La bibliothèque de compétences nylo

Is your tracking lying to you? Sweeps every campaign for tracking gaps — spend with zero conversions, conversions with no value, and value/conversion ratios that suddenly break — so you trust your numbers before you optimize.

Grade your account architecture. Flags budget concentrated in too few campaigns, ad groups stuffed too thin or too dense, and orphaned entities — a weighted structural scorecard, not a metric dump.

Find every campaign and ad group that breaks your naming pattern. Infers the dominant scheme from your own entities, lists the outliers that won't parse cleanly, and emits a rename change-set you apply in your platform.

See where mobile eats budget and desktop converts (or vice versa). Splits spend, CVR, and CPA by device, ranks the gaps, and emits a device bid-adjustment change-set you apply through your own ad platform.

Find regions burning budget with nothing back and geos quietly converting cheap. Splits performance by location, flags wasted and high-CPA geos, and emits a geo bid/exclusion change-set to apply in your own platform.

Expose the hours and days that waste budget. Splits performance by hour-of-day and weekday, separates a real efficiency pattern from noise, and emits an ad-schedule bid-adjustment change-set you apply in your own platform.

Reconcile what each ad platform claims it drove against your ground truth (GA4 / store orders). Flags double-counting and attribution inflation across channels so your blended numbers add up before you reallocate budget.

Find the destination URLs your paid traffic lands on that spend big and convert poorly. Ranks landing pages by paid-traffic efficiency and flags the leaks worth a CRO sprint — distinct from organic page audits.

Are you on track to hit the number this month? Projects each campaign's month-end spend and conversions against goal with a real forecast, flags over/under-pacing, and stands up an alert so pacing problems surface before month-end.

Your metric moved and you need the real reason. Decomposes any spend/ROAS/CVR/revenue swing into the segments that actually drove it — ranked by contribution, with exact before/after numbers.

Something spiked or cratered on a specific day and you need to know what and why. Runs real statistical anomaly detection to pinpoint the exact dates, then isolates the segment behind each outlier.

Numbers are down — but is it just the usual weekly/seasonal pattern or a genuine problem? Decomposes the series into trend + seasonality + residual so you stop firefighting normal dips.

An aggregate number fell but the cause is hiding in one slice. Finds the single campaign, audience, product, or country dragging the whole account down — ranked by how much each subtracted.

Is the metric actually underperforming, or just where it should be? Fits a real forecast band and tells you whether today's number is inside the expected confidence interval or genuinely off-pace.

Conversions or ROAS dropped — but where in the funnel? Splits the change across impressions, CTR, CVR, and AOV so you fix the actual broken stage instead of guessing top-of-funnel.

You changed bids, creative, or budget and the metric moved — but did your change cause it, or did the whole account drift? Runs difference-in-differences against an unaffected control to separate signal from noise.

Conversions fell off a cliff — but is it real, or did your pixel/attribution break? Cross-checks platform-reported conversions against upstream signals (clicks, sessions, orders) to tell a tracking gap from a true performance drop.

Tells you if you're on track to hit a revenue, conversion, or lead goal by its deadline — forecasts the metric to the target date and reports projected attainment %, the gap, and the daily run-rate needed to close it.

Schedule a guard that forecasts a metric forward and alerts the moment actuals fall outside the prophet confidence band — your early warning that something broke the trend, even when the metric is trending or seasonal.

Decomposes any metric into trend + weekly/yearly seasonality + residual so you know your real peaks and troughs, then forecasts ahead respecting that seasonality — answers 'when do we peak' and 'is this dip just seasonal?'.

Forecasts product/SKU demand (orders or units) N days ahead with confidence bands so you can plan inventory and budget — surfaces accelerating vs decaying products and emits a restock/markdown change-set for your store or 3PL.

Projects inbound lead or conversion volume by channel for the coming weeks so sales and ops can staff and budget — reports expected leads, the confidence range, and which channel is trending up or down fastest.

A tight few-days-ahead forecast for fast-moving or non-seasonal metrics using ARIMA — best for 'what does the next 3-7 days look like' when prophet's seasonality assumptions are overkill or the series is short.

Forecasts each campaign to period-end, compares projected delivery against budget, and emits a reallocation change-set — shift spend from over-pacing or decaying campaigns to under-pacing or accelerating ones — for you to apply in your ad platform.

Flags days where any KPI jumped or collapsed far beyond its normal range using Nylo's standard-deviation model — not eyeballing — so you only hear about real anomalies on spend, conversions, traffic, or revenue, not noise.

