How Do I Score My Counter Staff Across Branches?

The job a cross-branch counter scorecard is actually hired to do

The job is not "measure sales." Every point-of-sale system already measures sales. The job is to answer a specific, uncomfortable question that multi-branch operators cannot answer honestly today: *if I moved this person from the busy store to the quiet store, would they still be good?*

That question is why raw revenue fails. A counter person at a metro branch doing $1.4M in annual counter revenue looks like a star next to a rural counter person doing $480K. But strip out traffic and the picture can invert. Suppose the metro branch rings 240 transactions a week and the rural branch rings 70. If the metro person's average ticket is $112 with a 19% attach rate, and the rural person's is $164 with a 41% attach rate, the rural person is selling substantially better on every line that reflects skill. They are just standing in a smaller room. Revenue measured the room. The scorecard has to measure the person.

So the job breaks into four concrete deliverables:

Comparability. Every KPI on the matrix must be a rate, a ratio, or a level — not a total. Average ticket instead of total ticket dollars. Attach rate as a percentage of eligible transactions instead of add-on dollars. Units per transaction instead of units. Quote-to-order conversion as a percentage instead of orders won. The moment a raw total lands on the matrix, branch size leaks back into the score and the whole exercise collapses into a traffic report.

Completeness. A counter person's job is wider than selling. They quote, they source substitutes, they check stock, they handle returns, they keep a will-call shelf honest, they keep an account from drifting to the competitor down the road. If the matrix only carries selling lines, staff optimize for selling lines and let accuracy rot. A miskeyed part number on a $60 fitting costs a truck roll, a callback, and a credit memo — call it $180 to $400 of erased margin depending on how far the job site is. Accuracy belongs on the matrix precisely because it is invisible in a revenue report.

Legibility. The scorecard exists to change behavior, and behavior only changes when the person being scored can see the score, understand how it was computed, and identify the single move that raises it most. A composite that arrives once a quarter as a number from corporate with no visible arithmetic behind it is not a scorecard, it is a verdict. Publish the weights. Publish the level definitions. Let a counter person compute their own score on a napkin.

Teeth. A scorecard with no consequence decays within two quarters. The consequence can be pay (a bonus component keyed to composite), advancement (Counter I → II → III bands with composite thresholds), or coaching allocation (the branch manager owes 30 minutes a week to anyone under 3.0). Any of the three works. None of them working is the common failure.

There is an adjacent version of this same job that shows up all over distribution and RevOps: inside sales rep scoring, service advisor scoring at dealerships, parts pro scoring at auto-parts chains, and even dispatcher scoring at HVAC operations. The mechanics are identical — rate-based KPIs, weights, 1–5 levels, one composite — because the underlying problem is identical: a person's output is entangled with the volume of the queue they were handed, and you need to separate the two.

How the scorecard fits the RevOps stack

The scorecard is not a system. It is a layer that sits on top of systems you already run, and its quality is capped by the quality of the data underneath it. Before you weight anything, walk the data path backward from the composite to the source record and confirm each hop actually exists.

The source of truth for counter transactions is almost always the ERP or the distribution POS — Epicor, Infor,"whatever the branch has run since 2009." That system holds the line items, the customer account, the salesperson code on the ticket, and the timestamp. Three things routinely break here. First, the salesperson code is stale: tickets get rung under a shared branch login or under whoever's terminal was open, and 15% of your transactions have no reliable owner. Second, quotes live in a different module than orders, so quote-to-order conversion cannot be computed without a join nobody has built. Third, "attach" has no definition in the data — the system knows what was on the ticket but not what *should* have been on the ticket, so attach rate needs an eligibility rule (e.g., "any ticket containing a pump is eligible for the fitting/seal attach") that you author yourself.

Fix those three before you score anyone. Scoring on bad attribution is worse than not scoring, because it destroys trust in one cycle and you will not get a second launch.

Above the ERP, most operators land in one of three configurations. Configuration one: spreadsheet on export. Someone pulls a monthly extract, a pivot table computes the rates, and a formula rolls the composite. Cheap, transparent, and fragile — it survives exactly as long as the person who built it stays interested. Configuration two: BI layer. The extract lands in a warehouse or a BI tool and the matrix becomes a modeled dashboard. More durable, needs an analyst, and re-weighting takes a ticket instead of an afternoon. Configuration three: CRM-hosted. If counter and quote activity already syncs into a CRM, the scorecard can live next to the orders, and branch managers see the score in the same place they see the pipeline.

