to be repeated over time to measure progress.

The shared actions included launching the federal Chief Data Officers Council by January 2020 and creating an OMB Federal Data Policy Committee. GSA was to build a curated catalog of data skills training by November 2020 and a data ethics framework by December 2020. A data protection toolkit and a repository of resources at resources.data.gov were also promised. Each milestone had a target date, a measure, a reporting route and a label saying whether it was required or encouraged. The plan said each agency’s Chief Data Officer would be the main point of contact for progress reporting.

Read today, the 2020 Action Plan looks like a technology program. It talks about inventories, catalogs, metadata tools and a repository of resources. It is easy to read it as a set of data projects for technology staff to deliver.

Our argument is that the plan’s foundation was organizational, and that agencies which treated it as a technology plan built on sand. Its first actions asked agencies to supply three things that no system can provide. The first was authority: a body with the standing to make decisions about data across the agency. The second was hours: named staff whose time belonged to that work. The third was skills: a measured picture of what staff could do with data, and a plan to close the gap.

These three are distinct, and together they cover what the plan needed from agencies before any tool could help. Authority decides what data work happens. Hours determine whether it happens. Skills determine whether it is done well. The government’s own progress tracking, read with hindsight, shows that the launch milestones were met on time and the people milestones lagged.

A first section explains how the plan was built and scored, because its design shaped what agencies did. The three supplies follow in turn. A fifth section tests the argument against the strongest objections. The last considers what the plan offers state and local agencies, which were outside it but faced the same choices.

How the plan was built, and how it was scored

Milestones, measures and labels

The plan’s structure was unusually explicit. Each action had milestones. Each milestone had a measurement, a target date, a reporting mechanism and a label: required for all agencies, required for agencies covered by the Chief Financial Officers Act, or encouraged. The reporting mechanism was often a progress report tool, with a unique link for each Chief Data Officer.

That design had a predictable effect. Milestones measured by completion of a document were easy to report. Publishing a governance body’s charter on an agency web page by January 31, 2020 was a yes or no. Milestones that required sustained work, such as dedicating staff or closing skills gaps, were measured by counts or by completion of an assessment. They depended on budgets and hiring, which the plan did not control.

Starting from questions, not data

The first Agency Action set the logic for the rest. Agencies were to identify the data needed to answer their priority questions. The plan tied this to the learning agendas the Evidence Act required of the large agencies: multi-year lists of the most important questions about their programs, policies and regulations. It called the approach a shift. Research would start from the agency’s questions rather than from whatever data happened to exist.

The milestones asked agencies to consult their Chief Data Officers, Evaluation Officers and Statistical Officials, to identify data needs in an interim learning agenda, and to plan how to get the data. The plan noted that the data might already exist in the agency, in another agency or outside government. If so, the next step was access or an acquisition strategy. If not, a new collection might be justified.

This action already depended on people. Three officers had to work together, and program leaders had to agree on which questions mattered. The data work followed from that agreement. Where the agreement was missing, the inventory and catalog work that came later had no clear purpose.

Required versus encouraged

The labels mattered as much as the dates. Publishing governance materials and selecting a maturity model were required of all agencies. Dedicating staff to the governance body, documenting how it received its authority and conducting the maturity assessment were encouraged. The staff skills assessment and gap analysis were required only of the large agencies covered by the CFO Act, and encouraged for the rest.

An agency with limited resources reading these labels would sensibly do the required items first. That is what the labels invited. The consequence was that the parts of the plan most likely to change how an agency worked were the parts most easily deferred.

A ten-year frame with a one-year plan

The strategy itself looked ahead ten years. The action plan covered one. The plan described its role as building capacity, establishing processes and aligning existing efforts. That framing was honest. It also meant that the most valuable outcomes of 2020 would be invisible in 2020: a governance body that worked, staff who could do the job, and the habits that made later plans possible.

Authority: who decides about data

A body, not a person

The plan’s second action asked each agency to constitute an inclusive and empowered data governance body, chaired by the Chief Data Officer. Senior staff from business units, data functions and financial management were to take part. By November 2020, agencies were to have begun enterprise-wide data governance, including capital planning for data assets and preparing data for use in artificial intelligence.

