River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Multi Location Operations Template

Four documents and four sheets that separate the differences you approved from the ones nobody asked about, then rank your sites on what is left.

Free download  ·  No account needed

How Compliance Is Scored

Wexford, 14 locations, 40 standards

One audit, three rankings. 560 possible cells, 54 out of scope, 506 scored: 388 met, 76 approved variance, 32 unapproved deviation, 10 not met.

#By naive rateBy adjusted rateBy real findings
1Ironbridge92.3%Alderfen100.0%Alderfen0
2Pinehurst91.9%Glasswater100.0%Glasswater0
3Quarrygate90.0%Pinehurst97.3%Cobbsfield1
      
12Vanhorn65.0%Foxhollow90.0%Milbrook10
13Trellick53.1%Milbrook75.0%  
14Alderfen50.0%Vanhorn70.0%Vanhorn12

All 14 positions differ between the first ranking and the third

Alderfen is an airport concession, 620 sq ft. Last in the estate at 50.0%. Its 32 scored cells are 16 met and 16 approved variances, every one requested, reasoned and signed. Zero findings. First on both other measures.

Vanhorn is freestanding, 3,100 sq ft, open since 2015. Mid-table at 65.0%, which is 15.0 points above Alderfen. Ten unapproved deviations and two not met, the worst in the estate.

A conditional-format scorecard puts the best-run manager under review and leaves the worst alone.

A compliance percentage per site is the standard artifact here, and it is the reason standards documents get ignored across an estate. It scores an airport concession with a 540 square foot footprint and a landlord who controls the music identically to a freestanding site that quietly stopped holding performance conversations. Managers can tell those apart even when the scorecard cannot, so they decide the scorecard is wrong and then discount the document behind it.

So every location-standard cell here resolves to one of four states. Two of them are the ones a normal scorecard cannot separate: approved variance, a difference this company agreed to in writing, and unapproved deviation, where the site never asked. On the worked estate, a 14 location chain called Wexford with 40 standards, that produces three rankings off one audit, and all 14 positions differ between the first and the third.

The site that comes last at 50.0% has 16 approved variances and zero findings. The site sitting twelfth at 65.0% has twelve. Then read the register down the standard column: four standards are varied at over half their sites, and rescoping those moves the headline rate to 82.2% while changing nothing on the ground. The same read voids two variances granted against a notice period that is statutory in that state. Send River the audit results, or take the files blank. The standards belong in an SOP library, and the rate is a defined metric.

Read the register down the standard column, not across the location column

The register with a reason class and a substitute on every row, the memo that turns it into a standards defect, and four quarters of trend showing which number moved.

Variance Register

Seven of 76 rows. The substitute column is what turns a variance from permission into an obligation, and it is the column most exception processes leave blank.

RefSiteStdWhat the site does insteadReasonReviewValid
032GlasswaterS25Rota at 7 days with the local notice ruleregulatorypermanentVOID
042RedhavenS25Rota at 7 days with the local notice ruleregulatorypermanentVOID
055AlderfenS22One supervisor on shift plus phone escalationphysicallease breakyes
021FoxhollowS12Oil changed daily rather than filteredtemporary15 Apr 2026yes
009MilbrookS06NOT RECORDEDgrandfatheredpermanentyes
049VanhornS40NOT RECORDEDgrandfatheredpermanentyes
061TrellickS38Landlord cleaning schedule, copy held on siteleasepermanentyes

Two rows are void, not lenient. Both are variances against a 14 day advance rota, at the two sites in a state where that notice period is statutory for chains in this industry. Approved in good faith by the People Director and filed correctly. Withdrawal is immediate rather than queued, and the wider fix is the column: every standard needs a tag for the jurisdictions where it restates a legal floor.

Seven rows are grandfathered, meaning on the register, approved at some point, with no reason and no named approver. Nobody can say whether the constraint still exists. That is the cheapest thing here to fix: ask the site, then write the reason down or withdraw the variance.

Four of the 76 variances carry a date. The other 72 are marked permanent, including several plainly temporary ones. A register where almost nothing expires is a register nobody re-reads.

Standards Review Memo

Per standard, the share of its applicable sites holding a variance against it. Over half is a defect in the standards document.

StdRequirementVariedRateFormats
S02A-frame pavement sign at entrance9/1464%inline, mall, airport
S38External area litter-picked twice daily9/1464%inline, mall, airport
S06Music from the corporate stream8/1457%all four
S39Customer toilets checked hourly8/1457%all four
S22Two supervisors at all trading hours5/1436%mall, airport
S27Annual performance conversation0 varied71% devevery format

The first four have one cause. They were written when Wexford was five freestanding units with their own lot, frontage and facilities. It then grew into four inline units, three mall units and two airport concessions, and nobody rescoped the document. A site with no external area cannot litter-pick one twice daily, and it has been marked non-compliant for that every quarter since it opened.

