What a reporting reconciliation engagement should deliver, and how to scope one

Six artifacts, and if the proposal in front of you does not name all six it is a dashboard project with a reconciliation title. The deliverables, the acceptance criteria to write into the contract, the residual running cost, and what to ask before signing.

You have two numbers for the same metric, someone senior has noticed, and a proposal is now in front of you. This is what it has to contain.

The short version: six artifacts. If the proposal does not name all six, what you are buying is a dashboard project with a reconciliation title, and it will produce a third version of the contested number rather than settling the first two.

The six deliverables

The six artifacts a reconciliation engagement has to hand over

If a proposal does not name all six, it is a dashboard project with a reconciliation title.

• A written definition per disputed metric, with a named owner. Timing, grain, scope, adjustments and currency answered in prose a non-technical reader can check. An owner who is a person, not a committee.
• A reconciliation bridge from each old number to the certified one. Line by line, so every person who defended the old figure can see exactly which rule moved their number and by how much.
• Traced lineage for every number in scope. Source table to transformation to reporting surface, readable without opening the code.
• One enforcement point where the calculation lives. A metric layer the reporting tools query, so the definition cannot be re-implemented differently in the next dashboard.
• A standing forum with the authority to settle the next dispute. The definitions will be challenged again. Without a named forum the answer reverts to whoever shouts.
• A change log on the definitions themselves. Needed the first time someone asks why last quarter's figure moved.
• A new dashboard showing the certified number. Useful, and not a deliverable. A third version of a contested metric is the problem, not the resolution.

Kaiser Permanente's engagement delivered the definition, the certified layer, the lineage and a governance council, and left the platform alone. Reconciliation labour fell from $2.1M a year to under $180K.

Source: Thinklytics case library, the 11 semantic layer and metric governance engagements and the 3 master data management engagements.

A written definition per disputed metric, with a named owner. In prose a non-technical reader can check, answering five questions: on what date does this count, one row per what, what is in and out of scope, what gets subtracted and when, and at what exchange rate. The owner is a person, not a committee and not a function.

A reconciliation bridge from each old number to the certified one. Covered below. This is the deliverable that ends the argument, and it is the one most often missing.

Traced lineage for every number in scope. Source table to transformation to reporting surface, readable by someone who cannot read the code.

One enforcement point where the calculation lives. A metric layer that the reporting tools query, so the definition cannot be quietly re-implemented differently in the next dashboard somebody builds. Without this the register becomes a document and the problem returns within two quarters.

A standing forum with the authority to settle the next dispute. The definitions will be challenged again, and they should be. What matters is that there is a named body that decides rather than the question reverting to whoever argues hardest in the meeting.

A change log on the definitions themselves. You need it the first time someone asks why last quarter's figure moved.

The reconciliation bridge is the one that ends the argument

What a reconciliation bridge contains

One row per rule difference, walking from the number a team has been defending to the certified figure. This is the artifact that ends the argument.

Ask for a worked example of this on one real metric before signing. A firm that cannot produce the bridge shape has not done the work before.

Source: Thinklytics engagement pattern across the 11 semantic layer and metric governance engagements in the case library.

A certified number handed down without a bridge is a fourth opinion with a consultant's name attached. The people who have been defending the old figures have no way to check it, so they keep their own version running in a spreadsheet, and within two cycles you are back to two numbers.

The bridge is one row per rule difference, walking from the figure a team reports today to the certified one. Timing reclassification. Grain correction. Scope in and out. Adjustments. Currency restatement. Each row isolates one rule, shows the amount it moves, and says where that rule now lives.

What it buys is consent. A revenue leader who can see that $900K of the gap was deals moving period on an agreed close-date rule, and $500K was services revenue the other team had excluded, can agree to the certified figure without conceding that their own number was wrong. It never was. The rule was just never chosen.

Ask for a worked example of a bridge on one real metric before you sign. A firm that cannot produce the shape of one has not done this before.

