Cart Abandonment Tracking Article — DataCops

Body copy (no H1, no frontmatter)

---

Every article about cart abandonment tracking starts in the same place. The 70% statistic. The $260 billion in recoverable revenue. The three-email recovery sequence. How to segment by cart value and recency. How to write a subject line with urgency.

None of those articles ask the question that matters before any of that: are the abandonment events you're tracking actually real?

They are not. Not all of them. On most stores, a meaningful slice of what fires as AddToCart and InitiateCheckout in your pixel, your CAPI, your Klaviyo flows, and your Meta retargeting audiences is bot traffic. Automated scrapers, carding bots, competitor intelligence crawlers, traffic fraud operators, and AI agents that browse like humans but never buy. They add to cart. They initiate checkout. They fire every JavaScript event you have. And your conversion infrastructure swallows all of it without question.

This is not a minor data quality issue. It is the root failure that makes everything downstream unreliable. Your abandonment rate looks worse than it is. Your recovery email list gets poisoned with fake addresses that torch your sender reputation. Your Meta retargeting audience contains bots, and Meta learns from them. Your Lookalike Audiences improve at finding more users who look like the bots that abandoned. Your Cost Per Acquisition goes up, and you conclude the creative was wrong or the offer was weak, when the actual problem is that a percentage of your conversion signal never represented a human being.

Shopify communities have been documenting this for years. Merchants report thousands of fake abandoned checkouts daily, using rotating IPs across 18,000+ addresses, generating synthetic email addresses in predictable patterns, filling in "street 10 apt 2" as the address, and bypassing every front-end protection because they access checkout via Shopify's backend API directly, not through your storefront. These aren't edge cases. This is infrastructure-level fraud targeting the conversion funnel you've spent years optimizing.

Advanced Conversion Tracking: The Technical Implementation Guide covers the full picture of why your conversion data fails. This article focuses specifically on the cart abandonment layer and what every tool in the stack actually does about it.

---

The real question your cart abandonment rate can't answer

When you see a 70% abandonment rate, two questions go unasked. First: how many of those carts were added by humans who had any real purchase intent? Second: how many of those "abandoners" are bots, scrapers, and automated agents that your tracking stack counted as people?

The honest answer is that the tools most stores use cannot tell you. Analytics fires client-side. The pixel fires for every AddToCart event. Server-side tracking doesn't fix this because server-side still depends on the browser sending the event first. If a headless bot using Playwright executes your AddToCart flow on your storefront, the event fires. It hits your pixel. It routes through your CAPI. It gets counted as an abandonment. It joins your retargeting audience. And nowhere in that chain does any tool you're paying for ask whether the session was human.

This is what Layer 4 of broken data infrastructure looks like in practice. Your analytics is half-blocked, half-bot. Twenty-five to thirty-five percent of real humans are never recorded because ad blockers suppress third-party scripts. Thirty to forty percent of the traffic that does get recorded is bots, VPNs, proxies, and AI agents. Server-side was sold as the solution. It solves the ad blocker problem. It does nothing about the bot problem, because it still depends on the contaminated browser event arriving first.

ChatGPT Ads Manager launched May 5, 2026, and 70.6% of LLM-driven traffic is currently misclassified as direct in GA4. Your abandonment attribution models are operating on data where a growing category of non-human traffic is invisible or mislabeled. This is getting worse, not better.

---

Quick answers

How do I know if bots are inflating my cart abandonment rate?

Look for abandonment rate spikes that don't correlate with any change in your creative, offer, or checkout flow. Look for abandoned checkouts with email patterns like name followed by three numbers, or recurring fake addresses. Look at your Klaviyo bounce rates on abandoned cart sequences: if your list bounce rate is above 2%, bot-generated emails are likely in your flow. If your abandonment rate jumped significantly after Black Friday or Cyber Monday with no corresponding revenue movement, bots are almost certainly the explanation.

Does server-side CAPI fix the bot problem in cart abandonment data?

No. Server-side routes events from your server to Meta instead of from the browser, bypassing ad blockers. But the triggering event still originates in the browser. If a bot executes the AddToCart event in a browser session, that event fires to your CAPI the same as a human's. Server-side solves data loss from ad blockers. It does not filter out events that were generated by automated traffic.

What percentage of AddToCart events are typically bots?