Catches the moment a metric departs from its recent moving-average trajectory — a slow CPA creep or gradual CTR decay you'd miss in a single-day glance. Powered by Nylo's moving-average model.

Knows your normal weekly/seasonal shape (weekend dip, Monday surge) and alerts only when a day deviates from that expected pattern — not just from a flat average. Uses Nylo's seasonal-decomposition model.

Weights the most recent days most heavily to catch a shift while it's still forming — an early-warning radar for momentum building before it shows in a slower average. Uses Nylo's exponential-smoothing model.

Runs the spike/drop detector across every connected provider at once and returns a single ranked board of where things broke today — so you triage the whole account in one pass instead of channel by channel.

When a top-line KPI spikes or drops, drills into the breakdown to find the exact campaigns or segments driving it — turning 'spend is weird' into 'these 3 campaigns caused 80% of it'.

Captures any anomaly monitor into a Nylo cloud flow that runs server-side on a schedule and pings Slack, email, or Teams the instant a breach fires — guaranteed even with your laptop closed.

After an anomaly is confirmed (e.g. a runaway-spend campaign), emits the exact pause or budget-cap change-set and hands it to your own ad-platform MCP to apply — Nylo stays read-only, you stay in control.

Prove a campaign's TRUE incremental lift by comparing treated vs held-out regions before and after launch. Runs the real difference-in-differences model with significance — not a naive before/after eyeball.

Did the new campaign, creative refresh, or pricing change actually move the needle? Measures the real causal step-change at a cutoff date against a comparison group, with confidence, so you don't credit a seasonal wave.

Settle the creative debate with statistics, not vibes. Compares variant vs control on your success metric and reports whether the difference is real or just noise at 95% confidence.

Tested a new landing page, checkout, or offer? Get a statistically sound verdict on whether the new version truly lifts conversions or revenue versus the control, before you roll it out everywhere.

See how your reported conversions and ROAS swing across attribution windows (1-day vs 7-day click). Quantifies how much of your 'performance' is a windowing artifact so you read numbers honestly.

Bridge platform-claimed conversions and your real incremental lift. Pairs the platform's self-reported numbers with a holdout/DiD lift estimate and reports the over-claim ratio so budget decisions use verified impact.

Before you run a geo or holdout test, check it's actually measurable. Validates the split dimension, baseline balance, and sample size so you don't burn weeks on an underpowered experiment.

Keep a running test honest with a scheduled watch on the treated-vs-control gap. Alerts your team via Slack or email when the divergence crosses a threshold or the test reaches a decision point.

Run a Bayesian media-mix model across all your channels to get ROI per channel and the optimal budget split for a fixed total — then hands the spend changes to your ad platforms. Use for 'where should my budget go'.

Find where each channel hits diminishing returns — the spend level past which extra budget stops paying off. Use for 'are we maxed out on Meta', 'saturation point', 'is this channel tapped'.

Rank every channel by true modelled ROI and contribution, not last-click — see which channels actually drive the outcome per dollar. Use for 'ROI by channel', 'modelled contribution'.

Equalize marginal ROAS across channels — move budget from the channel with the lowest next-dollar return to the highest until they balance. Use for 'the next $10k', 'shift spend to the better margin'.

Measure the true incremental return of a spend change with a difference-in-differences or A/B test before scaling it. Use for 'did the budget increase actually work', 'validate the scale-up'.

A fast, deterministic weighted scorecard ranking channels or campaigns by spend efficiency (ROAS, CPA, conversion rate) — the cheap pre-read before committing to a full MMM. Use for 'rank channels by efficiency'.

Schedule a recurring check that re-grades your channel mix and alerts when a channel's efficiency drifts far enough to warrant rebalancing. Use for 'alert me when the mix needs rebalancing', 'weekly budget health'.

Find which hook, format, emotion, talent, or color actually drives your best CTR/ROAS — AI tags every ad image, then ranks the traits that track real performance, not your gut.

Audit the make-or-break opening of every video ad. AI scores hook effectiveness from the 0-3s frames and shows which opening emotion, on-screen text, or face presence wins thumb-stop and retention.

Catch ads dying of fatigue before they bleed budget. Tracks each creative's daily CTR/CPM trend and flags the statistically real decline curve — so you refresh on evidence, not vibes.

Settle the UGC-vs-polished debate with statistics, not opinions. AI labels each ad's format, then runs an A/B significance test to prove whether UGC actually beats studio on your KPI — and by how much.