NaiveAdjustedCellsVariancesFindings
Before76.7%91.7%5067642
After rescope82.2%91.1%4724242

Findings are 42 either side, and that is the control. A rescope only removes cells already agreed in writing to be excused, so it moves the rate and cannot move the findings. Every freestanding site moves under two points; every mall and airport site moves seven or more, which is the tell that the old rate was measuring format.

S27 is the opposite pattern and must not be confused with it. Deviated from at 10 of 14 sites with zero variance requests. Nobody claims it is impossible, they are just not doing it. That needs a decision about whether the standard survives, not a rescope.

Compliance Trend

Estate level, four quarters. Prior quarters were scored on the naive basis only, because that is all the estate was measuring.

Q2 2025Q3 2025Q4 2025Q1 2026
Naive rate74.7%76.5%76.7%76.7%
Real findings65564542

The naive rate moved 2.0 points across three quarters and looks like a plateau. The findings count fell by 23, which is 35% of them.

That is not a reporting quibble. The estate has been improving steadily for a year and the number the board sees has been flat throughout, because most of what that number measures is which format a site trades in rather than how it is run. Anybody reading only the naive rate would conclude the standards programme had stopped working and would go looking for a new one.

Per site, the trend column says which of two things happened. Findings fell while the rate held still, meaning the rate could not see the improvement. Or findings rose while the rate held still, meaning the rate hid the drift. Both appear in the estate, and neither is visible on a single percentage.

The rescope column on this sheet restates each site's rate with the four defective standards scoped correctly, so next quarter's comparison is against a measure that means something.

What's in the pack

01

Standard by Location

One row per standard, one column per site, and four possible states per cell rather than a pass or a fail.

02

Variance Register

Every approved difference with its reason class, its named approver, its review trigger and what the site does instead.

03

Exception Approval Log

Kept deliberately separate from the register: the register is what is true now, the log is who decided it and when.

04

Compliance Trend

Four quarters per site on both bases, with a column stating whether the rate could see the change in findings or hid it.

05

Standards Policy

What a standard carries here, including the scope rule and the jurisdiction tag that most standards documents omit.

06

How Compliance Is Scored

The four states, the three rankings, and the two named sites where the naive rate and the findings count disagree completely.

07

Standards Review Memo

The register read down the standard column, with the rescope costed and the findings count shown as the control.

08

Approved Variance Note

A worked note for one variance, including the substitute, the evidence, the review trigger and what is explicitly not varied.

How to use it

  1. 1

    Open in River, or take it blank

    Open the pack in River and hand it the standards list, the audit results and your site attributes, or download the Word documents and CSV sheets and work through them yourself.

  2. 2

    Give every standard a scope rule first

    Applies everywhere, only where there is a kitchen, only above a stated floor area. Scope is decided once as a property of the standard, never argued during an audit.

  3. 3

    Score each cell into one of four states

    The question that separates an approved variance from an unapproved deviation is whether the site ever raised it with anybody. That is usually one sentence per cell.

  4. 4

    Rank three ways, then invert the register

    Naive, adjusted and a plain findings count, never combined. Then read down the standard column to find the standards that are wrong rather than the sites.

Frequently asked questions

Is this template free?

Yes, and nothing is gated. The zip holds four Word documents and four CSV sheets, no account and no card. Edit with AI is the optional half: River scores the grid, computes the three rankings and reads the variance register in both directions. More packs sit in the template library.

What format are the downloaded files?

Four Word documents and four CSV sheets, zipped. Excel, Numbers and Google Sheets open the sheets directly, and the policy, scoring note, variance note and review memo open in Word or Pages. The grid sheet is one row per standard and one column per location, so it stays readable at fourteen sites and scales past that.

Why not just report one compliance percentage per site?

Because most of what it measures is which format a site trades in. On the worked estate the naive rate moved two points across three quarters while the findings count fell by 23, so anybody reading only the rate would conclude the programme had stopped working. Three columns travel together, and a composite of them is the same mistake.

Is a variance register not just a way to excuse everything?

It would be without the two checks that come with it. A variance needs a reason class, a named approver and what the site does instead, which makes it an obligation rather than permission. And the register gets read down the standard column, where a variance rate over half turns into a defect in the standards document.

How is rescoping a standard different from massaging the number?

The findings count is the control. Rescoping only removes cells the company had already agreed in writing were excused, so it moves the rate and cannot move the findings. On the worked estate the rate goes from 76.7% to 82.2% and the findings stay at 42 either side, which is what makes the change defensible.

What about standards we are not allowed to vary?

That is a column the register needs and usually lacks. An emergency action plan must be in writing unless the employer has ten or fewer employees, and Oregon requires covered chains to post a written schedule at least 14 calendar days ahead. Neither is yours to excuse.

Does this run the site visit or open a new location?

No. This space holds the standards, the grid and the register. Walking the floor and re-checking last visit's closures is a site visit and field audit job, and getting a new site trading is a location opening one. A franchised site is governed by its operations manual instead, and passing information between shifts without losing it is a shift handover log job.

Separate the differences you agreed to from the ones nobody asked about

Take the Word documents and CSV sheets blank, or open this pack in River and let it score the grid and read the register both ways first.

Edit with AI