Digital Analytics & Measurement
We build the measurement that lets you argue about everything else honestly.
Happy Cog's analytics team handles GA4 and Google Tag Manager implementation, server-side migrations, consent configuration, attribution modeling, warehouse-backed reporting, and the testing programs that come after. Most engagements open with an audit, because the numbers a company is making decisions on have usually gone unexamined for several years and nobody inside the building is sure which ones to trust. The audit ends with a prioritized roadmap, and our team can take on the fixes from there.
The number the bidding algorithms optimize toward.
The number in the monthly marketing report.
The number sales and finance will defend in front of the board.
Next quarter's budget gets set off the top bar, the marketing team is judged on the middle one, and the business is run on the bottom one. All three are in the same meeting.
A gap under ten percent is normal. At thirty percent, at least one of these systems is wrong, and the audit is how you find out which one.
We've been describing what happens on websites since 1999.
We came out of the web standards movement, which means we were writing markup that accurately described a page long before anyone was paying us to. Specifying a dataLayer is that same discipline pointed at a different problem, and it's why our implementations tend to survive the next redesign. The tools have changed several times and the vocabulary has changed more often than that, but the job of describing what a person did in terms a machine can count has stayed the same the whole way through.
Measuring websites since before analytics was a product category.
Our bid and budget technology has been making live decisions, as often as every thirty minutes, since 2013.
Slate inquiry records matched to on-site behavior for higher education clients.
-
1999Founded
-
2000sWeb standards and accessibility
-
2013Machine learning in production, and our first enrollment-to-behavior joins
-
2023GA4 migration across the client base
-
Going ForwardServer-side, consent mode, and AI referral tracked as its own channel
Selected clients
Five situations bring most people to this page.
Nobody at your company knows who installed the tracking.
When we started with a financial risk advisory firm, we found two Google Ads conversion pixels added in November 2020 by someone no one there could identify, and neither had been touched since beyond a consent setting. Their marketing lead couldn't get into her own GA4 property. This is more common than not, especially at organizations that have changed agencies once or twice, and it's the reason an audit is usually the first thing we sell.
Two platforms report the same revenue differently, and you don't know which one is right.
A beauty supply retailer we run media for had a trigger change quietly point five platform conversion tags at a dataLayer that used different variable names for revenue. Bing, LinkedIn, Simpli.fi, and Yahoo all kept reporting conversions at the normal rate while the revenue attached to them drifted, and because the conversion counts looked fine, nothing surfaced for weeks.
The dashboard is green and the pipeline isn't.
Every channel reports a win, the board asks why revenue is flat, and nobody in the room can answer it. Last-click attribution credits whichever channel sat closest to the form, so branded search takes the win for demand that programmatic and organic created, and next quarter's budget follows the credit instead of the cause.
Your best conversions happen after the form, and the ad platforms can't see them.
Lead volume climbs, sales quality drops, and the algorithms keep optimizing toward the only thing you've given them to optimize toward. Without offline conversion import, both Google and LinkedIn are bidding hard to produce more of the leads your sales team is already ignoring, and they'll keep getting better at it.
Consent changed what you can measure, and nobody told you.
Hard-coded tags firing outside the consent manager entirely, a GTM container that doesn't load until someone accepts the banner (which suppresses far more than anyone expected), and modeled conversions the marketing team doesn't trust because no one has explained where the modeled portion comes from. We've walked into all three.
Analytics is four jobs, and most vendors sell one of them.
Most vendors handle one piece of this (installing the tags, say, or building the monthly dashboard). That leaves no one accountable for whether the dashboard shows the right number or able to fix it when it doesn't. There are four jobs here, and a measurement program that skips any of them will eventually produce a confident answer to the wrong question.
- Instrumentation.
- Getting the events, the dataLayer, and the conversion definitions right, so the data describes what actually happened rather than what somebody assumed would happen.
- Attribution.
- Deciding how credit gets assigned across channels, and closing the loop from an ad click through to recognized revenue. This is a modeling problem, and no amount of clean tagging solves it on its own.
- Interpretation.
- Working out why a number moved, which takes someone who knows your business and your campaigns rather than someone who only knows your property.
- Definition.
- Deciding what the number ought to be in the first place. Almost nobody sells this, and it's usually the thing holding a program back.
The work covers six areas, and you can engage us on one of them or all of them.