Grade every thumbnail's scroll-stopping power. AI reads text-overlay density, face presence, product prominence, and color contrast, then deterministically ranks thumbnails by impression-weighted CTR.

Map the emotional journey of your video ads frame-by-frame. AI tracks the emotion arc and detects the shift points, revealing which narrative shape (calm to excited, problem to relief) holds attention and converts.

Find the perfect moment to show the product. AI detects when each video ad first reveals the product and ranks reveal-timing buckets against ROAS — so you know whether to lead with product or hook first.

Never let a fatigued or winning creative slip past you. Schedules a guaranteed cloud watch that alerts Slack or email when any ad's CTR drops below threshold or a fresh top performer emerges — on autopilot.

Pace spend against a period budget: month-to-date burn vs ideal run-rate, projected end-of-period total, and a +/-% on-track verdict. Use for 'are we pacing right', 'will I overspend this month'.

Forecast where spend LANDS at period end (prophet, with confidence intervals + weekly seasonality) and compare to the budget cap — so you see overspend before it happens, not after.

Detect the exact day daily spend breaks its normal band (standard-deviation anomaly) so a runaway bid, broken cap, or feed glitch is caught in hours, not at month-end. Use for 'spend spiked', 'why did cost jump'.

Rank weekday and hour-of-day efficiency (CPA/ROAS by daypart) to expose when budget is wasted on dead hours vs starved on peak windows — emits an ad-schedule bid-adjustment change-set you apply in your own platform.

Surface campaigns hitting their daily cap (lost impression share to budget) vs chronic underspenders, so you reallocate idle budget to demand you're leaving on the table. Emits a budget change-set for your platform.

Track a fixed-budget burst (sale, launch, event flight) intraday: daily burn vs remaining days, projected exhaustion date, and whether you'll overspend before the flight ends or leave budget unspent. Use for 'will the sale budget last'.

Turn any pacing check into a guaranteed Nylo cloud flow that fires only WHEN spend pace breaches a guardrail (projected over budget, or daily spend over cap) and alerts Slack or email — runs server-side even with your laptop closed.

One true blended ROAS and CAC across every paid channel plus store revenue. When channel-reported ROAS looks great but the bank account disagrees, this reconciles paid spend against actual store revenue.

Where should the next dollar go? Runs a Bayesian marketing-mix model over per-channel spend vs a single outcome to get ROI per channel and a budget reallocation plan — then hands the spend changes to your ad platforms.

Are two channels fighting over the same conversions? Detects spend overlap and cannibalization (brand search vs Meta retargeting) by reading both channels at a comparable grain and testing whether scaling one suppressed the other.

Stitches paid ads with organic (GA4, GSC, organic social) into one funnel so you see total demand, not just the paid slice. Answers 'how much traffic and revenue is paid actually incremental to organic?'

Reframes channel value beyond last-click — surfaces which channels assist conversions vs claim the final click (GA4 attribution paths), so upper-funnel channels aren't unfairly defunded.

Ranks every channel on one weighted scorecard (spend, ROAS, CAC, growth) so you can say objectively which channel is your best — not just whichever platform's dashboard you opened last.

Set a blended ROAS or CAC target across all channels and get alerted the moment combined performance breaches it — guaranteed cloud run, even with your laptop closed. Turns the blended view into an always-on guardrail.

Build the monthly or weekly retention curve for your store: which sign-up cohorts keep coming back and where drop-off accelerates. Spot the period when customers churn so you can time win-back offers.

Compare how customers acquired from different sources (channel, campaign, UTM, first-touch) retain and repeat. Find the channel that buys loyal customers vs one-and-done buyers — not just the cheapest click.

Rank your customer segments by realized lifetime value and repeat-purchase rate. See which first-order product, plan, or audience produces the highest-CLV customers so you can double down on what attracts them.

Cross paid-acquisition CAC with cohort lifetime value to see how many months each channel takes to pay back. Surfaces channels you're scaling that never recoup spend — the real efficiency read behind ROAS.

Track the share of customers who place a second order and the time-to-second-purchase, broken out by cohort. Schedules a guaranteed cloud alert when repeat rate or second-order lag degrades versus prior cohorts.

Split revenue, AOV, and margin between first-time and returning customers, and track the mix over time. Reveals whether growth is healthy repeat demand or expensive new-customer churn.