The choice of a body over a single officer was deliberate. Data in a federal agency belongs to programs. The people who collect it, fund the systems that hold it and answer for its quality sit in different parts of the organization. A Chief Data Officer acting alone can advise. A body that includes the owners of the data can decide.

Where the authority comes from

One milestone asked agencies to document how the governance body received its authority. It was encouraged rather than required, and it was the milestone that mattered most. A body chartered by the agency head, with decision rights written into the charter, can settle disputes between programs. A body convened by the Chief Data Officer without a delegation from above can only recommend.

The Evidence Act did not settle the question either. As GAO later noted, the Act did not specify where a Chief Data Officer should sit or to whom the officer should report. OMB’s guidance said the officer should be placed high enough to engage regularly with agency leadership. In practice, placement varied. GAO found that two of six agencies it reviewed placed their Chief Data Officers in offices led by the Chief Information Officer. The other four placed them in offices with planning and data management responsibilities.

What authority looks like on paper

For an agency head, the useful test is whether the governance body can make three kinds of decision without escalation. It should be able to set data standards that programs must follow. It should be able to prioritize data investments across programs. And it should be able to resolve requests to share data between programs, within the law. If any of these needs the agency head’s personal intervention each time, the body has not been given the authority the plan assumed.

A charter should say so in plain words: what the body decides, what it recommends, and who resolves disagreements. It should also say how often it meets and who attends, which the plan asked agencies to publish. The publication requirement had a further effect. It made the charter a public document that staff, oversight bodies and the public could hold the agency to.

Authority over data that leaves the agency

Some of the plan’s later actions required decisions that cross legal lines. Open data plans, under the fifth action, were to set processes for publishing data in open formats and for engaging the public. They were to name a point of contact for quality complaints, and set a process to evaluate and improve timeliness, completeness, accuracy and usefulness. Each of those is a decision about what leaves the agency and on what terms.

A Chief Data Officer cannot make those decisions alone. Privacy, records, legal and program officials each hold part of the answer. A governance body that brings them together, with authority to decide, turns a series of separate approvals into one. Without it, each dataset goes through its own negotiation, and publication slows to the pace of the most cautious reviewer.

Authority shows up in the budget

The plan offered agencies three encouraged activities for 2020, one per quarter in any order. They were a data strategy or road map, a plan for capital planning for enterprise data assets and infrastructure, and a master data management program. The capital planning item deserves attention. It asks whether data assets are planned and funded like other capital investments, with owners, life cycles and replacement costs.

A governance body that has a voice in that planning has real authority. One that is consulted after budgets are set does not. The progress page later showed that ten agencies had developed such a capital planning approach in 2020, and seven had adopted master data management. For the other agencies, the 2020 record shows no such plan yet.

Signs a governance body lacks authority

Some signs show that a governance body has the form of authority without the substance:

  • Members send deputies who cannot commit their programs.
  • Decisions are recorded as recommendations and then reopened by programs that disagree.
  • Data investment requests go straight to the budget office without passing through the body.
  • The body’s agenda is mostly briefings, with few items for decision.
  • Disputes about sharing data between programs are settled by the agency’s front office case by case.

Each of these can be fixed in the charter or in practice. In our view the fix needs the agency head or deputy to say, once and in writing, that the body’s decisions stand unless the head overrules them.

Authority and the CIO

Placing the Chief Data Officer under the Chief Information Officer had advantages. It put data governance close to the systems and budgets that carry data. It also carried a risk. Data questions would become infrastructure questions, settled by what the technology could do rather than what the programs needed. Placing the officer elsewhere had the opposite risk: governance without access to the systems.

Neither placement is wrong. What matters is that the charter compensates for the weakness of the chosen one. An officer inside the CIO’s organization needs program leaders on the governance body with real votes. An officer outside it needs a standing seat in IT investment decisions.