What the timeline looks like

In our case library the engagements that resolved disputed metrics on the platform already in place ran 8 to 20 weeks, and the driver was the number of contested metrics rather than the size of the company.

Five competing ARR definitions at a growth-stage SaaS platform took 8 weeks, and resolved a $2.1M discrepancy the board had already noticed. See five ARR definitions reduced to one certified figure. Fourteen regional definitions of a patient encounter at Kaiser Permanente took 11 weeks, after two internal attempts had failed. See the Kaiser Permanente metric governance engagement. Six revenue metrics across four reporting surfaces at an enterprise SaaS company took 14 weeks and closed a $1.4M gap on a $22M base. Forty-two metrics across eight administrative units at a public university took 20.

If a proposal quotes you a duration without first counting the contested metrics by name, the number is a guess. Counting them is a conversation, not a discovery phase, and it should happen before the statement of work is signed.

Acceptance criteria to write into the contract

Acceptance criteria to put in the contract

Written before the work starts, tested on the last day. Each one is observable by someone who was not in the project.

• Every reporting surface in scope returns the same figure for the same period. Name the surfaces in the statement of work. Enterprise SaaS engagement: four surfaces on one number.
• The gap between the highest and lowest version is quantified and explained. $2.1M on a growth-stage SaaS platform, $1.4M on a $22M base at an enterprise SaaS company. The number itself is the proof the bridge works.
• A named person can state the definition and source without opening a spreadsheet. Test it by asking them in front of the project sponsor.
• Automated tests run on every data refresh and fail loudly. A certified number that silently drifts is worse than a contested one, because nobody is checking it any more.
• Reporting kept running throughout, with no blackout period. Kaiser's clinical reporting stayed fully operational across all 11 weeks. Ask for this explicitly.
• The residual annual cost of running the governance is stated. Kaiser's was $180K against $2.1M eliminated. A proposal with no ongoing cost is hiding one.
• A date by which the next disputed metric gets heard. Proves the forum exists rather than appearing in a slide.

Four of these are checkable on the final day by a sponsor who never attended a working session. That is the point of writing them down first.

Source: Thinklytics case library, delivery outcomes as published per engagement.

Write these before the work starts and test them on the last day. The useful property of all seven is that a sponsor who never attended a working session can check them.

The two that get left out most often are the residual annual cost and the date for the next dispute. Both are uncomfortable to put in writing, which is exactly why they belong there. A proposal with no ongoing cost is hiding one, and a forum with no first meeting scheduled is a slide.

What should not be in scope

Three things get added to these engagements because they are easier to scope than the disagreement, and all three push delivery past the point where the sponsor's attention runs out.

Replacing the platform. Almost never required to settle a definition dispute. Kaiser's certified layer was built on the enterprise data warehouse already in place, and clinical reporting ran without interruption across all 11 weeks. If a platform rebuild is being proposed as part of the reconciliation, ask which of the four pipeline conditions in fix your KPI definitions or rebuild your data pipelines has been demonstrated.

Rebuilding reports that are not in dispute. A report nobody is arguing about is working.

Defining metrics nobody is arguing about. Tempting, because it looks like completeness, and it is where metric registry projects go to die. Settle the contested ones, prove the forum works, then extend.

What to ask before signing

Five questions, and the answers tell you more than the methodology section.

Show me a reconciliation bridge from a previous engagement, on one real metric. Which of my contested metrics are in scope by name, and which are explicitly out. Who on my side has to be in the room, and for how many hours a week. What is the residual annual cost after you leave, and what does it buy. And what happens if two of my teams refuse to agree on a definition.

The last one matters most. That is a governance decision rather than a technical one, and a firm that has done this work before will have an answer involving escalation to a named decision maker rather than a workshop.

What we would do first

Before the procurement conversation, count the contested metrics. Write each one on a line with the number of versions in circulation and the name of the most senior person who has complained about it. That list is the scope, and it is also what tells you whether you need this engagement or the broader standing capability in data governance consulting: the first 90 days.