The RevOps question underneath all three is ownership. The scorecard touches sales, branch operations, HR, and finance. If nobody owns it, the weights drift, one region quietly runs its own version, and within a year you are back to comparing zip codes. Assign a single owner with authority to change weights, and put the change on a schedule everyone knows about.

Downstream, the composite feeds more than reviews. It feeds hiring profiles (what does a 4.0 look like on day one versus month eighteen?), staffing decisions (a branch whose median composite is 2.4 has a coaching problem, not a headcount problem), and transfer decisions (moving a 4.3 into a struggling branch is a real intervention with a real expected lift). It also feeds vendor conversations — when a manufacturer funds a program, being able to show attach-rate movement by branch is worth actual co-op dollars.

Choosing the KPIs, setting the weights, defining the levels

This is the part people rush, and it is the part that determines whether the scorecard is fair. Three separate decisions live here.

Pick 6–8 KPIs. Not four, not fourteen. Under six and the matrix is too coarse to distinguish a good counter person from a fast one. Over eight and each weight gets so small that no single line is worth chasing, and the whole thing becomes noise. A workable set for a distribution counter:

• Average ticket — total counter revenue ÷ transactions, per person, trailing 90 days.
• Attach / add-on rate — attached-item tickets ÷ eligible tickets, using your written eligibility rule.
• Units per transaction — line count or unit count per ticket; catches basket depth that average ticket can miss on high-price single items.
• Quote-to-order conversion — orders ÷ quotes issued, trailing 90 days. This is the single most skill-revealing line on most counters.
• Accuracy — error tickets (wrong part, wrong quantity, wrong price, avoidable credit memo) ÷ total tickets, inverted so low error = high level.
• Customer satisfaction / service — survey score if you have one, structured manager observation if you don't.
• Account growth — share of the person's assigned or frequently-served accounts showing period-over-period growth.
• Margin discipline (optional eighth) — realized gross margin percentage versus branch target, which catches the seller who buys volume with discounts.

Set weights as a group, in one room, once. Get branch leadership, sales leadership, and one or two respected counter people in the same conversation. Weights must total 100. A common starting distribution for a counter network trying to move from order-taking to selling: average ticket 20, attach rate 20, quote-to-order 15, units per transaction 10, accuracy 15, satisfaction 10, account growth 10. If margin discipline is in, pull five points from average ticket and five from units.

The weights encode strategy, and they should visibly change when strategy changes. If a vendor program lands in Q3 and attach on that line is the priority, moving attach from 20 to 30 for the quarter is a legitimate, transparent, one-afternoon change — and every branch re-aims the next day. That responsiveness is the main advantage a weighted matrix has over a fixed commission plan, which takes a legal review and a quarter's notice to touch.

Define the levels with numbers, not adjectives. This is where most matrices fail. "Level 4 = strong attach performance" is unusable; two managers will score the same person two levels apart. Write it as thresholds:

Set your own bands off your own trailing-12 data, not off numbers from an article. The correct method: pull the distribution of each KPI across all counter staff at all branches, put level 3 at the median, level 2 and 4 roughly a half-standard-deviation out, and levels 1 and 5 at the tails. That guarantees roughly a normal spread on day one and gives you an honest baseline to improve against. If everyone scores a 4 in month one, your bands are too soft and the scorecard will tell you nothing.

One caution on level design: do not let a level be defined by a number the person cannot influence. Average ticket at a branch that serves industrial MRO accounts will run structurally higher than one serving residential contractors. Either band by branch type, or accept the noise and weight that line lower. Fairness is a design constraint, not a nice-to-have — the first time a counter person can point at a line and say "there is no way I could ever hit a 5 on that," you have lost the room.

Rolling it out across branches without losing the room

A technically perfect matrix launched badly still fails. The rollout is a change-management problem, and it has a shape that works.

Run it silent for one full cycle first. Compute composites for 60–90 days before anyone's pay or review touches them. Two things come out of a silent cycle. First, you find the data bugs — the person scoring 1.2 because their tickets are landing under a shared login, the branch whose returns are coded so oddly that accuracy is meaningless. Second, you get the real distribution, which lets you fix bands before anyone has an emotional stake in a number.

Show every person their own card before you show anyone a ranking. The first exposure should be private, individual, and explanatory: here are your seven lines, here is your level on each, here is the arithmetic, here is your composite. A cross-branch leaderboard as the first contact reliably produces the reaction "this is a stack rank to fire people," and the reaction is not irrational — plenty of scorecards have been exactly that.