Hours: staff the governance body can use

The milestone that counted people

The plan asked agencies, after assessing staff skills, to dedicate staff to support the governance body. The measure was the number of full-time equivalents dedicated. The target date was September 30, 2020, later moved to October 31 because of the pandemic. Like the authority milestone, it was encouraged, not required.

This was the plan’s most concrete test of commitment. Charters and assessments cost little to produce. Dedicated staff cost salary, and they come from somewhere. An agency that named staff whose time belonged to data governance had made a budget decision. One that did not had made a different one.

What the counts showed

The strategy’s progress page, last updated in January 2021, reported that 24 agencies had dedicated staff to their governance bodies. The average was 2.7 full-time equivalents. The count of agencies covered those that had reported a designated Chief Data Officer to OMB.

Two or three people can do a great deal of coordination. They cannot, on their own, run an inventory of an agency’s data assets, assess maturity across dozens of programs and build open data plans. Those tasks depend on the time of staff in the programs themselves. The plan counted dedicated governance staff. It did not count the program staff time its other actions consumed, and no agency budget line covered it by default.

Where the hours come from

For an agency planning data work, the useful step is to make that hidden time visible. Each action in a data plan can be costed in hours from three sources: the central data team, program data stewards and IT staff. A program that will provide stewards should know how many hours, for how long, and what work will stop to free them.

Data steward roles are one mechanism. A steward is a named person in a program who answers for a data asset’s definitions, quality and access. The role works when it appears in the person’s position description and performance plan. It fails when it is added as an unpaid extra. The 2020 plan did not prescribe stewards, but its inventories and open data plans assumed someone in each program would do that work.

Costing one action

Take the maturity assessment as an illustration. The plan asked agencies to select a model and then conduct and document an initial assessment. On paper that is two milestones. In an agency with many programs, it means interviews or surveys in each program, a review of policies and systems, scoring against the model and a report the governance body can act on.

The central data team can design the assessment and write the report. It cannot answer for each program’s data. Program stewards or managers have to describe their data, systems and practices, and IT staff have to describe the infrastructure. If each program gives a few days of a knowledgeable person’s time, the total across a large department is substantial. None of it appears in the milestone’s measure, which was simply completion.

A costing like this does two things. It tells programs what is being asked of them before the request arrives. And it shows leaders the true price of the plan, so they can decide what to fund and what to drop. Plans that skip it risk discovering the price later, as missed dates.

Hours and the pandemic

The pandemic tested this directly. Consistent with OMB’s guidance on operational alignment during COVID-19, several target dates in the plan were extended. Agencies were directed to prioritize COVID-19 response data above other data assets. Staff time that might have gone to inventories and maturity assessments now competed with urgent reporting.

That priority was right. It also showed how thin the capacity was. An agency with dedicated governance staff and named stewards could absorb an emergency data task without stopping everything else. One that relied on borrowed time could not. The emergency did not create the capacity gap. It exposed it.

Skills: assess, find the gap, plan to close it

The four-step method

The fourth Agency Action asked the large agencies to assess the data literacy and data skills of their staff. The plan gave a four-part method. Identify the critical data skills the agency needs, and assess current staff capacity for them. Then perform a gap analysis to prioritize needs, and identify and carry out approaches to fill them.

The plan was specific about scope. It said agencies should consider a minimum level of data literacy for all staff, including those whose roles were not traditionally seen as data roles. It listed options for closing gaps: new software tools, easy-to-use dashboards, training, rotations, communities of practice, and hiring and retention strategies. It also said the assessment should form part of the agency’s ongoing human capital management, rather than a separate exercise.

What the counts showed

The skills milestones moved slowest. According to the progress page, nine agencies had performed an assessment of staff data literacy and skills by July 31, 2020. Five had conducted a gap analysis by September 30. Six had established a baseline and a performance plan to close the gaps, against a December 31, 2020 target.

