How to set up an innovation lab in a government entity

To set up an innovation lab in a government entity, start with a written mandate signed by a senior sponsor, place the lab where it can reach the services it is meant to change, staff it with a small mixed team, and prove its worth on two or three real challenges before you ask for a room or a brand. Government labs live or die on governance, not creativity, so most of the steps below are about authority, ownership and reporting.

This article is the practical, government-specific companion to What an innovation lab is, and why most fail.

Step 1: Get the mandate in writing

A verbal "yes, go ahead" from a senior leader is not a mandate. In government, people change roles often, and an unwritten lab disappears with its sponsor.

Write one page and get it signed. It should state:

1. Purpose. Which kinds of problems the lab works on, in one sentence. For example: "redesigning citizen-facing services that generate repeat contact or complaints".
2. Sponsor. The named executive who owns the lab, plus a deputy sponsor from operations.
3. Decision rights. What the lab can do without further approval: talk to citizens, run small tests, access service data, invite staff from other departments for a set number of days.
4. Limits. What the lab will not do: build IT systems, replace existing project offices, or run public consultations that belong to another unit.
5. Review point. A date, usually after twelve months, when the sponsor decides whether to continue, change or close the lab, based on agreed evidence.

Step 2: Choose where the lab sits

Placement decides what the lab can touch. There are three common options, each with a trade-off:

• Under the head of the entity or the strategy office. Strong authority and visibility, but risk of being seen as the leadership's project rather than a service to departments.
• Under a service or operations directorate. Close to real problems and easier adoption, but harder to work across directorates.
• Under digital transformation or IT. Access to data and technology, but the lab gets pulled towards technology projects and away from policy and process.

Whatever you choose, map the relationships with a stakeholder map before launch: who must support the lab, who could block it, and who simply needs to be kept informed. If you are still deciding whether a lab is the right structure at all, compare it with the alternatives in Innovation lab vs centre of excellence vs innovation unit.

Step 3: Build a small, mixed team

Four to six people is enough to start. You need, between them:

• Someone who has run research with service users.
• A skilled facilitator who can hold a room of senior officials.
• Someone who can make quick prototypes: forms, scripts, mock screens, service walk-throughs.
• A respected insider who knows how decisions are really made in the entity, and who can get a meeting with any director.

That last role is often missing, and its absence is the main reason labs stay on the edge of the organisation. Second staff from other directorates into each challenge for a fixed period. They bring knowledge in and carry the methods back out.

Clarify roles early with a RACI for each challenge: who owns the decision, who does the work, who is consulted and who is informed.

Step 4: Pick two or three first challenges with owners

Your first challenges are your business case. Choose them with these tests:

1. A named director owns the problem and wants it solved.
2. The problem affects citizens, businesses or staff in a way people can see.
3. A first useful result is possible within three months.
4. Changing the service does not need new legislation.
5. Data or access to users is available.

Use a problem tree with the owner to separate causes from symptoms before you commit. Avoid flagship transformations as first challenges; they are slow, political and hard to attribute. The detailed selection method is in How to choose your lab's first challenges.

Step 5: Solve procurement, data and ethics before you need them

Government labs often stall in month two, not because of the work, but because a small test needs a permission nobody has. Prepare in advance:

• Procurement. Agree a simple route for small purchases such as printing, incentives for research participants, or a short contract with a specialist.
• Data access. Agree with the data owner which service data the lab may see, in what form, and who signs off.
• Research with citizens. Write a short consent form and a protocol for interviews and observation, cleared with legal and privacy colleagues.
• Testing in live services. Agree in advance what a small live test looks like, for example a concierge test where staff deliver a new service manually to a few citizens before any system changes.

Step 6: Report decisions, not activity

From the first quarter, report what changed because of the lab: decisions taken, services adjusted, projects stopped or redirected. Build a simple KPI tree that links the lab's work to measures the entity already cares about, such as repeat contact, processing time or complaints.

Plan how you will communicate with sponsors, directors and staff using a stakeholder communication plan. Short, regular updates to directors matter more than a polished annual report.

Step 7: Design the handover from the start

Every challenge needs an answer to one question at intake: if this works, who runs it, and with what budget? Without that answer, successful pilots wait for months. See From lab to operations: the handover nobody plans.

Common mistakes

• Launching the room before the work. A branded space and an opening ceremony raise expectations the lab cannot yet meet.
• One sponsor only. When that person moves, the lab is orphaned.
• Taking challenges nobody owns. The result will be admired and ignored.
• Becoming the entity's events team. Labs asked to run every workshop and hackathon lose time for real challenges.
• Hiding stopped projects. In government, a cheap early "no" is valuable evidence. Report it.
• Chasing external awards before internal results. Recognition outside the entity does not protect a lab whose directors see no benefit.

An illustrative example

Illustrative: an invented organisation, described to show the pattern.

A national tax administration in a mid-sized European country set up a lab under its deputy director-general, with the head of taxpayer services as co-sponsor. The mandate fitted on one page and named a review date twelve months out.

The team of five included a long-serving tax officer who knew every directorate head by name. The first challenge, owned by the head of taxpayer services, was why so many small traders filed late in their first year. Interviews showed many did not realise their first deadline differed from later ones.

The lab ran a concierge test: for one month, staff phoned a few dozen new traders before their first deadline. Late filing in that group dropped noticeably. The service directorate then built the reminder into its standard process, with its own budget.

At the twelve-month review the lab presented four decisions, two stopped ideas and one adopted process change. The sponsor renewed the mandate and widened it.

Questions people ask

Does a government innovation lab need its own budget? It needs a small, protected budget for testing and research, and the salaries of its core team. It should not need large project budgets; those belong to the directorates that adopt the results.

Who should a government innovation lab report to? To a senior executive with authority across directorates, with a co-sponsor from operations. The reporting line matters less than having two committed sponsors and a written mandate.

How is a government lab different from a digital transformation office? A digital transformation office delivers technology change across the entity. A lab tests whether a service, process or policy change will work for the people affected, often before any technology is chosen. They work best as partners.

Where to go next

The guides most useful in a government lab's first months are the stakeholder map, problem tree, concierge test and KPI tree. For the wider picture, read What an innovation lab is, and why most fail and 5 reasons innovation labs get closed in year two.