Most programs pull from several. Each area is scoped and priced on its own, and you can hire us for a one-time audit with a prioritized roadmap and nothing else.
Our analytics lead starts with the business questions your leadership is asking, then works backward to the small set of metrics that answer them and the event architecture that produces those metrics. Part of that work establishes what a good number would look like for your business rather than what the current number is. What comes out is a measurement plan naming each KPI, its target, and the reason the target is what it is.
Our team writes dataLayer specifications, works with your developers to get them implemented correctly, and QAs them against live containers rather than trusting preview mode. The audit pass that usually precedes this finds the redundant pixels, the duplicate conversion definitions, the tags still firing for campaigns that ended three years ago, and the events that were named inconsistently enough that nothing downstream can group them. We also work in Adobe Analytics when that's the environment you're in.
Our engineers build and run server-side GTM containers, configure Google Consent Mode and enhanced conversions, and handle consent management platform work across OneTrust, Termly, and the rest. Server-side tagging improves data accuracy and gives you control over what leaves your site, and our engineers size the setup to your traffic volume and risk profile. On consent, we work out with you what loads, when it loads, and what happens when a visitor declines, then document it so anyone on your team can explain the setup.
Our team captures the click identifiers each platform sends through (gclid for Google, msclkid for Microsoft, li_fat_id for LinkedIn, vmcid for Yahoo) and persists them in a first-party cookie, so someone who clicks a paid ad and then browses for ten minutes before converting still carries the identifier into the form. Those identifiers flow into your CRM through HubSpot or Salesforce, and back out to the ad platforms as offline conversions once a deal reaches the stage that matters to you, whether that's a qualified lead, a closed-won deal, or a specific revenue threshold. Where the spend justifies it, we run media mix modeling alongside the deterministic work, because attribution and incrementality answer different questions.
Our team builds the warehouse layer, sets up GA4 BigQuery exports, and joins behavioral data to CRM records, platform spend, and whatever else has to sit in the same table for the question to be answerable. Dashboards get built in Looker Studio, Tableau, or Power BI depending on what your organization already runs. We rebuilt our own agency reporting stack on this model, moving off a third-party platform and onto warehouse-backed reporting we control, so we're recommending a setup we run ourselves.
Our team runs structured CRO programs: developing hypotheses from the analysis, designing A/B and multivariate tests, and reading results with the statistical rigor to tell a real effect from a noisy week. Behavioral tools (Crazy Egg and Microsoft Clarity, mostly) show where people are getting stuck, and the tests get tied back to revenue or qualified leads rather than to a micro-conversion that's easy to move and doesn't matter. Before a test launches, we check that your traffic can bring it to significance in a reasonable time, so each test is set up to produce a result you can act on.
Connects all six
Attribution connects your analytics to where the media budget goes.
Every ad platform optimizes toward whatever you've told it a conversion is. That makes the conversion definition the input that matters most in a media program, ahead of the bid strategy, the audience, and the creative, because a badly chosen definition means the algorithm spends the next quarter getting better and better at producing something you don't want. Feed Google Ads a form fill and it will find you more form fills, including all the ones your sales team deletes.
The fix is to send the platforms the outcome that matters, which means capturing the click identifier at the front of the journey and sending the result back from your CRM at the end of it. For our own clients we've set this up so the conversion Google optimizes toward is a closed-won deal rather than a form submission, which changes what the bidding algorithm chases. We recommend starting measurement alongside the media program rather than six months afterward, since this work touches analytics, paid media, and whoever controls the website.
On the organic side, the same problem shows up differently. AI-referred sessions from ChatGPT, Perplexity, Gemini, and Claude arrive as their own channel in GA4 and need to be tracked that way, while the citations that never produce a click still shape somebody's shortlist and have to be measured separately from traffic. A search program reported only in sessions can look like it's failing in month four of a twelve-month program, even while it's working.
No attribution model is correct, whatever the people selling attribution software tell you. They're all approximations, so we choose the one whose distortions you can live with for the decision you need to make.
Attribution windows and upper funnel
Platform attribution is generous by design.
Every marketing leader has sat in this meeting. The display or CTV vendor presents strong reach, healthy engagement, and a conversion volume that doesn't appear anywhere in your analytics. Every platform competes for your budget against every other platform, and the one that shows the most conversions tends to win, so the default windows and models are set to give the platform the best possible picture rather than the most accurate one.