By contrast, the launch milestones were complete. OMB launched the Chief Data Officers Council on January 31, 2020, and the progress page recorded 65 officers and representatives at the first meeting. The OMB data policy committee had been meeting monthly since November 2019. The curated data skills catalog and the data ethics framework were both published by the end of 2020. The shared tools existed. Measuring the people who were supposed to use them lagged behind.

The gap is clearer when the counts sit side by side. Thirty-three agencies had published governance materials, a required milestone. Twenty-four had dedicated staff. Nine had assessed skills. Each step further from a document and closer to people saw fewer agencies complete it. The skills milestones were required only of the large agencies covered by the CFO Act, so fewer agencies were expected to finish them. Even so, fewer than ten had done so by the target date, in the middle of a year in which data skills were in sudden demand.

Literacy for everyone, depth for some

The plan’s distinction between minimum literacy for all staff and deeper skills for data roles is the useful one. Literacy means being able to read a chart correctly, question a figure and know where the agency’s data comes from and what it cannot show. Depth means being able to clean, link, analyze and model data, and to manage it over time.

The two need different methods. Literacy can be raised through short training, plain-language data documentation and dashboards designed for non-specialists. Depth needs hiring, sustained training, rotations and communities of practice, which the plan listed. An agency that treats them as one problem risks sending everyone on the same course. The result can be too much for some staff and too little for those who need depth.

Why skills work stalls

Skills assessments stall for reasons that have little to do with data. They need human resources offices to define competencies and position descriptions. They need managers to rate staff, which raises labor relations questions. They need a budget for training once gaps are found. And they produce uncomfortable findings. An assessment that shows most staff cannot work with the agency’s data invites the question of why the agency has been making data-based decisions.

They also lacked a home in the personnel system. In 2020, there was no federal occupational series for data science. OPM had allowed agencies since June 2019 to add the parenthetical title “Data Scientist” to positions in existing series. That left data science work spread across existing occupations, which made data roles hard to count and compare across agencies.

Measuring whether it worked

A skills plan also needs a way to tell whether it is working. Course completions are easy to count and say little. Better measures look at the work. How long does it take to answer a priority question with agency data? How many requests for data or analysis wait in a queue, and for how long? How often do program reports need correction because a figure was misread?

These measures connect skills to the first Agency Action, which started from priority questions. If the skills plan works, the agency answers more of its questions, faster, with its own staff. If those measures do not move, more training of the same kind is unlikely to help, and the mix of hiring, contracting and redesign should change.

Train, hire, contract or redesign

Each skills gap presents the same four options, with different trade-offs. Training existing staff keeps program knowledge but takes time, and trained staff may leave for better-paid roles. Hiring brings skills quickly but competes with the private sector on pay, and new hires take time to learn the programs. Contracting fills gaps fastest but can leave the agency unable to understand or maintain what contractors build. Redesigning the work, for example by automating a routine report, removes the need for the skill altogether.

The right mix depends on whether the skill is core to the mission and how often it is needed. A skill used daily in a core program should be held by staff. A skill used for one project can be contracted, with a requirement to transfer knowledge. Writing these choices down for each gap is what turns a skills assessment into a workforce plan.

What a usable skills plan contains

We would give a skills plan a small number of parts:

  • A short list of critical data skills, tied to the agency’s priority questions and the systems it actually uses.
  • A baseline from a light self-assessment for all staff and a deeper one for roles that work with data daily.
  • A decision for each gap about whether to train, hire, contract or redesign the work.
  • Measures that show change, such as the share of priority questions answered with agency data within a set time.
  • An owner in the human capital office, working with the Chief Data Officer.

The last point is the easiest to miss. A skills plan owned by the data office alone has no access to hiring, classification or training budgets. One owned jointly with human capital can use them.

The strongest objections

The pandemic, not the design, explains the gaps

The most obvious objection is timing. The plan’s first year collided with COVID-19, which pulled data staff into emergency work and pushed back target dates. Judging the plan on 2020 is unfair.

The pandemic explains some delay. It does not explain the pattern. Launch milestones that depended on central bodies were met, many on time. Milestones that depended on agencies finding staff and measuring skills lagged. The pandemic affected both kinds of work, so the difference points to how the plan was designed and resourced.