Global invalid traffic runs at 20.64% of all digital traffic (Fraudlogix 2026). On Instagram, measured invalid traffic reaches 38%. The Audience Network hits 67%. For ecommerce, the estimate ranges from 20-40% depending on vertical, traffic source mix, and whether you're running paid campaigns that attract click fraud. High-traffic events like BFCM concentrate bot activity significantly.

How does bot cart abandonment affect Meta Lookalike Audiences?

Every event you send Meta trains its optimization algorithm. If bot AddToCart and InitiateCheckout events flow into your CAPI, Meta treats those sessions as signals of high-intent users and builds Lookalike Audiences that resemble them. You end up optimizing toward more bot-like traffic and paying more for it. Project Andromeda, fully deployed October 2025, acts on contaminated signals within hours, meaning the contamination propagates faster than it used to.

Will switching from pixel-only to server-side CAPI improve my cart abandonment attribution?

Yes, meaningfully, but with a ceiling. CAPI vs pixel-only delivers a 17.8% lower CPA on average (Meta via AdExchanger). That recovery is real. But if your CAPI is forwarding bot events along with human events, you are sending cleaner human data and dirtier bot data to the same destination. The attribution improvement is real. The optimization quality improvement depends on what you're filtering before the event fires.

Why do my abandoned cart recovery emails have such high bounce rates?

Checkout bots generate synthetic email addresses. When your Klaviyo or Klaviyo-equivalent flow triggers on every InitiateCheckout event, it triggers on bot-generated addresses too. Those addresses bounce. Each bounce damages your sender reputation, depresses deliverability on the emails going to real humans, and increases the effective cost of every legitimate recovery you achieve.

What should my cart abandonment rate actually be if I filter bots?

No universal answer exists because it depends on your vertical and traffic mix. But if your rate is above 80%, bots are almost certainly contributing meaningfully. A clean signal from verified human traffic typically settles 60-75% across most ecommerce categories, which aligns with Baymard Institute's meta-analysis of 49 studies showing 70.22% as the global average.

---

Who this article is for

The framing below splits tools into two distinct jobs. First, the tracking infrastructure layer: server-side CAPI tools that determine what events get sent to Meta, Google, TikTok, and LinkedIn, and whether any filtering happens before they fire. Second, the recovery layer: email, SMS, and push tools that act on those events to re-engage abandoners. Both layers matter. Most stores have the second layer optimized to a degree that far exceeds the quality of the data feeding it. That imbalance is where the money is getting lost.

---

The tracking infrastructure layer

These tools determine the quality of the signal. They sit upstream of every recovery email, retargeting audience, and attribution model you have.

DataCops

DataCops is the only tool in this category that filters bot traffic at the IP layer before any CAPI event fires. The 361-billion-IP database, covering 146.4B datacenter and cloud IPs, 11.9B VPN endpoints, and 620M+ proxy and anonymizer IPs, runs against each session before an event is dispatched to Meta, Google, TikTok, or LinkedIn. This means bot-generated AddToCart and InitiateCheckout events are suppressed at the source rather than forwarded and counted. The Lookalike Audience your Meta campaigns train on contains human abandoners, not automated sessions. Recovery email audiences contain people who actually browsed your store.

The architecture is first-party by design. One script tag plus one CNAME record points traffic through your subdomain, not a third-party CDN. Ad blockers that block GA4, the Meta pixel, OneTrust, and Cookiebot by name do not block a script running on datacops.yourdomain.com. The TCF 2.2 CMP is bundled and loads from the same first-party subdomain, which matters for the June 15, 2026 Google Consent Mode v2 enforcement deadline affecting all EEA advertisers. Consent is captured correctly, anonymous analytics flow after rejection (because anonymous data is always legal), and cookieless persistent identity activates for users who consent, with no ITP decay and no cookie expiry.

SignUp Cops layered on top closes the fake email address loop in checkout flows. The PillarlabAI case study documented this in hard numbers: 4,560 signups, 4 weeks, 730 real, 84% fraudulent, 650 accounts from one laptop. That ratio does not surprise anyone who has looked at their Klaviyo bounce rates honestly.

What it does not do: Pinterest CAPI, Snapchat CAPI, and deep Shopify-native order-level event fidelity at the millisecond level that Elevar provides for high-volume DTC. SOC 2 Type II is in progress. It is a newer brand versus Stape and Elevar.