Most platforms also use last-touch attribution inside their own ecosystem. Run Meta, Google, TikTok, and programmatic at the same time and you get credit overlap, where several platforms each claim the same conversion in full. That is mathematically impossible, and your analytics platform will confirm it. We report each platform's claims next to what your analytics shows, so you can see where they overlap.
| Platform | Default click lookback | Default view lookback |
|---|---|---|
| Google Ads, display | 30 days | 1 day |
| Google Ads, YouTube | 30 days | 3 days |
| Meta | 7 days | 1 day |
| TikTok | 7 days | 1 day |
| 30 days | 30 days, one of the most aggressive defaults in the industry | |
| 28 days | 7 days | |
| Yahoo DSP and The Trade Desk | 30 days, varies by campaign | 30 days, varies, audit per account |
Click-through conversions
A traceable action connects the ad to the outcome, so we trust these more. Sensible windows still matter.
View-through conversions
Someone saw the ad, didn't click, and converted later. The platform assumes the ad influenced the conversion. It may have, but how much you count (and over what window) should reflect that uncertainty.
CTV, the special case
There is no click on connected television, so view-through is the only path available and we count it at full value. So we set the window carefully: a 30-day CTV window on a considered purchase captures a great deal of conversion the ad had nothing to do with.
A starting framework for mid-funnel display
100% of click-through conversions within a 30-day lookback.
20% of view-through conversions within a 2-day lookback.
Platforms won't count 20% of view-throughs for you. We pull the raw number, apply the multiplier, and report the raw and adjusted figures side by side with the discount factor named. Both levers, the weighting factor and the window, get set from your own conversion lag and time lag reports rather than from a default, and we override the guideline whenever your data says something different.
Worked example
182 click-through conversions at full value, plus 640 view-through conversions counted at 20%, which is 128 adjusted. Total adjusted conversions: 310. The unadjusted number a DSP would have reported is 822.
Three lenses, three cadences
No single measurement approach tells the whole story, so we layer three.
Platform attribution
Right for pacing and day-to-day optimization. Overstates individual channel contribution, and walled gardens can't deduplicate against each other.
Incrementality testing
True causal lift against a holdout, which is the investment most programs are not making. Platform-reported view-throughs commonly overstate causal influence by two to five times, with retargeting the worst offender. It needs volume, and not every platform supports it.
Media mix modeling
Portfolio-level revenue by investment, including offline effects. Slow to update, expensive for smaller programs, and it needs a long data history.
We also track evidence that doesn't depend on any platform's view-through math (branded search trend, direct traffic during flight periods, and GA4 attribution paths), which gives you an independent check on what the platforms report at any budget size.
Runs through all six
AI is only as good as the signal you hand it.
Our paid media technology has been making live bid and budget decisions since 2013, long enough to find out what a model needs from the data underneath it and what happens when it doesn't get it. Automated bidding, Performance Max, and every AI campaign type on the market are conversion-signal amplifiers, so the cleaner your conversion definitions are, the more automation pays off.
We use AI throughout the analysis work, including anomaly detection across accounts and the first pass on large query and behavioral datasets, so our analysts can cover far more data than they could by hand. We build the review and approval steps around how your team wants findings delivered.
If you're hiring us to design or build the site, the measurement work happens during the build.
Plenty of clients come to us for a redesign or a replatform with no intention of buying an analytics retainer. Our analytics team does the measurement work during the build, since that's when most of the decisions that let you evaluate the new site get made.
The dataLayer is a build requirement.
Our analytics lead writes the specification during development, and it lands as tickets with acceptance criteria in the same sprint as the components, rather than as a document handed over at launch for somebody to retrofit.
Event naming gets settled during content modeling.
The same discipline that decides entry types and fields decides what an event is called and what parameters it carries, and inconsistent naming is one of the few analytics problems that gets more expensive every month you leave it.
Consent architecture is decided before launch.
We decide which tags load, when they load, what happens for a visitor who declines, and how the consent state reaches the platforms. Settling them during the build means every tag runs through the consent manager from launch day.
We capture a baseline before the old site comes down.
Almost every redesign launches without a before number, and then the argument about whether it worked runs for a year with nothing to settle it. Our team records the metrics we'll be judged on while the old site is still live, and we agree with you on what would count as success before the new one ships.