Small agencies were never going to do this

A second objection is scale. The plan applied to large departments and small commissions alike. A small agency cannot staff a governance body or run a skills assessment the way a cabinet department can.

That is true, and the plan partly recognized it by requiring the skills work only of the large agencies. But the argument applies with more force to small ones. A small agency has fewer people, so the authority and hours questions are sharper. Two people who already hold other jobs cannot be both the governance body and its staff. For small agencies, shared services and shared roles across agencies are the realistic answer. The shared actions, such as the skills catalog, were meant to support every agency, including them.

Technology was the binding constraint

A third objection is that many agencies lacked the basic tools: catalogs, metadata management and secure ways to share data. Without them, governance had little to govern. The plan’s own shared actions, including a metadata tool and a research data portal, recognized that need.

Tools matter. But the progress record shows that tools were delivered faster than the organizations that would use them. A metadata tool without stewards to maintain the metadata produces an incomplete catalog. A data portal without staff who can answer requests produces a queue. Technology was a constraint. It was not the one that bound first.

The law already required this

A further objection is that the plan added little. The Evidence Act already required Chief Data Officers, learning agendas and a council. OMB’s 2019 guidance already required governance bodies. The action plan, on this view, restated obligations with new dates.

Partly so. But restating obligations as dated, measured milestones was the plan’s contribution, and it produced the record that shows where agencies stalled. The law created offices. It did not supply the authority, hours and skills that make offices effective. The plan’s milestones were the first government-wide attempt to measure those, and the measurements are the evidence for this argument.

Governance slows delivery

Another objection comes from program and technology leaders who need results quickly. A governance body adds meetings and approvals. A skills assessment takes months. Meanwhile, urgent data work waits.

Badly designed governance does slow delivery. Good governance speeds it up, by settling standards and sharing questions once instead of for every project. The difference lies in what the body decides and how fast. A body that meets monthly to review every dataset will become a bottleneck. One that sets standards, delegates routine decisions to stewards and handles only exceptions will remove one. The plan’s emphasis on authority was an argument for the second design.

Plans like this are box-ticking

A sixth objection is cynicism: government-wide plans produce compliance documents, and agencies treat them that way. The plan’s labels and progress tool rewarded exactly that behavior.

There is something to this, and it supports the argument. The milestones easiest to tick were the documents. The hardest were the ones that required reallocating people. An agency that wanted more than a ticked box needed to treat authority, hours and skills as its own priorities, not as items in OMB’s tool.

What state and local agencies can take from it

Outside the plan, facing the same choices

The Federal Data Strategy bound federal agencies only. States, counties and cities were outside it. Those that build their own data programs, with chief data officers, open data portals and governance councils, face the same choices. The federal plan offers them a useful template, because its milestones are explicit and its progress record shows where the effort stalled.

The federal plan’s reporting design is itself worth copying. Each milestone had a measure and a date, and progress was published agency by agency. A state that publishes its own data governance milestones in the same way gives its legislature and public a clear view of what has been done. It also gives the data office evidence when it asks for staff. A published record of stalled skills milestones is a stronger budget argument than a general plea for capacity.

Start with the charter

For a state agency or a city, the first step is the same as the federal one: a governance body with written authority. A city data council convened by a chief data officer without a mandate from the mayor or city manager is advisory. One chartered by executive order, with decision rights over standards and sharing, is not. State agencies face the added complication that some of their data belongs to federally funded programs with their own rules. The charter should name those constraints rather than ignore them.

Cost the hours honestly

Where budgets are tight, hidden time is costlier. A county that launches an open data portal without costing the program staff time to maintain each dataset will find the datasets go stale. A better approach publishes fewer datasets, each with a named steward and a refresh schedule the program has agreed to.

Measure skills before buying tools

Finally, a state or local agency can run the four-step skills method at small scale before buying analytics tools. If the assessment shows that few staff can use a self-service dashboard, the dashboard purchase can wait. Training or a small central analytics team may come first. The federal plan listed dashboards and tools among the ways to close skills gaps. That only works where staff can use what is bought.

