Guide · Portfolio measurement governance

Definition drift in portfolio reporting

By Derwin Lucas · · Revised

Definition drift in portfolio reporting is what happens when the meaning of a measure changes over time while its name stays the same. The figure carries on under the old label, and the trend line then shows a change that isn’t in the business, or hides one that is.

Drift is harder to see than a difference between brands, because there is no second brand to compare against; the evidence is the history of the definition itself. This guide follows one illustrative brand through thirteen months, first as drift usually unfolds and then as it could have been handled. All figures are illustrative.

01The exhibit

One brand and its portfolio, thirteen months, as reported

Exhibit 1. One brand and its portfolio, thirteen months, as reported

Illustrative: cases from this site applied to one brand.

MonthWhat changedWhat the report showed
1Nothing. A job counts as installed when it meets the portfolio’s completion criteria.Installed revenue for the month, on those criteria
4The brand moves to a new field-service system, which marks a job installed when the crew closes the work order.$630,000 more installed revenue: 35 jobs at $18,000 counted before they met the completion criteria
5Nothing operational.The operating review credits the brand with a strong production month
9The platform begins allocating shared-service costs to brands.Brand EBITDA $1,350,000 on a business that reported $1,500,000 before the allocation
10Three brands’ submissions have not arrived.A portfolio total covering 22 of 25 brands, presented beside last month’s total covering 25
13Nothing operational.Year-over-year installed revenue and EBITDA each compare two different calculations, against a plan built on month 1’s definitions

Nothing in the brand’s performance changed in any of these months. Every change in what the report showed came from a definition or a scope.

02Month 4

Month 4: the change nobody recorded as a change

The system migration was an operational project. Its plan covered data transfer, crew training and a go-live date. It did not cover the fact that the new system would mark a job installed at a different point: when the crew closes the work order, which can come before the job meets the portfolio’s completion criteria. The status was still called "installed," the report still said "installed revenue," and the figure went to the platform as before. The 35 jobs moved into an earlier operating period; that is a question of which month the work belongs to, not by itself a revenue-recognition error. See the installed revenue case

That is how drift typically arrives: with an ordinary change that no one records as a change of definition. The same thing happens when a new controller books commissions or financing fees in a different place, when a brand changes its cancellation window, when an add-on joins the portfolio reporting under the definitions it brought with it, or when a reporting template gains a column that brands fill in different ways. None of these is a mistake. Each becomes a problem only when the figure carries on as if nothing changed.

Exhibit 2. The same thing in a portfolio: one brand moves commissions

Gross margin, sum of brands · %, July 2025 to June 2026 · reported against common basis · axis from 28%

28%30%32%34%36%38%JulAugSepOctNovDecJanFebMarAprMayJun20252026March: one brand moves commissionsReported 36.9%Common 36.3%

ReportedCommon basisOpen month: not every brand reported

Produced in Goverisyn on a 30-brand demonstration portfolio with synthetic data. From March one brand books sales commissions as operating expenses instead of cost of goods; by June the portfolio’s reported margin is 0.6 points above the common basis.

03Month 5

Month 5: drift read as performance

A $630,000 rise in a brand’s installed revenue is the kind of movement an operating review notices, and the natural reading is that the brand had a strong production month. Nobody asks why. The real cause sits in a system configuration that neither the brand nor the platform is looking at, and the 35 jobs it moved forward still have to meet the completion criteria.

Exhibit 3. Installed revenue, thirteen months

One illustrative brand · $ thousands · as reported against on the portfolio’s completion criteria

01,0002,0003,00012345678910111213MonthMonth 4: new field-service system+630: 35 jobs counted at work-order close,before they meet the completion criteriaOn the criteriaand as reported

On the completion criteriaAs reported

Illustrative. The two lines are the same in every month but month 4, when the new system counts 35 jobs ($630,000) at work-order close, before they meet the criteria.

Drift that flatters a figure rarely prompts a question, and it can hide a real decline: a brand whose production is genuinely slowing can report a steady figure if its definition becomes more generous in the same month. Drift in the other direction makes a brand look weaker than it is, and sends its leadership looking for an operating cause that isn’t there.

04Two kinds

Drift and other portfolio reporting inconsistencies

Drift and inconsistency between brands are both portfolio reporting inconsistencies, and both make a number mean less than its label suggests. Inconsistency is two brands measuring differently in the same month; it distorts comparisons between brands. Drift is one brand measuring differently over time; it distorts comparisons within a brand: growth rates, year-over-year changes, trailing twelve months and anything measured against plan.

A portfolio can solve the first and still suffer from the second. In month 1 of the exhibit, every brand may have agreed what counts as an installed job. A definition agreed once does not stay agreed on its own.