Most measurement programs never define what a good number would be.
A company will tell us its conversion rate is 2.1% and ask whether that's good. It depends entirely on what the traffic is, what the product costs, how long the buying cycle runs, and what the business needs the site to produce, and an industry benchmark answers none of those. A benchmark tells you what average looks like across companies that aren't yours.
So we do this part first. Our analytics lead runs a working session with the people who own the outcome, and what comes out of it is a short list of metrics, a target for each one, and a written reason the target is what it is, derived from your revenue model rather than from a report someone downloaded. That document is short on purpose. If the list of KPIs runs past six, we haven't finished the conversation.
Then we set the cadence.
Each month, we send a written read on what moved, why it moved, and what we're doing about it.
And we're specific about what counts as evidence.
One good week isn't a result. Our team is direct about which changes we can attribute confidently, which ones we can only correlate, and which ones would need a holdout test to settle properly, because a program that treats all three the same eventually makes an expensive decision on a coincidence.
Measure what's meaningful, not what's easy.
Where you are now
Five phases, and you have to walk them in order.
Most organizations we meet collect plenty of data and struggle to turn it into decisions. We assess maturity across five categories (data collection, measurement framework, reporting, analysis and application, and audience segmentation), and the phase that comes back tells you which work is next. You can't run a credible testing program on collection you don't trust, so the sequence isn't a preference.
-
Phase 1
Unstructured
"What data?"
Out-of-the-box pageviews, no defined KPIs, reporting when somebody asks.
-
Phase 2
Descriptive
"What happened?"
Custom events on the interactions that matter, dashboards that show a trend, regular reporting.
-
Phase 3
Diagnostic
"Why did it happen?"
Segmentation, hypotheses you can test, and analytics joined to CRM and platform data.
-
Phase 4
Prescriptive
"What should we do?"
A structured testing program, and analysis that changes design, content, and budget decisions.
-
Phase 5
Predictive
"What might happen next?"
Forecasting, predictive audiences, and lead scoring built on a warehouse you control.
The whitepaper
Beyond the Numbers: A Strategic Roadmap to Digital Analytics Maturity
The full model, with the characteristics of each phase, the practical steps out of it, and a worked example of what each one looks like inside a real marketing team.
A Strategic Roadmap to Digital Analytics Maturity
Here's how an analytics engagement runs.
After you sign off on the measurement plan, our team sets up the tagging to match it.
-
01 · Weeks 1 to 3
Audit and baseline
Our team audits the full container and property: every tag, trigger, and variable, every conversion definition, the consent configuration, and the data quality problems that follow from all of it. You get a written findings document, a tracking document that inventories what exists and what it's supposed to do, and a prioritized roadmap, walked through on a working call. We write the roadmap so it's ready to implement, whether our developers handle the fixes or yours do.
Full tag and property audit Consent compliance check Tracking document Prioritized roadmap -
02 · Weeks 2 to 4
Measurement plan and KPI definition
The business questions, the metrics that answer them, the target for each one, and the dataLayer specification that produces the data. You sign off here, because this decides what the implementation is for.
Business questions KPIs and targets dataLayer specification Your sign-off -
Sign-off gate. Implementation starts once you approve the measurement plan.
-
03 · Weeks 4 to 10
Implementation and QA
Tagging built to the specification, server-side migration where it's in scope, consent configured, and attribution and offline conversion import wired end to end. QA runs against live containers with real transactions rather than in preview alone, because preview mode passes things that production doesn't.
Tagging and dataLayer Server-side and consent Attribution and offline import Live QA -
04 · From roughly week 8, ongoing
Reporting and analysis
Dashboards built where your team already works, the monthly read, and the analysis that turns a number into a recommendation somebody can act on.
-
05 · Ongoing
Testing
The CRO program starts once there's enough clean data behind it to design a test worth running. That way every test runs on measurement you can trust.
Each phase is scoped and priced on its own. On timing, fixes from the audit take effect within days of deployment, attribution improvements usually show inside a quarter, and testing programs run on a schedule set by your traffic. We'll tell you which of those a recommendation falls into.
Results
What the measurement work paid for.
Return on ad spend improvement within six months for a direct-to-consumer eyewear brand, on a mature paid program we took over and re-instrumented.
Conversion lift for a B2B life sciences company after a site rebuild and an integrated media program.
Drop in cost per acquisition in 90 days for an education publisher.