CAPI starts at Business, $49/month for 50,000 sessions with Meta, Google, TikTok, and LinkedIn all included. Free plan covers 2,000 sessions with analytics and CMP. Full details at joindatacops.com/pricing.

Right for: Stores on any platform, running paid traffic across multiple channels, who want one architecture that filters bots before they corrupt CAPI, attribution, and recovery audiences. Value: 9/10.

---

Elevar

Elevar is the agency default for high-volume Shopify DTC brands. Mature server-side tracking with deep Shopify-native data layer, order-level fidelity, and clean event deduplication across Meta CAPI, Google Enhanced Conversions, TikTok, and Pinterest (which DataCops and most alternatives don't offer). The Session Enrichment feature is genuinely differentiated: it stitches fbclid and gclid attribution data into server-side events, significantly improving EMQ scores on cart events. Many agencies run Elevar as the data layer and Triple Whale on top for reporting.

The honest problems: Shopify-only, $200/month for 1,000 orders, escalating to $950/month at 50,000 orders. No bot filtering. Bot-generated AddToCart and InitiateCheckout events flow through to Meta the same as human events. No built-in CMP. No multi-platform: if you run WooCommerce, Webflow, or a custom stack, Elevar does not apply. At the $950/month tier, you are paying to forward clean events from your Shopify order stream and contaminated events from your traffic, because the tool has no mechanism to distinguish them.

Right for: Shopify-only brands doing $500K+ GMV who want the deepest server-side Shopify integration available and have in-house or agency GTM expertise. Value: 7/10. Starts at $200/month.

---

Stape

Stape is server-side GTM hosting infrastructure, not a finished CAPI product. The distinction matters. You bring your own GTM container, your own tags, your own data layer configuration. Stape provides the server at $17/month Pro, and then you configure everything yourself. The 80+ template library helps, but GTM expertise is required. For a team that has a dedicated tagging engineer or an agency that lives in GTM, Stape is the lowest-cost path to server-side infrastructure. For a founder or performance marketer who wants a working CAPI without building it, Stape is three to six months of configuration work that looks like a $17/month subscription until it isn't.

No bot filtering. No built-in CMP. The Bounteous research that found 80% of server-side GTM setups detectable to ad blockers applies to Stape deployments where the GTM server domain hasn't been customized correctly: first-party CNAME setup is possible but requires the configuration work most Stape users don't complete.

Right for: In-house GTM engineers and technical agencies who want full container control and maximum flexibility. Value: 8/10 for that buyer. $17/month Pro plus Cloud Run infrastructure costs $50-300/month depending on volume.

---

Tracklution

Tracklution is a clean, EU-leaning server-side CAPI tool with SOC 2 and ISO 27001 certifications, simple setup, and coverage across Meta, Google, TikTok, and Pinterest. The compliance credentials are real and relevant for EU agencies. No GTM required. The weaknesses are consistent with the category: no bot filtering, so bot-generated cart events flow through to Meta, and no built-in CMP. For a European agency managing multiple client accounts and needing documented compliance certifications in client contracts, Tracklution has a credible argument. At €31/month Starter, the price is accessible.

Right for: EU agencies and brands where ISO 27001 and SOC 2 certifications are contractual requirements. Value: 7/10. €31/month Starter.

---

Littledata

Littledata built its reputation on Shopify-to-GA4 accuracy and has since layered Meta CAPI and server-side routing on top of that foundation. The genuine differentiator is subscription and recurring revenue tracking through ReCharge integration, which handles recurring order events more cleanly than most CAPI tools. The Klaviyo integration stitches marketing channel attribution onto email profiles and backfills historical events, which is useful for recovery audience quality.

The pricing is the problem. Littledata scales by order volume and connector, and the cost curve at meaningful order volumes lands in the same range as Elevar. Like Elevar, no bot filtering, and the focus remains Shopify. No built-in CMP.

Right for: Shopify subscription and DTC brands with meaningful recurring revenue who want GA4 accuracy as the primary goal and server-side CAPI as a secondary benefit. Value: 6/10. $199/month Standard and up.