05Months 9–10

Months 9 and 10: the portfolio drifts too

Drift is not only something brands do. In month 9 the platform itself begins allocating shared-service costs, and a brand reporting EBITDA of $1,500,000 now reports $1,350,000: 10% lower, with nothing changed in the business. The decision was a sound one, and it may be the moment brand EBITDA became comparable across the portfolio. It is still a definition change, and every brand’s trend line breaks at it. See the brand EBITDA case

In month 10 the change is in scope rather than definition. A monthly total that covers 25 brands one month and 22 of 25 the next falls for reasons that have nothing to do with performance. Unless each total states which brands it includes, the reader cannot tell a weaker month from a smaller one. See the reporting coverage case

06Month 13

Month 13: where drift does the damage

Drift does its damage wherever a figure is measured against a fixed reference, and by month 13 the brand meets all of them.

  • Prior year. Year-over-year installed revenue compares jobs counted at completion with jobs counted at work-order close. Year-over-year EBITDA compares a figure without the allocation with one that carries it.
  • Plan. A budget built on month 1’s definitions can be missed by definition changes alone.
  • Covenants. A covenant test that holds its threshold steady while the measure underneath it changes can pass or fail for the wrong reason.
  • Diligence. A trailing twelve months that spans month 4 or month 9 mixes two measures in one figure, and a buyer’s quality-of-earnings team is likely to find the seam.

Exhibit 4. One trailing twelve months, three definitions

One illustrative brand · months 1 to 13 · what each month’s figure was built on

12345678910111213Installed revenuecriteriawork-order closeBrand EBITDAcriteriawork-order closeclose + allocationTrailing twelve months at month 13: months 2 to 13Year over year: month 13 against month 1, across both changesEBITDA: 3 basesRevenue: 2

Illustrative. Month 4: installed counted at work-order close instead of on the completion criteria. Month 9: shared-service costs allocated to brands, $1,500,000 of EBITDA becoming $1,350,000.

The principle

A trend is only as stable as the definition underneath it.

07Governed

The same thirteen months, with the definitions recorded

Exhibit 5. The same brand, with each change recorded

MonthWhat the record shows
1Installed defined by the portfolio’s completion criteria, effective month 1, approved by the CFO
4The system change flagged in advance as a likely definition change. The completion criteria stay; the 35 jobs closed in the new system but not yet complete, $630,000, are reported separately until they meet them
5The operating review sees installed revenue on the same criteria as month 1, with the $630,000 shown as an exception rather than a gain
9Shared-service allocation recorded as a portfolio definition change, effective month 9, with EBITDA shown on both bases for the months around it
10The total stated as covering 22 of 25 brands, with the three missing brands named
13Year-over-year figures compare like with like, or say plainly where they cannot

None of these entries required data the brand did not already hold. Each required a record made at the moment of change.

08Detection

How to detect definition drift

Drift is caught by recording definitions, not by watching the numbers. Four practices do most of the work.

Record each definition with the date it took effect. For each measure the board and sponsor rely on, write down how each brand calculates it and when that calculation began. A definition without a start date cannot tell you whether last year’s figure was built the same way as this year’s.

Treat a change in a submission as a question. A new system, a new column, a new person submitting, or a rate that arrives without the counts behind it are all reasons to ask what changed before accepting the figure. In the exhibit, the system change was planned before it went live; that was the moment to ask. Why a template alone doesn’t hold a definition

Reconcile to source records every period. A reconciliation that tied out last month and doesn’t this month can be the first sign of drift. Figures traced to the records behind them each period carry their own history; figures submitted only as totals don’t.

Screen for unexplained step changes. A jump at a month boundary that lines up with a system, staffing or policy change is a reason to investigate, not proof of drift. Only the definition’s history can confirm it.

09Handling

How to handle a definition change

Give the new definition an effective date, and keep earlier figures under the definition they were built on. Where a brand’s records allow, restate prior periods under the new definition alongside the originals, so the trend compares like with like. Where they don’t, show the series as broken at the change rather than splicing two definitions into one line. Missing history is stated as missing, never estimated.

Exhibit 6. Three ways to show a definition change

The same series, with a change at the dotted line

Restated alongside

restatednew

Prior periods shown on the new definition next to the originals. The trend compares like with like.

Broken at the change

series broken

Where the records don’t allow a restatement. The gap is stated, never filled.

Spliced: not used

step reads as performance

Two definitions joined in one line. The step reads as a change in the business.

Illustrative.

Decide who may approve a change before one arrives. The CFO should approve changes to the measures the board and sponsor rely on, brand controllers should raise a change before it takes effect, and FP&A should record it.

Figures that have already gone to a board or lender need the same care. In Goverisyn, each issued board pack is versioned and lists what changed since the previous version, so a reader can see when a figure’s basis moved. What an issued board pack records

10Where to start

Start with the measure the board watches

Pick the one measure your board watches most closely, and write down how each brand calculates it today and since when. That record alone shows where the trend can be trusted and where it can’t, and it is the month 1 entry that makes every later change visible.

Goverisyn’s Portfolio Measurement Review builds that record across selected brands, after a free 45-minute Portfolio Review that needs no data in advance. For differences between brands rather than over time, see How to compare performance across acquired companies and EBITDA comparability across portfolio companies.

Start with a Portfolio Review.

45 minutes. No data needed in advance.

By Derwin Lucas, founder of Goverisyn, with a career in management and information systems in home services. Published . Revised . About Goverisyn