How to choose a decision intelligence or analytics partner

Start from your decisions, prove the governance, price fixed phases, show platform depth, transfer the skills, say what not to buy, measure adoption.

You choose a partner to deliver decision intelligence or advanced analytics by testing seven things before you sign: whether they start from the decisions your business makes rather than from a technology; whether they can show how they agree definitions and govern the data, because every model, dashboard and agent amplifies whatever it is given; whether they price fixed phases with named outputs rather than open-ended days; whether they have delivered on the platform you run, Microsoft Fabric and Power BI, Pyramid Analytics or ServiceNow ADA (Autonomous Data Analytics); whether enablement is built in so the capability moves into your team; whether they will tell you what not to buy; and whether they measure success by decisions made rather than reports delivered. Ask for a first phase of around four weeks that produces a decision map, signed definitions and a costed plan before you commit to a build, and ask to speak to a client of similar size. This post sets out each test, the questions that expose a weak answer, and the red flags.

Test one: they start from the decision

Decision intelligence is the discipline of making specific business decisions better and faster with data, from the evidence to the action. A partner who opens with a platform demonstration is selling the platform. A partner who asks which decisions matter most, who makes them, how often, with what information, and what happens next, is starting in the right place. Ask them to name the first three decisions they would work on in your business after an hour's conversation. What decision intelligence actually means sets out the distinction.

Test two: they can show the governance

Every advanced analytics project rests on definitions: what revenue, margin, an active customer or a late order mean, signed by the people who own them. Ask how they agree definitions, where they are written down, how they are tested against the ledger, and who owns each one after the project ends. A partner who treats governance as a later phase will hand you a model that produces confident wrong answers. How do you build a single trusted analytics layer? describes how to do it.

Test three: fixed phases with named outputs

Ask what the first phase produces, what it costs and when it ends. A good answer names the outputs: a decision map, signed definitions, a source and ownership register, an architecture and a priced backlog. A weak answer is a day rate and an estimate. Fixed phases protect you from scope that grows because nobody decided what was in it. Fixed price, not time and materials explains how the pricing works.

Test four: depth on your platform

Decision intelligence is delivered on a platform, and the partner should know yours in depth. On Microsoft, that means Fabric, OneLake, a governed Power BI semantic model, Copilot and Fabric data agents, and Activator for triggers. On Pyramid or ServiceNow ADA, it means the semantic layer, generative BI, KPI monitoring and the workflow hand-off. Ask for certifications held by the people who will do the work, not the firm's logo wall, and ask what they would build on the platform you already pay for before suggesting another. How can decision intelligence be built on Microsoft Fabric and Power BI? shows the Microsoft route.

Test five: the capability moves to you

Ask who will run it in a year. A good partner plans enablement from the start: your analysts trained on the model, your power users running the next sessions, documentation your team recognises, and routine changes done without a call. A partner whose model only works while they are engaged has built dependency, not capability.

Test six: they will tell you what not to buy

The most useful thing a partner says in the first meeting is often no: you do not need a new platform, you need the one you have governed; you do not need an agent yet, you need a threshold someone owns. Ask what they would not do in your position. If the answer is nothing, the advice is a sales pitch.

Test seven: they measure decisions, not deliverables

Reports delivered and logins counted measure activity. Ask how they will know the work is succeeding: which decisions are now made with the platform, how much faster, and with what result. What does a decision intelligence maturity model look like? sets out the levels and how to measure progress through them.

Red flags

A proposal that leads with licences. A team you will not meet until after signature. Day rates with no defined end. Governance described as a phase two. AI agents in the first month. References only from much larger organisations. And no answer to the question of who owns each definition when the partner leaves.

How we answer the seven tests

We are a Microsoft Solutions Partner, Pyramid Analytics' UK Partner of the Year and a ServiceNow partner, and we deliver through the Analytics Acceleration Programme: a four-week Establish phase, a build against a priced plan, and a monthly continuity rhythm, at published indicative prices of £52.7k, £73.9k and £114.5k over 24 months by business size. Our AI and analytics page describes the work, and the free audit is a 45-minute conversation that will tell you which decisions to start with, whether or not you work with us.

How do you choose a partner to deliver decision intelligence or advanced analytics? Seven tests before you sign