Name the one move. The most useful output of a scorecard is not the composite, it is the derivative: which single line, if improved one level, raises this person's composite most? That is a mechanical calculation — for each KPI, weight × 1 — filtered to lines where the person is not already at 5. A person at level 2 on attach with a weight of 20 has a 20-point move available; a person at level 4 on satisfaction with a weight of 10 has 10. Tell them the 20. Coaching that says "raise your attach rate from 21% to 27% and your composite moves more than anything else you could do" is actionable in a way that "your composite is 3.1" never is.

Anticipate the four objections, because they always come. *"The busy branch has an unfair advantage"* — it does not, because every line is a rate; walk them through one. *"My customers are different"* — sometimes true, which is why branch-type banding exists on ticket-size lines. *"I do things that aren't on the matrix"* — often true and worth hearing; if it recurs from multiple people, the matrix is missing a line. *"This is just a way to cut people"* — answer it directly and honestly, because whatever you say, your actions in the first two quarters will settle it.

Give it teeth on a published schedule. Quarter one silent. Quarter two visible with coaching attached but no money. Quarter three, the comp or advancement component turns on. Announcing that calendar upfront converts the scariest part of the rollout into something predictable.

Watch for the gaming patterns. Every scored system gets gamed, and knowing the patterns lets you counter them cheaply. Quote-to-order conversion invites people to stop quoting the hard ones — so track quote volume alongside conversion and flag anyone whose quote count drops more than 25% while conversion jumps. Attach rate invites junk attachment: a $2 item added to a $900 ticket to tick the box — so define attach eligibility by category relevance, and spot-check. Average ticket invites talking a customer into more than they need, which shows up two weeks later as returns — which is exactly why accuracy and returns sit on the same matrix. A well-built matrix is partially self-policing because the lines constrain each other.

Recalibrate on a cadence, not on a complaint. Review weights quarterly, review level bands annually. Changing bands mid-cycle because someone is unhappy destroys the credibility of every score already issued. Changing weights quarterly because strategy moved is the system working as designed. Publish both changes in writing with the effective date.

The buyer decision framework: spreadsheet, BI, or a bought platform

Once the method exists, the tooling question is narrow: where do you want the effort to live? There is no configuration that eliminates effort — it only moves between build effort, maintenance effort, and license cost.

Spreadsheet. Free, fully transparent, changeable in an afternoon. The real cost is a recurring monthly chore — pulling the extract, refreshing the rates, distributing the cards — and the well-documented failure mode where the sheet quietly stops being updated after the person who built it changes roles. Under roughly 40 scored people it is genuinely the right answer, and it is always the right *first* answer because building the matrix in a sheet forces you to confront every definitional gap before you spend money.

BI or CRM-hosted. The scorecard becomes a modeled asset that refreshes on its own. Durable, auditable, survives turnover. Costs you an analyst's time to build and a ticket every time weights change, which is a real friction if you intend to re-weight quarterly. Worth it once the manual refresh is a multi-hour monthly job or once regional leaders need the same view without asking anyone.

Bought platform. Three distinct categories, and mixing them up is the most common purchasing mistake. *Scorecard and coaching platforms* automate the matrix and attach coaching cadences to it — closest to the method, and the natural upgrade when visibility is the goal. *Gamification and recognition platforms* run leaderboards, contests, and broadcast displays — they drive short-term behavior well and branch-versus-branch competition genuinely moves attach rates, but they lean toward motivation over rigorous weighting, so they complement a matrix rather than replace it. *Incentive compensation platforms* administer multi-component plans and pay accurately at scale — the right buy when the teeth live in pay, when plan components differ by branch, and when getting commission math wrong is a real liability.

Evaluating any of them, ask five questions and insist on demo answers using your own data shape: Can I define my own weighted KPIs, or am I locked to the vendor's metric set? Can I re-weight without vendor services hours? Does the individual see their own card, or only managers? Does it read from my ERP, or am I uploading a CSV monthly forever? And what does it cost at full rollout — every counter person, every branch — not at the pilot price for 20 seats.

Budget the way you would any RevOps tooling purchase: pilot at one region, insist on a real integration during the pilot rather than a promised one, and hold the annual commitment until the scorecard has survived one full quarter with teeth on. If the vendor will not do a paid pilot, that is information.

The honest sequencing for most operators: build it in a sheet, run it silent for a quarter, fix the data attribution problems you find, put teeth on it in a sheet, and only then buy something — because by that point you know exactly which of the three platform categories you actually need, and the demos stop being persuasive and start being answerable.

Related questions

How often should I recompute the composite?

Monthly for the score, on a trailing 90-day window for each KPI. Monthly keeps it current enough to coach against; the 90-day window smooths out a slow month or a two-week vacation that would otherwise swing a rate wildly.