Above its revenue target, where a B2B consultancy's paid program finished after we rebuilt the site and took over search engine marketing.
We can fix what we find, which is rarer in this discipline than it should be.
Most analytics engagements end at the findings document. You get an audit, a spreadsheet of recommendations, and a quarterly call, and then the recommendations sit in a queue behind everything else your development team is doing until somebody runs the audit again next year and finds the same issues. We're 87 specialists across strategy, design, engineering, search, and media, so the person who finds the broken tag can hand it to a developer down the hall, and the person defining the conversion is on the same team as the person bidding against it.
Two things don't change, whatever you hire us for. Every analytics property, every tag manager container, every warehouse project, and every platform login stays in your name, and you keep them if you leave. And you see what we see, in near real time.
If the problem turns out to be the offer, the landing page, or the product, we'll say so and bring in our strategy, design, and media teams.
What's worth measuring changes depending on what you're selling.
The technical foundation is the same everywhere. What changes is which outcome matters, how long it takes to arrive, and whether it happens anywhere near the website.
Higher education
We've been joining Slate inquiry records to on-site behavior since 2013 (in GA4 today), so admissions teams can see which pages, campaigns, and keywords produced applicants rather than sessions, on programs including Adelphi University and St. Joseph's University New York. The measurement problem here is the length of the cycle: a prospective student's first visit and her deposit can be eighteen months apart, which breaks every default attribution window in every platform.
Non-profit
We track donations and cost per donation, and we measure funds raised against media spend, which is the number a board asks about. We've done this for the Central Park Conservancy, Results for America, and the Posse Foundation. The harder half is measuring awareness where there's no click at all, and for that we track branded search volume as the leading indicator, since a campaign that works shows up as more people searching your name before it shows up as more people giving.
Franchise and multi-location
We attribute each call or form to the right location, treat Google Business Profile interactions as a measured channel rather than a black box, and give corporate marketing a view across the network that individual franchisees can't dispute. We use branded search volume against competitors as a baseline here too, because in most of these categories the visibility gap is the whole problem.
E-commerce
Revenue accuracy is the recurring issue, since the platform, the analytics property, and the finance system will all report a different number for the same month and somebody has to reconcile them. We implement GA4 ecommerce tracking, keep revenue variables consistent across every platform tag, and agree with you on which number is the number.
B2B and long sales cycles
The decision happens months after the first touch and usually off the website entirely, so offline conversion import and CRM integration are the core of the program here. Without them, your ad platforms are optimizing on the first cheap signal they can find.
Questions we get asked before an analytics engagement starts.
The quickest test is to pick one conversion and check whether GA4, your ad platforms, and your CRM agree on how many of them happened last month. They almost never match exactly, and a gap under about 10% is normal. A gap of 30% or more means something is wrong, and the usual culprits are duplicate tags, a consent configuration nobody documented, or a conversion defined differently in each system.
Our audit covers every tag, trigger, and variable in your container, every conversion definition and where it's used, your consent configuration and whether your tags actually follow it, data quality across your key events, and the attribution setup connecting your channels to outcomes. You get a written findings document, a tracking document, and a prioritized roadmap. We scope it as a fixed fee with a defined deliverable, so you know the cost before it starts, and what drives the number is the size of your container and the number of platforms in play.
Almost always we can fix it. A rebuild makes sense when the event naming is inconsistent enough that no historical data can be grouped reliably, or when the property was configured with filters and settings that permanently excluded data you needed. Short of that, fixing what exists preserves your history, which a rebuild can't give back.
Offline conversion import sends outcomes that happen after someone leaves your website (a qualified lead, a closed deal, a specific revenue amount) back to Google, Microsoft, LinkedIn, or Meta, so their bidding algorithms optimize toward the thing you actually want. Without it, the platforms optimize toward form fills, which means they'll get progressively better at producing leads your sales team ignores. It's the highest-return measurement work we do for most B2B clients.
Server-side tagging improves data accuracy by moving measurement out of the browser, past ad blockers and browser restrictions, and gives you control over what data leaves your site. It pays off fastest when you're spending enough on media that a 10 to 20% measurement gap is a large dollar figure, or when you're in a regulated category where controlling outbound data matters on its own. We'll look at your media spend and your data needs and recommend the setup that fits.