Project where the current cohort's retention curve is heading and forecast the month customers are most likely to lapse, with confidence intervals — so you can launch retention offers before the drop, not after.

Identify the lapsing cohort — customers past their typical repurchase window with no recent order — and emit a ready-to-import win-back audience change-set for your email or ads platform. Nylo reads; you activate.

Rank your catalog by true profit, not revenue. Builds a margin/contribution metric from COGS, discounts, and refunds, then scorecards every SKU so you scale winners and cut quiet money-losers.

Did that promo actually grow profit or just discount sales you'd have made anyway? Uses difference-in-differences to isolate the incremental revenue lift of a discount window vs a control, net of margin given away.

Catch refund and return spikes per product before they eat your margin. Statistical anomaly detection on refund rate by SKU, with a scheduled cloud alert when any product breaks its normal band.

Pinpoint what actually moves average order value: units-per-order vs price-per-unit, broken down by channel, customer type, or collection — so you know whether to bundle, upsell, or reprice.

Quantify how much store revenue your email program really drives. Joins Klaviyo flows and campaigns to store orders, splits flow vs campaign vs broadcast, and frames attributed vs incremental contribution.

Forecast unit demand per SKU and compare to on-hand stock to flag stockouts and dead inventory early. Emits a restock/markdown change-set you push to your store or 3PL — Nylo stays read-only.

Find the products you pay to advertise that lose money after COGS, discounts, and refunds. Joins ad spend to store profit per product and emits a pause/bid-down change-set for your ad-platform MCP.

Find queries where you pay for clicks you'd win organically — join GSC organic rank to paid search-term spend, flag overlap, and design a safe pause test to prove incremental paid value.

Classify every query by intent at runtime (rule-based dimension on query text), then surface high-volume intents you under-serve — no matching keyword or landing page — so you can build the campaign or page that captures unmet demand.

Decompose where CTR and Quality Score leak — rank keywords by the gap between expected and actual CTR at their position, isolate the biggest fixable levers (ad relevance, low rank), and emit a prioritized change-set.

Forecast search demand (impressions / search volume) with seasonality and confidence bands so you can pre-load budget before peaks and avoid throttling — a real Prophet/ARIMA fit, not a trailing average.

Match the same keywords across Google Ads and Microsoft Ads, expose CPC/CTR/CPA gaps, and emit a reallocation change-set — shift budget to the cheaper engine and port your winners to the one you under-invest in.

Cluster raw search terms into themes (runtime concatenate/rule dimension), rank themes by converting volume, and propose new ad groups or single-theme campaigns — positive opportunity mining, the inverse of negatives.

Rank where you lose impression share to rank vs budget, size the recoverable clicks and conversions for each, and emit a prioritized change-set so you fix the highest-value rank or budget gaps first.

Catch branded-query demand and CTR breaks early — split brand vs non-brand queries at runtime, run real anomaly detection on the branded slice, and schedule a guaranteed cloud alert when brand health moves.

Define a true blended ROAS or MER — total revenue over total ad spend across every channel — as a reusable custom metric, then chart it. Use for 'build blended ROAS', 'MER metric', 'account-wide return'.

Turn raw campaign or keyword names into a reusable Brand vs Non-Brand (or Generic/Competitor) dimension via rule-based matching, so every future report can group by intent. Use for 'split brand vs non-brand'.

Build a real profit or contribution-margin metric (revenue minus COGS, fees, shipping, and ad spend) as a reusable custom metric so you optimize to profit, not just revenue. Use for 'profit metric', 'net of COGS'.

Map campaigns or ad sets to a reusable funnel-stage dimension (Awareness / Consideration / Conversion / Retargeting) by rule, so spend and ROAS can be read by funnel stage on demand. Use for 'tag by funnel stage', 'TOF/MOF/BOF'.

Stitch two providers together on a shared key (Shopify orders to Meta by UTM, or Google Ads to GA4 by campaign name) so blended cross-channel queries resolve. Use for 'join Meta and Shopify', 'connect Google Ads to GA4'.

Build any derived efficiency or unit-economics metric the workspace lacks — target CPA gap, AOV, new-customer CAC, cost per qualified lead — as a reusable formula metric. Use for 'create a CPA metric', 'cost per lead'.

Bucket a numeric field into reusable tiers — order-value bands, AOV ranges, spend tiers, creative-age buckets — as a custom dimension so you can group performance by band. Use for 'bucket by order value', 'spend tiers'.