Then put a number on it. The reporting improvement business case is the worksheet we use with clients to size the current reconciliation labour against the residual running cost, which is the comparison a finance committee will ask for.

The delivery capability sits in semantic layer engineering for the enforcement point, data governance consulting for the register and forum, master data management when the dispute is about which records refer to the same entity, data cleaning and preparation when the source needs work first, and SAP data quality and governance on SAP estates. The full set of work in this area sits under we cannot trust the numbers.

Frequently asked questions

What should a reporting reconciliation engagement deliver?

Six artifacts. A written definition per disputed metric with a named owner. A reconciliation bridge from each old number to the certified one, line by line. Traced lineage for every number in scope. One enforcement point where the calculation lives, which the reporting tools query. A standing forum with the authority to settle the next dispute. And a change log on the definitions themselves. A new dashboard is not on that list. A third version of a contested metric is the problem rather than the resolution.

What is a reconciliation bridge?

A line-by-line walk from the number a team has been defending to the certified figure, with one row per rule difference: timing reclassification, grain correction, scope in and out, adjustments, and currency restatement. It matters more than the certified number itself, because it is the artifact that lets the person who defended the old figure see exactly which rule moved their number and by how much. Without it, the certified figure is just a fourth opinion with a consultant's name on it. Ask for a worked example on one real metric before signing.

How long should it take?

In our case library the engagements that resolved disputed metrics on the platform already in place ran 8 to 20 weeks, and the driver was the number of contested metrics rather than company size. Five ARR definitions took 8 weeks. Fourteen regional definitions of a patient encounter took 11. Six revenue metrics across four reporting surfaces took 14. Forty-two metrics across eight administrative units took 20. If a proposal quotes a duration without first counting the contested metrics, the number is a guess.

What acceptance criteria should go in the contract?

Seven, all observable by a sponsor who never attended a working session. Every named reporting surface returns the same figure for the same period. The gap between highest and lowest version is quantified and explained. A named person can state the definition and source without opening a spreadsheet. Automated tests run on every refresh and fail loudly. Reporting kept running throughout with no blackout. The residual annual governance cost is stated. And there is a date by which the next disputed metric gets heard.

What does it cost to keep running afterwards?

There is always a residual, and a proposal that shows none is hiding it. At Kaiser Permanente the reconciliation labour attached to 14 regional definitions fell from $2.1M a year to under $180K, so roughly 9% of the eliminated cost stayed behind as the ongoing cost of running the governance. Ask for that figure explicitly and for what it buys: the forum, the test suite, the change log and the person who owns the register.

What should not be in scope?

Replacing the platform, rebuilding reports that are not in dispute, and defining metrics nobody is arguing about. All three get added because they are easier to scope than the disagreement, and all three push the delivery date past the point where the sponsor's patience runs out. Kaiser's engagement kept the existing enterprise data warehouse and clinical reporting ran without interruption across all 11 weeks. Scope creep here is not a budget problem, it is the main reason these engagements fail.

How is this different from a data governance engagement?

Scope and trigger. A reconciliation engagement starts from a specific disputed number that someone senior has already noticed, and it finishes when that number is settled and enforced. A governance engagement builds the standing capability: the policies, the council, the catalogue, the roles. The reconciliation work is the better first purchase when there is an active argument, because it produces a resolved figure the organisation can see, which is what funds the broader programme.

What should I ask before signing?

Five questions. Show me a reconciliation bridge from a previous engagement on one real metric. Which of my contested metrics are in scope by name, and which are not. Who on my side has to be in the room and for how many hours. What is the residual annual cost after you leave. And what happens if two of my teams refuse to agree on a definition, because that is a governance decision rather than a technical one and the answer tells you whether the firm has done this before.

The work behind this

Eleven engagements in the case library carry semantic layer and metric governance, three carry master data management, and 18 carry governance, privacy and security. Each one states the duration, what was delivered and the measured outcome, including the residual running cost where there is one.

Semantic layer and metric governance, 11 engagements.