---

SignalBridge

SignalBridge enters the comparison as the lowest-cost path to server-side CAPI with bot filtering included, which distinguishes it from most of the category. At $29/month, it covers Shopify, WooCommerce, and other platforms with bot filtering using VPN detection and data center IP blocking. The bot filtering is not at the scale of DataCops' 361B-IP database, but it is present, which puts SignalBridge ahead of Elevar, Stape, Tracklution, and Littledata on this specific dimension.

The caveats: newer brand, less mature than Elevar or Stape for enterprise use cases, narrower integration catalog. But for a founder-run store that wants bot filtering in the CAPI pipeline at $29/month without an enterprise commitment, SignalBridge is worth evaluating.

Right for: SMBs wanting bot-aware CAPI on a tight budget across platforms. Value: 7/10. $29/month.

---

Aimerce

Aimerce positions as a no-code server-side tracking platform for Shopify and Shopify Plus, with a focus on recovering ad-blocker-lost events and improving Meta EMQ scores. The setup experience is cleaner than Stape for non-technical users. No bot filtering. No built-in CMP. Shopify-only. At $299/month base plus usage-based pricing above 1,000 orders, the cost escalates quickly for growing stores. Multiple user reviews on Trustpilot note the pricing curve as a concern.

Right for: Shopify Plus brands with a dedicated analytics person who want managed server-side without GTM complexity. Value: 6/10. $299/month base.

---

Analyzify

Analyzify is a "done-for-you" GTM and data layer setup service for Shopify, not a monthly SaaS subscription. You pay for expert GTM configuration rather than infrastructure. For a Shopify brand that tried Elevar, found the setup frustrating, and concluded the problem was their GTM container rather than the concept of server-side tracking, Analyzify makes sense. The one-time fee model is distinctive and honest: you pay once for correct configuration rather than a monthly subscription for infrastructure you may not be fully using.

No ongoing bot filtering. No built-in CMP. The GTM container Analyzify sets up for you still depends on client-side event triggers for CAPI routing. What it fixes is configuration quality, not signal quality.

Right for: Shopify brands that want a correct GTM and data layer setup without learning GTM themselves, where the problem is configuration errors rather than missing bot filtering. Value: 7/10. One-time fee model, contact for pricing.

---

Meta 1-Click CAPI

Meta launched its free, one-click CAPI integration on April 15, 2026, and the floor for Meta-only CAPI is now $0. For a single-platform Shopify store running only Meta ads and not needing Google, TikTok, or LinkedIn CAPI, the 1-click native integration does the job. The EMQ improvement is real. The setup is genuinely one click.

The limits are equally real: Meta only, no bot filtering, no built-in CMP, no multi-platform routing. Every other channel you run requires a separate solution. And the bot problem remains entirely unaddressed: bot AddToCart events flow through the native 1-click integration the same as they flow through any other pixel-only setup. Meta builds its optimization model on whatever you send, including automated traffic.

Right for: Single-store basic Meta-only advertisers who don't need multi-platform and aren't concerned about bot signal contamination. Value: 10/10 for that buyer (it's free).

---

Google Tag Gateway

Google launched the Tag Gateway in January 2026, providing free Google-only server-side routing via GCP, Cloudflare, or Akamai. Same category position as Meta 1-click: free, Google-only, no bot filtering, no multi-platform, no consent management. For Google Enhanced Conversions on a budget, it is a viable starting point.

Right for: Brands with meaningful Google Ads spend who want server-side Enhanced Conversions without paying for infrastructure. Value: 10/10 for that buyer. Free.

---

Triple Whale

Triple Whale is an attribution dashboard and reporting platform that includes server-side tracking as a component. It is not primarily a CAPI infrastructure tool. The Pixel is the foundation, the attribution models are the value, and the server-side layer improves the accuracy feeding those models. At $179/month annual, you are buying reporting and attribution modeling. The CAPI layer you get is real but secondary to the dashboard. No bot filtering. No built-in CMP.

The common Shopify stack at seven-figure GMV is Elevar for data layer plus Triple Whale for reporting. They serve different jobs. Triple Whale on its own without a clean server-side foundation is beautiful charts built on the same broken pipe as GA4.

Right for: DTC brands that want a unified attribution dashboard and can accept that the server-side tracking bundled in is less mature than dedicated infrastructure tools. Value: 7/10. $179/month annual.

