Putting Delivery Claims Next to Public Sentiment

Two versions of the same district.

A government client had no shortage of reporting. Departments submitted progress against their own development claims, and the numbers added up. What none of that reporting could answer was whether citizens in the same areas recognised the delivery being claimed on their behalf.

The two pictures sat in different places and arrived on different timetables. Progress reporting came up through departments on a departmental cycle. Public feeling arrived through coverage, complaints and constituency pressure, usually only once it had already become a problem. Senior officials were therefore asked to make planning decisions with one half of the picture in front of them.

Build one screen, not two reports.

Ahmad Mehboob, Co-Founder and COO, co-led the design and deployment of a real-time system whose whole purpose was to hold both readings side by side. The governing design constraint was the one that always applies in government: a senior official will not use a dashboard that needs explaining.

Establish the claim side

Departmental development claims were reduced to a small set of comparable indicators — what was committed, where, and its reported status — so that delivery could be read by area rather than by department.

Establish the sentiment side

Public sentiment for the same areas was collected through perception surveys, stakeholder interviews and monitoring of public and media commentary. Responses were coded to recurring themes, so that sentiment could be counted, compared between areas and tracked across periods rather than quoted anecdotally.

Join them on a common frame

Both streams were mapped to the same geography and the same reporting period. That single decision is what made the system useful: any area could be read in both directions at once, and a gap between the two showed up without anyone having to go looking for it.

Make it current, and make it checked

Field data was captured on mobile-based tools, transferred securely, and validated remotely before it reached the dashboard, so that a reading was current and verified rather than merely fast. The dashboard itself was kept plain: areas, themes, direction of travel, and nothing on the screen that did not support a decision.

The gap was the finding.

Delivery and perception did not move together. There were areas where reported progress was strong and public sentiment was not, and areas where sentiment ran ahead of what the reporting alone would predict. That told the client something a progress report structurally cannot: delivering work and being credited with it are two separate tasks, and the second one was failing in specific, nameable places.

The coded themes explained part of why. Where sentiment lagged delivery, the recurring themes were rarely about the quality of the work. They were about not knowing it had happened, not knowing whom to ask, or not encountering the part of it that touches daily life. That is a communication and sequencing problem rather than a delivery one, and it calls for a different kind of fix.

Manage the gap, not the average.

The recommendation was to treat the distance between claim and sentiment as an operating measure in its own right, and to review it on the same cycle as delivery.

Review both readings in the same meeting

Delivery reviews were advised to take the sentiment reading for the same area at the same time, so that a department reporting completion and a public that does not recognise it become one agenda item instead of two unconnected ones.

Treat a widening gap as an early warning

A gap that widens between reporting periods was to be escalated on its own, before it arrives as complaint volume or public pressure. Early is the only point at which the response is cheap.

Respond where the themes actually point

Where coded themes showed that citizens did not know work had been done, or did not know where to raise an issue, the recommended response was communication and access — not further capital work.

Keep the instrument honest

Survey wording, sampling and coding definitions were to be held stable between periods, and field validation kept in place. A monitoring system loses its value the moment a change in the reading might only be a change in the method.

What the client could then do.

The system gave the client one place where delivery claims and public sentiment for the same area could be read together, current enough to act on inside a planning cycle rather than after it.

With both readings in front of them, senior officials could:

• Separate areas that were genuinely under-served from areas that were served but poorly understood, and treat those as different problems
• Test a departmental claim against an independent reading of the same area, in the same meeting
• Direct communication and follow-up to the places the themes pointed at, rather than spreading it evenly
• Report progress in terms of what citizens could see, including against Sustainable Development Goal benchmarks, rather than activity completed

The system decided nothing on its own. It made the disagreement between two versions of the same district visible early enough for someone to act on it.