Sales Automation 2026: Practical Stack for DACH

Sales Automation · 18. September 2026 · Mohsen Ghulami

Sales Automation 2026 for DACH manufacturers: Choose your CRM, data, and AI SDR stack correctly, calculate ROI, and avoid GDPR pitfalls.

In the hotel lobby at Stuttgart Airport, the smell of wet coats and carpet cleaner still lingered in the air. Martin, CEO of a component manufacturer from Heilbronn, placed his iPad on the small glass table and said: "We have HubSpot, Apollo, and some LinkedIn tool. Yet, no one knows which leads are good." On the display, 1,842 contacts blinked, 317 of them without an industry, 96 with private Gmail addresses, and surprisingly many "Head of Something" from companies that didn't even fit his target segment. I was silent for a moment because the answer wasn't pleasant. Sales Automation 2026 rarely fails due to tool price – it fails due to broken handovers between data, sequence, and CRM.

Sales Automation 2026: Why Your Stack is Now Costing Revenue

When medium-sized manufacturing companies don't set up sales automation cleanly, a catastrophe usually doesn't happen immediately. Something worse happens. The pipeline looks fuller, activity numbers rise, sales managers see more sent emails, more LinkedIn touches, more automatically generated tasks – and yet, at the end of the month, no reliable forecast emerges. At a mechanical engineering company in East Westphalia, I saw in March 2025 how 14 percent of the supposedly new target accounts were already customers. Some of them were handled through spare parts service, others through a dealer in Austria. The system didn't know. Sales only realized it when an existing customer wrote back: "Why are you treating us like a stranger?" This isn't an automation problem. This is a revenue operations problem with an automation accelerator.

The second damage is more expensive but more invisible: your best sales team loses faith in the machine. And once field sales reps believe that "the leads from the system" are junk, it's hard to win them back. At Brose, Schaeffler, Festo, or Phoenix Contact, every good salesperson knows that a plant, a line, a standard, a spare parts event, or an OEM program matters more than a generic job title. Anyone who still believes in 2026 that you can simply dump 10,000 contacts into a sequence in industrial sales and the pipeline will somehow work out, is confusing reach with relevance. Honestly? That's the most expensive confusion I'm seeing in B2B sales right now.

The tool landscape has shifted. Teams no longer compare "automation" versus "no automation." They decide between three stacks: all-in-one CRMs like HubSpot, Salesforce, and Microsoft Dynamics 365; prospecting and data engines like Apollo, Clay, Cognism, ZoomInfo, or Lusha; plus sequencing and AI SDR tools like Reply.io, lemlist, Instantly, Smartlead, Salesloft, Outreach, 11x, or Salesforge. According to Salesforce State of Sales 2026, cited by KNK Outbound among others, 54 percent of salespeople already use AI Agents, and almost 90 percent expect to use them by 2027. Sounds fast. And it is. But speed without data hygiene is just noise with a dashboard.

What this Practical Guide for Sales Automation 2026 Explains

I'm writing this from my work as a GTM Engineer at Amplifa, not from an analyst's box. In implementations for B2B teams in DACH, I see the same questions repeatedly: HubSpot or Salesforce? Apollo or Cognism? Clay just a toy for growth people or serious for industrial prospecting? Buy AI SDR or build sequences with human approval? And the most important question, which almost no one asks at the beginning: What does a qualified meeting really cost when credits, data verification, deliverability, CRM cleanup, and human control are factored in?

The guide walks through five steps that I almost always follow in this order in projects:

• Step 1: Define ICP and account logic so that sales automation doesn't chase the wrong companies.
• Step 2: Place data sources, enrichment, and verification before the sequence – not after.
• Step 3: Clearly separate CRM, sequencing, and AI execution to keep costs and responsibilities clear.
• Step 4: Build multichannel outbound with stop rules, GDPR process, and human-in-the-loop.
• Step 5: Measure ROI not by seat price, but by cost per qualified meeting and pipeline per rep.

Step 1: Sales Automation 2026 Starts with ICP, Not the Tool

The most common mistake looks harmless: a sales manager exports 5,000 companies from a database, filters for "manufacturing," "Germany," "50-500 employees," and calls it ICP. Well, almost. For software, that might sometimes be enough. In manufacturing, it's too crude. A manufacturer of seals for hydraulic applications sells differently than a provider of testing technology for battery production. Both can have 180 employees, both are located in Baden-Württemberg, both appear in databases under similar industry labels. Nevertheless, they need different triggers, different contacts, different timing signals.