---

Northbeam

Northbeam is media mix modeling and attribution for brands spending $50K+ per month on ads. The analysis is genuinely sophisticated: multi-touch modeling, incrementality testing, channel-level budget recommendations. At $1,500/month entry, it is priced for that buyer. Like Triple Whale, the tracking quality feeding Northbeam's models determines the quality of its recommendations, and Northbeam does not filter the events it ingests.

Right for: Performance marketing teams at large spenders who want incrementality testing and media mix modeling. Not a CAPI infrastructure tool. Value: 7/10 for that buyer. $1,500/month.

---

The recovery layer

These tools act on abandonment events. Their quality ceiling is set entirely by the tracking infrastructure above.

Klaviyo

Klaviyo is the foundation of most ecommerce recovery stacks, and that position is earned. The segmentation is genuinely powerful: 500+ behavioral and demographic data points, predictive CLV modeling, churn risk scoring, and AI-optimized send timing. The abandoned cart flow built on Klaviyo, properly configured with 1h/24h/72h sequences and dynamic product blocks, outperforms most alternatives for email recovery.

The problem isn't Klaviyo. The problem is what feeds it. Klaviyo triggers on whatever InitiateCheckout events your pixel or CAPI sends. If your pipeline doesn't filter bots before those events fire, Klaviyo receives bot-generated email addresses and triggers recovery flows to them. Those emails bounce. Your sender score degrades. Your deliverability on real human abandonment emails goes with it. Klaviyo cannot fix upstream data quality; it reflects it.

Right for: Any ecommerce brand doing serious email marketing. The question is what you've cleaned before the events reach it. Starts at $45/month for 15,000 emails.

---

Attentive

Attentive is the enterprise SMS and email platform with deep personalization and custom pricing. The AI send-time optimization and behavioral trigger logic are mature. The honest constraint is that Attentive's personalization is only as good as the identity resolution underneath. For EU traffic where consent was never captured because your CMP loaded from a third-party CDN and got blocked by uBlock, Attentive has no identity to personalize against. The signal gaps upstream cascade into personalization gaps downstream.

Right for: Enterprise and fast-growing brands wanting SMS at scale with advanced personalization. Custom pricing.

---

Postscript

Postscript is built specifically for Shopify SMS. Deep native Shopify integration, clean compliance tooling, good deliverability, and solid abandonment flow automation. Two-way SMS enables conversational recovery, which works well for higher-AOV products where a human-feeling message outperforms a discount blast. The per-message cost model is transparent. The Shopify-only constraint is the limit.

Right for: Shopify stores that want SMS recovery as a serious channel alongside email. $49/month plus message costs.

---

Omnisend

Omnisend is the all-in-one multi-channel marketing automation platform: email, SMS, push notifications, and Messenger in one flow builder. The workflow logic handles email plus SMS plus push in sequence without managing three separate tools. For brands that want multiple recovery touchpoints without multiple platforms, Omnisend is frequently the most operationally efficient choice. The segmentation is less granular than Klaviyo. The deliverability track record is solid.

Right for: SMBs wanting multi-channel recovery without the complexity of separate tools. Starts at $16/month.

---

Recart

Recart focuses on SMS and Facebook Messenger recovery. The Messenger channel is differentiated: most competitors have moved away from Messenger as a recovery channel, but for certain audiences it still converts well. The SMS layer is less mature than Postscript or Attentive. Shopify-focused.

Right for: Shopify brands with an audience that engages with Messenger and wants to add it as a recovery channel. Custom pricing.

---

CartStack

CartStack is an independent cart recovery platform covering email and SMS with a particular focus on anonymous abandonment recovery through exit-intent capture and partial email address recognition. The identity resolution component is the differentiator: it attempts to identify abandoners who didn't reach checkout by matching behavioral signals. The data quality question applies here in amplified form: if anonymous session identification captures bot sessions, the "anonymous abandoner" audience CartStack builds includes automated traffic.

Right for: Stores that want to extend recovery beyond identified email abandoners to the anonymous visitor pool. Custom pricing, $29/month and up.

---

Rejoiner

Rejoiner is a focused cart recovery email platform with a strong track record in high-volume enterprise ecommerce. Partial checkout submission capture, high-intent segmentation, and hands-on support are the differentiators. Less self-serve than Klaviyo, more agency-style engagement. The recovery rates cited are solid. Same upstream data quality dependency as every recovery tool in this list.