Federally funded programs bring their own rules

A state agency may hold data from programs funded by federal agencies, such as health, nutrition, labor and education programs. Those programs can carry federal rules on use, sharing and confidentiality. A state governance body cannot override them. It can map them, so that staff know which datasets can be shared, with whom and under what agreement. That map can be the most valuable early product of a state data council, and it depends on staff from each program to build it.

Borrow the federal tools

The shared products of the federal plan were public. The data skills catalog, the data ethics framework and the data protection toolkit were published for anyone to use, and resources.data.gov collected them. A state or local agency building its own program can adapt them instead of starting from nothing. The ethics framework, for example, was meant to help staff identify and weigh the benefits and risks of acquiring, managing and using data. That is the kind of decision a new governance body faces early.

The same sequence applies to health systems and banks building data governance, though their obligations come from other rules. Authority, hours and skills come before tools. The federal record from 2020 is a public demonstration of what happens when the order is reversed.

With hindsight

The Chief Data Officers Council met for the first time on January 31, 2020, completing milestone 7.1 of the plan. More than 60 officers and their representatives attended. Over its next few meetings the council was to elect a chair, develop a governance framework, set priorities and create subcommittees.

The first-year results arrived during a pandemic. The strategy’s site records that, consistent with OMB’s COVID-19 guidance, several target dates were extended and agencies were told to prioritize response data. The progress page, last updated in January 2021, showed the pattern described above. Launch milestones and most shared milestones were complete. People milestones were partly complete: 24 agencies had dedicated staff to governance, 20 had conducted maturity assessments by September 30, 2020, and fewer than ten had assessed staff skills.

A 2021 Action Plan followed in October 2021, according to GAO. It described 2021 as a transition year for the executive branch, with aspirational milestones that agencies might only begin before the end of the year. It let agencies finish actions not fully met in 2020 and continue their foundational work, and it also focused on community of practice and shared solution actions across agencies. It also set out a maturity ladder. Foundational governance and infrastructure came first, around 2020 to 2022. Enterprise standards and budgeting were to follow around 2023 to 2025, then optimized self-service analytics around 2026 to 2028.

GAO reviewed the Chief Data Officer Council in December 2022. It found that the council had acted on each of its five statutory requirements, and that its membership included Chief Data Officers from 82 agencies as of October 2022. GAO also noted that the council had lacked fully formed plans and performance measures until it updated them in August 2022. GAO said the Evidence Act ends the council two years after GAO submits that report to Congress and that, because the report issued in December 2022, the council’s authorization would expire in December 2024. Among agency examples, GSA had begun offering role-based data training in February 2022, after an assessment identified gaps in staff data skills.

The same report cited GAO’s 2020 survey of federal managers. Across government, only about one-third to half of managers reported that aspects of evidence-building capacity, such as staff with relevant skills, were present. Skills remained the weak link.

The personnel system caught up in part. On December 20, 2021, OPM released a classification flysheet and qualifications standard for a new Data Science Series, 1560, citing the Evidence Act. Agencies had twelve months to apply it to covered positions. The memo kept the option of the “Data Scientist” parenthetical for data science work done in other series, such as statistics or epidemiology, so data roles could still sit in several occupations; OPM’s memo named the Epidemiology and Statistician series among them. On January 15, 2025, OMB issued M-25-05, the Phase 2 guidance under the Evidence Act on open government data. It rescinded and replaced OMB’s earlier open data policy, M-13-13. It assigned implementation to each agency’s Chief Data Officer, acting with the data governance body, and required coordination with the privacy, FOIA, legal and information officers. The strategy’s own site now says it is being retired, with its content moving to resources.data.gov.

What held up was the 2020 plan’s order of work. Later guidance kept returning to the governance body and the Chief Data Officer as the engine. What did not hold up was the assumption that agencies would find the people from existing budgets.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • Federal Data Strategy
  • Evidence Act
  • chief data officer
  • data governance
  • workforce skills