Decide on real numbers

Most commercial decisions in mid-sized B2B are made on an export somebody built once and now maintains by hand. The numbers are stale, they disagree between departments, and nobody entirely trusts them, so the meeting reverts to instinct and the spreadsheet is used to justify the decision afterwards. The goal is that two people asking the same question get the same answer, quickly enough to act on it.

Why this is harder than it sounds

The first surprise is that the disagreeing numbers are usually both correct. Ask three departments for last month's revenue and you get three figures because you asked three different questions. Was it booked or invoiced? Net of credit notes? Does it include freight and packaging? Is it recognised at order date or at delivery date? Are intercompany lines in or out? Each answer is defensible. None of them is wrong. A reporting tool cannot resolve this, because the conflict sits in the definitions rather than in the data, and definitions are a management decision.

The second is that your online data describes a biased slice of your business. If most orders still arrive by phone and email, shop analytics describe the customers who already converted, which is a small and unrepresentative part of your customer list. Drawing conclusions about assortment or pricing from that slice is how companies end up optimising for the segment they already serve well. Getting offline orders into the same system is a precondition for this goal.

Third, B2B numbers are small numbers. If twelve people at four customers place every order in a category, conversion rate is noise, bounce rate is meaningless, and a two-week A/B test will never reach significance. B2B reporting works on accounts, cohorts and absolute counts rather than on the traffic metrics a B2C analytics course teaches.

And there is a human obstacle worth naming. The person who maintains the master spreadsheet has real standing attached to it: they are the one who knows where the numbers come from. Replacing that file with a dashboard removes a role someone has held for years. Involve them in defining the metrics and they become the best ally the project has. Go around them and the spreadsheet survives, in parallel, forever.

The levers

Get the transactional data in one place

What good looks like: orders from every channel (shop, phone, email, punchout) end up as orders in the same system, with the same customer and the same articles attached, so that a question about a customer has one place to be asked.

What it depends on: deciding which system owns which entity, and then connecting it. This is the unglamorous prerequisite for the entire goal.

1. Deciding your system of record
2. Integration patterns
3. Connect a system
4. Reconcile with your ERP
5. Monitor your syncs

Agree the definitions before you build anything

What good looks like: a short written list (one page is enough) naming each metric you will manage on, its exact definition, its source, and the person who owns it. "Revenue" appears once, not four times.

What it depends on: somebody with authority sitting in a room and deciding. Expect this to take longer than configuring the reports and to be worth more.

1. The metrics that matter in B2B
2. What counts as an order
3. Currency and time
4. Trusting your numbers
5. Customer data hygiene

Use the standard reporting before building your own

What good looks like: the recurring questions (which customers are ordering less than expected, what is growing, where are quotes stalling) are answered by the platform's standard reports and exports, and by Analytics Studio's dashboards once it is on your tenant. Custom work is reserved for the handful of questions specific to your business.

What it depends on: resisting the urge to rebuild your existing spreadsheet layout. The spreadsheet's shape encodes the limitations of how it was assembled rather than what you need to know.

1. Dashboards and KPIs
2. Reports and exports
3. Build a custom dashboard

Get the numbers to the people who act on them

A number nobody sees on Monday morning changes nothing. The last step is distribution: the right view, in front of the right person, without them asking.

What good looks like: each sales representative sees their own accounts without needing an export; the commercial lead gets the weekly figure without requesting it; and the alerts that matter arrive rather than being looked for.

What it depends on: roles and permissions being modelled, so that "everyone sees everything" is not the only workable setting.

1. Users, roles and permissions
2. Notifications
3. Reports and exports

A sensible order

Phase 1: one source and one definition sheet. Connect the transactional data and write down what each metric means. Do not build a single report yet. The deliverable of this phase is agreement, and its value is that arguments about whose number is right stop. It depends on the integration work, which is why it takes longer than anyone expects and why starting it late delays everything else.

Phase 2: the standard reports, with owners. Set up the reports and exports for the three or four questions you manage on, assign each one a person, and use them in the meetings where decisions are made. If a report is not used in a meeting, delete it. This phase depends on Phase 1: reports built on undefined metrics relocate the disagreement.

Phase 3: the derived questions. Cohorts, order-frequency decay as a churn signal, assortment gaps from zero-result searches, quote conversion by segment. These are the questions that make the investment pay, and they are only answerable once a full cycle of clean history exists. Attempting them in month one produces confident conclusions from three weeks of biased data.