Directly and substantially, and the size of the effect depends on choices most teams make by accident. If your container is gated so that nothing loads until a visitor accepts, you lose everyone who ignores the banner. If tags load and respect consent signals individually through Google Consent Mode, you keep modeled data for the visitors who decline. Both are defensible, and they produce numbers that can differ by a wide margin, so the decision should be deliberate rather than whatever the vendor's default was.
We'll recommend one after looking at your channels and how much conversion data you have. Last-click works well if most of your spend goes to search. GA4's data-driven attribution is a good fit once you have steady conversion volume. Media mix modeling shows whether a channel brings in sales you wouldn't have gotten otherwise. It pays off once your media budget is big enough that putting money in the wrong channel would cost more than the modeling. Whichever model you go with, we'll show you how it credits each channel.
Both, with different confidence levels, and we'll be clear about which is which. Attribution tells you what happened and where credit plausibly sits. Incrementality testing (a geographic holdout, a controlled spend reduction) tells you whether the channel caused anything, which is a stronger claim and a more expensive one to get. Most programs should run attribution continuously and incrementality tests occasionally, on the channels where the spend is big enough to justify the question.
Two separate things get tracked. Referral sessions from ChatGPT, Perplexity, Gemini, and Claude do arrive in GA4 as their own channel, so those get reported like any other source. Citations that never produce a click get tracked through prompt monitoring instead, where our SEO team establishes a set of representative prompts at baseline and watches how often you're named and how you're described over time. A citation with no click still shapes someone's shortlist.
You need one when the question you're asking requires joining data that lives in different places, which is most interesting questions past a certain size. GA4's BigQuery export is free to turn on and the storage costs are small, so we usually recommend enabling it early even for clients who won't use it for a year, because it only captures data from the day you switch it on and there's no way to backfill.
That's the point of building them. We build in Looker Studio, Tableau, or Power BI depending on what your organization already runs, we train your team, and we leave written documentation covering what each metric means and where it comes from.
The audit and measurement plan phase is a fixed fee with a defined deliverable, so you can see the roadmap before committing to anything else. Implementation is scoped per project against that roadmap. Ongoing analysis, reporting, and testing run as a monthly retainer sized to the hours the program needs, and what drives that number is how many properties and platforms are in play, whether your developers or ours are implementing, and whether testing is in scope.
Both. Our analytics lead writes the dataLayer specification during development so it ships with the components rather than getting retrofitted, settles the consent architecture before launch, and captures a baseline of the metrics you'll be judged on while the old site is still live. That baseline gives you a clear before-and-after comparison once the new site launches.
You do, always, including while we're working in them and after we're gone. Every analytics property, tag manager container, warehouse project, and platform login stays in your name. We ask for the access we need to do the work, and if the engagement ends you keep everything, with no export process and nothing to negotiate.
Yes, and a lot of our analytics work runs that way. Our team writes implementation-ready tickets with acceptance criteria, sits in your sprint rituals where that's useful, and QAs the work after it ships, which is a different arrangement from handing over a specification and hoping. Where your developers are implementing, what we need is a staging environment and a way to see what changed, since verifying that a fix landed correctly is part of the job.
The audit produces findings in the first three weeks, and the fixes that come out of it (a double-firing conversion, a broken revenue variable, a consent misconfiguration) usually take effect within days of deployment. Attribution and offline import changes show up in media performance inside a quarter, because the bidding algorithms need a few weeks of the new signal before they adjust. Testing programs depend on your traffic volume, and we'll tell you at the start whether you have enough of it.
Measurement work usually leads somewhere more specific.
Services this connects to
Industries we measure for
Related reading
- Beyond the Numbers: A Strategic Roadmap to Digital Analytics Maturity (PDF)
- Mastering Offline Conversion Import: A B2B Imperative Beyond the Form Fill
- Connecting KPIs to Goals and Objectives
- Tips for Nonprofits to Maximize Value from Web Analytics Data
- Beyond the Cookie: The Executive Playbook for Privacy-First Measurement
- Beyond the Blue Links: Surviving (and Thriving) in the Age of Answer Engines
- What ChatGPT's May Update Means for Your Brand's Website Traffic
If you don't trust your numbers, tell us what you're seeing.
Whether that's a full measurement program or a single audit that establishes where you stand, we'd like to hear it. We'll tell you what we'd do about it and what it would take.
Tell us what you're seeing.