Data Consulting Companies
Data consulting companies, and how to tell them apart
Most shortlists of data consulting companies mix four very different kinds of firm and compare them on price. A global strategy firm, a systems integrator, a staffing agency and a senior-led boutique are not alternatives to each other. They solve different problems, and picking the wrong category is the most expensive mistake in the process. This page sorts the categories, says what each is actually for, and gives you the questions that separate a firm that will fix the problem from one that will staff it.
What separates a strong Data consulting company
Work out which problem you actually have first
There are three. Nobody can agree which number is right. The numbers agree but nobody trusts the pipeline underneath. Or the data is fine and the reporting is slow. Each points at a different kind of firm, and a shortlist built before you know which one you have will be judged on price because there is nothing else to judge it on.
The people who pitch should be the people who build
Ask who will be on the engagement in week six, by name. The common pattern is a senior architect in the sales meeting and a junior team by the end of the first month. A firm that will not name the team is telling you the answer.
A deliverable list, not a bucket of hours
A strong proposal reads as named artefacts with dates: a certified metric layer, a documented pipeline, a runbook your team owns. An hours estimate with a rate card attached transfers all the delivery risk to you.
Willing to tell you the project is wrong
The most useful thing a data consultancy ever says is that the dashboard problem is really a data problem, and the six week engagement you asked for is the wrong six weeks. A firm that agrees with your framing in the first meeting is selling hours.
Outcomes stated in hours and dollars
Ask for a figure a client would recognise. Reconciliation labour removed. Report turnaround cut from days to minutes. A migration avoided. Firms that answer in capability statements rather than numbers usually have no numbers.
A handover that leaves you able to run it
Documentation, lineage, a named owner on your side, and a runbook. Without those the engagement ends with you dependent on the firm, which is a commercial outcome for them and a cost for you.
Proof
- $2.1M a year: Reconciliation labour removed by unifying 14 regional metric definitions
- 47s to 9s: Dashboard response time after retiring 4,380 dead workbooks
- $6.2M: Platform migration avoided once the reporting estate was rationalised
The four kinds of data consulting company
They are not competitors. They are different products, and the shortlists that go wrong are the ones that treat them as interchangeable.
Global strategy firms
Accenture, Deloitte, BCG, Bain, EY and McKinsey. Built for portfolio-level decisions across a large organisation: operating models, multi-year transformation, board-facing cases. Strong when the question spans many business units and needs political cover. Expensive and slow for a scoped technical fix, and the senior people who win the work rarely deliver it.
Systems integrators and platform partners
Slalom, Tredence, Analytics8, Beyond Key and the large platform resellers. Built for delivery at scale once you have chosen a platform. Strong when the architecture is decided and the job is to build it properly with a lot of hands. Weaker when the real question is which platform, because the answer tends to match the partnership agreement.
Staffing and contract shops
They place engineers against your backlog. The cheapest way to add capacity when you already know what to build and have someone senior directing it. A poor fit when you need the problem diagnosed, because a contractor paid by the hour is not incentivised to tell you the project should be smaller.
Senior-led boutiques
Smaller firms where the person who scopes the work is the person who does it. Built for a specific, diagnosable problem: metrics that disagree, a pipeline nobody trusts, a reporting estate that has grown past what anyone can maintain. Thinklytics is one of these. The constraint is capacity rather than capability, so they are the wrong answer for a fifty person programme.
What data consulting companies actually deliver
Underneath the category names the work is fairly consistent. A diagnostic establishes which numbers disagree and why, and who owns each one, which is almost always less glamorous and more valuable than it sounds. Data foundation work fixes the pipelines, the identity resolution and the ownership underneath the reporting, and it is where the majority of real engagement time goes.
A semantic or metric layer is what makes a number mean the same thing in two different tools, and it is the deliverable that ends the meeting where two directors disagree about margin. Analytics and reporting is the visible layer on top, and it is the part clients ask for first and benefit from last, because a dashboard built on uncertified data just distributes the disagreement faster.
Governance is ownership, lineage, access and the audit trail, and enablement is what leaves your team able to run all of it. An engagement that skips the last two ends with a working system and no one accountable for it, which is a slower kind of failure.
Questions that separate the shortlist
Five questions, and the answers sort a list of eight firms into two quite quickly. Who is on this engagement in week six, by name and by seniority. What are the named deliverables and their dates, rather than the hours. What would make you tell us not to do this project. Show me an outcome in hours or dollars from a client of our size. And what does handover include, specifically.
A firm that answers all five concretely is worth a second meeting. A firm that answers in capability statements, methodology diagrams or a rate card is describing how it sells rather than how it delivers.
Where Thinklytics fits
Thinklytics is the fourth kind of firm on that list. Every engagement starts with one written question, and the first stage that finds it is priced separately, so you can stop there if what we find does not justify going further. The person who sits in the meeting where two directors disagree about margin is the person who later writes the semantic layer that settles it, so nothing is lost in a handover between a diagnosis team and a delivery team. That shape suits a specific diagnosable problem and does not suit a fifty person programme, and we will say so in the first call rather than the third.
Frequently asked questions
What do data consulting companies do?
They diagnose why the numbers disagree, fix the pipelines and ownership underneath the reporting, build the semantic layer that makes a metric mean the same thing everywhere, put governance around it, and train your team to run it. Firms differ mostly in how many of those steps they actually deliver rather than in how they describe the work.
What are the different types of data consulting firm?
Four. Global strategy firms for portfolio-level decisions across a large organisation. Systems integrators and platform partners for delivery at scale once the architecture is chosen. Staffing shops for adding engineers to a backlog you already understand. And senior-led boutiques for a specific diagnosable problem where the person who scopes the work does it. They are different products, and most bad shortlists compare them as if they were the same one.
How do I choose a data analytics consulting company?
Work out which of three problems you have first: nobody agrees which number is right, nobody trusts the pipeline underneath, or the reporting is simply slow. Then ask five questions. Who is on the engagement in week six by name. What are the named deliverables and dates. What would make you tell us not to do this. Show an outcome in hours or dollars from a client our size. What does handover include. Those answers sort a shortlist faster than any rate comparison.
What is the difference between a data consultancy and a staffing agency?
A staffing agency places engineers against a backlog you have already defined and is the cheapest way to add capacity when you know what to build. A consultancy is accountable for a diagnosis and an outcome, which includes telling you the project should be smaller or different. If a firm bills hours against your backlog and never challenges the brief, you are buying staffing whatever the proposal calls it.
How much do data consulting companies charge?
The drivers are the state of the data, how many systems have to agree, whether governance and enablement are included, and how much of the work is diagnosis against delivery. Thinklytics prices by deliverable rather than by an hours bucket, so scope and cost are agreed before work starts. Request the audit for a fixed quote.
Do I need a data consultant or a data engineer?
Hire an engineer when you know what to build and have someone senior who can direct and review it. Bring in a consultant when the question itself is unsettled, when four systems give four answers and nobody can say which is right, or when the reporting estate has grown past what anyone can maintain. The second case is a diagnosis problem, and more engineering hours tend to make it worse rather than better.
How long does a data consulting engagement take?
A scoped diagnostic runs in weeks rather than months and should produce a written answer plus a costed plan. Build work depends on how many systems have to agree: a certified metric layer across fourteen regional definitions took 11 weeks, and a reporting rationalisation that retired 4,380 workbooks ran alongside normal operations. Be suspicious of a multi-year plan proposed before anyone has diagnosed the problem.