Should branch managers be scored on the same matrix?

No — score them on a different one built from their branch's aggregate: median composite of their staff, distribution improvement, retention, and coaching-hours delivered. Scoring a manager on their own counter sales pulls them behind the counter instead of behind their people.

What about part-time or new counter staff?

Exclude anyone under 90 days from cross-branch ranking, but compute and show their card from day 30 as a development tool. For part-timers, all rate-based KPIs work fine; set a minimum transaction floor (roughly 100 tickets in the window) below which a rate is too noisy to trust.

Can this replace our commission plan?

It can drive one, but do not bolt a matrix onto pay overnight. Run it visible for two quarters first, confirm the scores are trusted and the data is clean, then wire a bonus component to the composite while leaving base commission mechanics alone.

How do I handle a branch with genuinely different customers?

Band the structurally-different lines by branch type — average ticket and units per transaction usually — while keeping attach, conversion, accuracy, and satisfaction on universal bands. Those four are skill lines and travel across any customer mix.

FAQ

How many KPIs should be on the matrix?

Six to eight. Fewer than six and the composite cannot tell a genuinely good counter person from a fast one, because a single line dominates. More than eight and each weight shrinks to the point that no line is worth chasing, which produces a score people acknowledge and ignore. Six to eight also fits on one printed card, which matters more than it sounds — a scorecard that requires scrolling stops being read.

What if my counter staff push back on being scored?

Pushback almost always traces to fear of unfair comparison rather than objection to measurement. Publish the entire matrix — every KPI, every weight, every level threshold — so anyone can compute their own score. Show individual cards privately before any ranking exists. Explain that the composite reflects the full job, not one number, and that coaching is wired to the same system that scores. When the first cycle runs silent with no consequences, most of the resistance resolves on its own.

Can the same scorecard work for a busy urban branch and a quiet rural one?

Yes, and that is the entire point of using rates rather than totals. Attach rate, quote-to-order conversion, units per transaction, and accuracy are all independent of foot traffic. A rural counter person selling a fuller basket per customer will outscore a metro seller who rings fast and never attaches, even though the metro branch does triple the revenue. Average ticket is the one line that may need branch-type banding.

How do I define attach rate so it is actually measurable?

Write an eligibility rule before you compute anything. Attach rate is attached-item tickets divided by *eligible* tickets, and eligibility is a category relationship you author: a pump ticket is eligible for seals and fittings, a fixture ticket is eligible for supply lines and valves. Without that rule you are dividing by all tickets, which punishes anyone serving customers with nothing sensible to attach.

What is the fastest way to raise a low composite?

Compute the derivative rather than staring at the total: for each KPI where the person is below level 5, the available gain is that KPI's weight. Coach the largest one. A person at level 2 on a 20-weight attach line has twice the upside of a person at level 4 on a 10-weight satisfaction line, and telling them that specific number is far more useful than telling them their composite is 3.1.

How does this connect to the rest of our RevOps reporting?

The composite becomes a person-level dimension you can join to almost everything else — branch performance, account retention, margin, hiring cohort. Once it exists, questions that were unanswerable become routine: do branches with higher median composites retain accounts better, does a specific hiring source produce higher scores at month twelve, does a coaching program move the distribution or just the top quartile.

Sources

• https://hbr.org/2016/10/the-performance-management-revolution
• https://www.gartner.com/en/human-resources/topics/performance-management
• https://www.shrm.org/topics-tools/tools/hr-answers/how-to-develop-employee-performance-metrics
• https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/harnessing-the-power-of-performance-management
• https://www.bls.gov/ooh/sales/retail-sales-workers.htm
• https://www.nist.gov/baldrige/baldrige-excellence-framework
• https://www.investopedia.com/terms/k/kpi.asp
• https://www.salesforce.com/resources/articles/sales-performance-management/
• https://www.naw.org/
• https://www.census.gov/programs-surveys/awts.html

Related on PULSE

• [How Many Attendants Should I Schedule Each Day at My Car Wash?](/knowledge/tl0067)
• [How Many Sales Reps Do I Need to Hire for My Logistics Company?](/knowledge/tl0058)
• [How Many Salespeople Do I Need to Hire for My Car Dealership?](/knowledge/tl0052)
• [How Many Producers Do I Need to Hire for My Insurance Agency to Grow My Book?](/knowledge/tl0015)
• [How Do I Figure Out How Many People to Schedule Each Day and at What Times for My Single Store?](/knowledge/tl0002)