Common mistakes

• Buying a reporting tool to fix a definitions problem. The tool will faithfully produce five disagreeing numbers, faster.
• Building sixty KPIs. A dashboard with sixty numbers is read by nobody. Four numbers with owners beat sixty without.
• Treating shop data as business data while most orders still arrive offline. The conclusions will be about the wrong customers.
• Importing B2C metrics. Conversion rate, bounce rate and session duration describe anonymous consumer traffic. Your buyers are twelve named people with purchase orders.
• Letting the old exports run in parallel "only for a transition period". They never stop, and as long as they exist the new numbers are optional.

How you know it is working

• The number of places an authoritative revenue figure can come from. The target is one. Count it honestly; most companies discover four or five.
• Time from question to answer. Time a real question ("how did customer Meyer GmbH develop against last year") before you start and again a quarter later. Minutes versus days is the whole point of this goal.
• Share of all orders, across every channel, that exist as structured orders in the platform. Anything short of all of them is a known blind spot; know its size.
• Manually maintained recurring spreadsheets still in circulation, counted by name. This should fall, and if it does not, Phase 2 has not replaced anything.
• Named metrics with a written definition and an owner. A metric with neither is a rumour.
• The reconciliation gap between the platform's revenue figure and finance's, for the same period. It will not be zero (timing and scope differ), but it must be explainable and stable. An unexplained gap that changes every month means the pipeline is broken rather than the definition.

Related goals

• Cut manual work: the same integration work pays twice.
• Raise order value: the first commercial decision worth making on real numbers.
• Consolidate systems: fewer systems is the structural fix for disagreeing numbers.

What you're building

Every order, from every channel, as a structured order against a clean customer record in one place, with a one-page definition sheet that says what each number means and who owns it. Reports and exports feed the people who decide, and your BI tool if you keep one. Analytics Studio's dashboards sit on the same data once they are available, so nothing you define now has to be redefined.

flowchart LR
  ERP["Your ERP: phone and email orders"]
  BI["Your BI tool, if you keep one"]
  subgraph XS["Experience Studio"]
    SF["Storefront orders"]
  end
  subgraph IS["Integration Studio"]
    WF["Workflows"]
    EXP["Exports"]
  end
  subgraph CS["Commerce Studio"]
    O["Orders: all channels"]
    C["Customers: organizations, segments"]
    RPT["Reports"]
  end
  AS["Analytics Studio dashboards (coming soon)"]:::planned
  SF -- "shop and punchout orders" --> O
  ERP -- "offline orders, nightly" --> WF --> O
  O --> C
  O --> RPT
  C --> RPT
  O --> EXP --> BI
  O -.-> AS
  C -.-> AS
  classDef planned stroke-dasharray: 5 5

What you need

• The system of record for orders and customers settled and written down — Deciding your system of record
• Offline orders flowing from the ERP into the platform as structured orders — Connect your ERP
• A customer master without duplicates, so one company is one organisation — Customer data hygiene
• A one-page definition sheet: each metric, its definition, its source, its owner — What counts as an order
• Rules for currency conversion, time zones and periods — Currency and time
• The spreadsheet owner in the room from the first meeting — The metrics that matter in B2B
• Roles that decide who sees which accounts — Users, roles and permissions
• A baseline: one real question, timed from asking to answering — Trusting your numbers

Build it

1. Settle which system owns orders and customers — Deciding your system of record
2. Bring every offline order from the ERP into the platform as a structured order — Connect a system
3. Reconcile one month of orders against the ERP and explain every difference — Reconcile with your ERP
4. Merge duplicate organisations so a customer is one record — Customer data hygiene
5. Write the definition sheet: what counts as an order, which revenue, which date — What counts as an order
6. Fix the currency and period rules once — Currency and time
7. Time one real question, end to end, as the baseline — Trusting your numbers
8. Set up the reports and exports for the three or four questions you manage on, one owner each — Reports and exports
9. Feed your BI tool from platform exports if you keep it — Bulk exports
10. Give sales representatives a view of their own accounts through roles — Users, roles and permissions
11. Route the alerts that matter to the people who act — Notifications
12. Retire the parallel spreadsheets by name, with a date — Dashboards and KPIs
13. After a full cycle of clean history, add cohorts and frequency-decay as a churn signal — Build a custom dashboard

Decisions you'll need to make

Related product areas

• Commerce Studio: Orders · Organizations · Segments · Contacts
• Integration Studio: Workflows · Runs · Import/export
• Analytics Studio (coming soon): Analytics Studio