Inventory every existing custom metric, dimension, and relationship, flag broken or stale definitions (missing leaves, dead join fields), and re-validate them so the workspace's semantic layer stays trustworthy. Use for 'audit my custom metrics'.

Turn a recurring exec one-pager into a guaranteed Nylo cloud flow: headline KPIs, period-over-period deltas, and a chart, delivered to email or Slack on a schedule even with your laptop closed.

Build a presentation-ready, multi-channel client deck where every number is queried, not invented: charts per headline, a written narrative spine, and a forward-looking forecast slide with confidence intervals.

Rank campaigns, channels, or clients into a single weighted health score across several metrics with normalization, so stakeholders see one defensible Top/Bottom scorecard instead of arguing over which KPI matters.

Write the plain-English 'what changed and why' summary for any period: cited week-over-week or month-over-month deltas plus the standout movers and anomalies that explain them, ready to paste into an email or update.

Stand up a persistent, always-fresh KPI board for a client or goal: headline metric tiles with period-over-period deltas and goal pacing, laid out with taste so stakeholders self-serve the numbers any time.

Assemble a named pack of saved analyses — the standard tables and breakdowns a report always needs — so the whole report rebuilds with current data in one re-run and can later be scheduled for delivery.

Report whether you're on track to hit the number: set or read a target on a metric, then show actual-to-date, required run-rate, projected end-of-period, and an on-track / at-risk / off-track verdict with the gap quantified.

Frame your results against the outside world: pull your actual metrics, then ground them with sourced industry benchmarks and seasonality context, so the report says not just what happened but whether it's good.

Wake up to your numbers. A cloud flow runs every weekday at dawn and drops yesterday vs the day-before across spend, conversions, and ROAS into Slack or email — no laptop required.

Only hear about it when it matters. Set a hard line — CPA above target, ROAS below floor, spend over cap — and a cloud flow pings Slack or email the moment a metric crosses it, instead of a daily report you ignore.

Never overspend or leave budget on the table. A daily cloud flow tracks month-to-date spend against your monthly cap, projects end-of-month from current run-rate, and alerts Slack or email when you're pacing over or under.

Get warned before you miss the number. A weekly cloud flow forecasts your key metric (Prophet, with confidence intervals) and alerts when the projected trajectory will breach a target or floor — early warning, not post-mortem.

Close the month on autopilot. A flow scheduled for the last day of each month delivers full-month performance with month-over-month and year-over-year deltas, as a charted recap to Slack or email or a saved analysis.

Spot competitive pressure early. A scheduled flow tracks impression-share lost, CPC drift, and search-impression trend, and alerts when auction competition is squeezing your best campaigns — so you react before efficiency erodes.

Babysit every new campaign or product launch. A short-lived daily flow watches just-launched entities for early-warning signals — zero conversions on rising spend, CPA blowout — and emits a reviewable pause/scale change-set you apply in your ad platform.

Stack your real CTR/CPC/CVR/ROAS against current industry benchmarks. Nylo pulls your actual numbers, researches the external standard with sources, and flags where you're ahead or behind.

Catch ad-platform algorithm, policy, and tracking changes before they hurt you. Nylo researches recent platform events for your connected channels, then ties each to the metric it could move.

Map upcoming demand peaks and dips for your category so you plan budget around them. Nylo researches the seasonality calendar and cross-checks it against your own historical pattern.

Understand where your category is heading — demand shifts, emerging formats, competitor moves. Nylo runs grounded market-trend research and frames every finding against your own channel mix.

Turn a finding into durable workspace memory. After any analysis, capture the conclusion as a versioned, sourced fact so it compounds into context for every future session.

Lock in what a term means for this client — 'a qualified lead is X', 'conversion = purchase not add-to-cart'. Nylo stores the definition so every future answer uses the same meaning.

Tell Nylo what to ignore for good — a test ad account, internal traffic, a defunct campaign. Once registered, every report and analysis honors the exclusion automatically.

Turn an industry benchmark or strategic decision into a tracked target. Nylo researches a defensible standard, anchors it to your real baseline, and creates a live goal on the metric.

Sweep campaigns, ad groups, keywords and search terms, then flag every issue ranked by impact.

Find keywords with zero conversions that spent over a threshold, ranked by spend.

Find ad groups above a CPA ceiling and hand off the pauses to your own ad platform.

Mine wasteful search terms and hand off the negatives across the account.

Build a presentation-ready audit deck with charts and recommendations.

Schedule a recurring check that runs in the cloud and alerts you in Slack or email.