Right for: Mid-market to enterprise brands wanting a managed cart recovery email program rather than a self-serve platform. Custom pricing.

---

Privy

Privy combines on-site conversion tools (exit-intent popups, spin-to-win, email capture) with basic email and SMS recovery flows. The on-site capture layer is the genuine value: capturing email addresses before checkout starts puts identified abandoners into recovery sequences even when a user never reaches the checkout page. This is the correct architectural instinct, because a well-run email capture funnel reduces the percentage of abandoners who are anonymous. The email platform component is less sophisticated than Klaviyo for complex segmentation.

Right for: SMBs that want on-site conversion capture plus basic email recovery in one tool. Starts at $30/month.

---

PushOwl

PushOwl is the leading browser push notification tool for Shopify cart recovery. The channel is genuinely differentiated: push notifications reach abandoners who gave no email and no phone number, only browser permission. No contact detail required at all. The recovery rate ceiling is lower than email, but the addressable pool is larger for stores that haven't captured email from all visitors. The bot signal question is less acute here because push requires an explicit browser permission grant, which bots do not typically complete.

Right for: Shopify stores wanting to reach anonymous abandoners without contact capture. Free plan available, paid from $19/month.

---

The buyer matrix

For stores on multiple platforms or with significant non-Shopify infrastructure, the choice of tracking infrastructure matters more than the choice of recovery tool. A great Klaviyo setup on a broken CAPI produces lower recovery rates and worse Meta performance than a basic Klaviyo setup on clean, bot-filtered events.

For Shopify-only stores under $500K GMV where Shopify's native checkout data is reliable and Meta is the primary ad channel, the Meta 1-click CAPI plus Klaviyo is the lowest-friction starting stack. The bot filtering gap exists but may not be worth the incremental cost at that scale.

For stores doing $500K to $5M GMV across multiple ad platforms, the multi-platform CAPI question becomes material. Elevar at $200-950/month is real money. DataCops at $49/month with Meta, Google, TikTok, and LinkedIn included starts to look like a different conversation, particularly when the bot filtering is built in at the pipeline level rather than treated as a separate project.

For stores above $5M GMV with in-house engineering, Stape plus a custom bot filtering layer is viable. The infrastructure cost is justifiable. The configuration requirement is manageable with dedicated resources. The total cost of ownership calculation includes engineering time.

---

When NOT to use DataCops

If you are on Shopify only and need Pinterest CAPI, DataCops doesn't offer it. Elevar or TrackBee do.

If you need SOC 2 Type II certification today for a compliance requirement, Tracklution has it. DataCops is still in progress.

If you are an agency or in-house engineer who wants full GTM container control and prefers to build your own tracking architecture, Stape is the right infrastructure choice. DataCops is a managed outcome, not a toolkit.

If your primary problem is attribution modeling and media mix reporting rather than data pipeline quality, you need Triple Whale, Northbeam, or Hyros. DataCops cleans the pipe. It does not build the dashboard.

If your store is genuinely early-stage, under 2,000 sessions per month, and you have no paid ads running, the free tier of DataCops covers you, but so does the Meta 1-click CAPI, and the operational overhead of a full conversion infrastructure setup is premature.

---

The question worth asking

The industry spent the last four years building better pipes. CAPI adoption is real. Server-side tracking is real. The EMQ scores are real. Attribution has measurably improved.

Nobody built a filter for the water.

The abandoned cart retargeting audiences you built last month, the Lookalike Audiences Meta trained on your highest-intent cart events this quarter, the recovery email sequences that fire on every InitiateCheckout in your flow: how much of the signal feeding all of that came from a human being who actually looked at your product and wanted to buy it?

If you don't have an answer to that question with a number attached, you're not optimizing a recovery program. You're optimizing a machine learning model to find more traffic that looks like whatever your pixel counted.

What's in your CAPI event stream right now, and how many of those AddToCart events can you prove were real people?

---

Related reading: API-to-API Conversion Tracking Setup covers the technical implementation layer. Best Click Fraud Protection Tools 2026 addresses the bot problem at the traffic layer. AI + Meta CAPI: The 2026 Conversion Stack covers the intersection of AI optimization and signal quality. B2B Conversion Tracking Best Practices extends the same framework to lead-gen. Fraud Traffic Validation is where DataCops' bot filtering architecture is documented in detail.