Therefore, I start with an account matrix. Not with personas. Accounts first. For a customer in automation technology in Nuremberg, in April 2025, we divided target companies into four clusters: OEM machine builders, special machine builders, system integrators, and maintenance organizations of large plants. Each cluster had different buying events. For OEMs, new product lines and international rollouts were relevant. For maintenance, downtimes, retrofit programs, and spare parts availability were important. This sounds like diligent work, but it's the point where sales automation either builds pipeline or produces spam.

A concrete setup rule: I don't let any tool automatically contact prospects before the account class is in the CRM. In HubSpot, this can be done via a mandatory field like "ICP Cluster," in Salesforce via an account field plus a validation rule, in Dynamics 365 via segment logic and business rules. For smaller teams, a custom field in Pipedrive is often sufficient, as long as someone owns it. Elegance is not important. What's important is that no sequence starts if the account is not assigned. Otherwise, you'll send texts about retrofits to an OEM buyer who is currently negotiating series prices. That's not embarrassing. That's pipeline-damaging.

"That doesn't work for us," Andrea, Head of Sales at a hidden champion in Bielefeld, told me when I suggested classifying accounts before contact research. Two weeks later, her team showed 38 percent less manual lead sorting in the CRM. Not because the AI got smarter. But because it received less nonsense.

— Mohsen Ghulami, GTM Engineer at Amplifa

Concrete example: ICP fields for a machine builder

For a medium-sized machine builder, 220 employees, export quota over 60 percent, I wouldn't build a setup that only filters by industry and revenue. I would create fields like "Machine type in target plant," "Role in buying center," "Dealer relationship known," "Installed base presumed," "Service upsell possible," and "Trigger source." The trigger source could be an exhibitor list from Hannover Messe 2025, a new Webasto plant, a job posting pattern for PLC technicians, or a LinkedIn post from a production manager about capacity expansion. Yes, that's granular. That's precisely why it works.

The business impact is simple: If a qualified meeting in industrial sales can generate 8,000 to 25,000 euros in weighted pipeline, then a bad data record is not just a wrong email address. It's a lost slot in a salesperson's calendar, who could have called a real target account instead. For a customer near Karlsruhe, we only released 1,180 accounts from 12,400 raw companies for the first outbound wave. The CEO initially found the reduction brutal. Afterwards, 41 qualified initial meetings came from the smaller list in nine weeks. From the large list, there might have been more responses. But responses don't pay salaries.

Step 2: Data Quality Before AI SDR – Always

Tool providers like to tell the story of the AI SDR that builds lists, finds contacts, writes emails, classifies responses, and books appointments. Part of that is true. Another part only exists in demos. The practical order remains clear in 2026: data source, enrichment, verification, segmentation, sequence. Only then AI execution. Anyone who reverses this pays credits to email the wrong people with pretty sentences.

Apollo, Clay, Cognism, ZoomInfo, and Lusha play different roles here. Apollo is often fast for international contact research and simple sequence proximity. Clay is strong if you want to combine multiple data sources, web scraping, AI classification, and custom logic. Cognism is often chosen in Europe because it strongly positions itself on GDPR-compliant EMEA data and phone-verified contacts; prices, according to several market comparisons, are often in the five-figure annual range. ZoomInfo is powerful in enterprise environments but not cheap. Lusha is an easy entry for some teams. None of these sources are magic. All need rules.

My standard workflow for manufacturing companies looks like this: Take an account from CRM or target account list, normalize the domain, enrich company data, check location, search for relevant roles, verify email, use phone number only with sufficient data origin, check for duplicates against CRM, match against suppression list, then only allow sequence. In Clay, you can build this with table logic, HTTP calls, Clearbit-like company data, Apollo search, and AI columns. In HubSpot or Salesforce, only contacts that meet certain thresholds end up. Below 70 out of 100 points? No automatic outreach. Period.

What we specifically see at Amplifa: In the last 12 months, industrial customers with 50 to 500 employees almost never had a pure tool problem, but rather a handover problem between database and CRM. In 7 out of 10 audits, we found duplicate rates between 9 and 22 percent in target segments before any sequence was even started. Particularly often: old dealer contacts, former employees, incorrectly mapped subsidiaries, and contacts from trade fair scans without opt-out status. As soon as we implemented a verification block with domain match, company status, role fit, and suppression matching before the sequence, manual lead rejection decreased by about a third for several teams. This isn't in any vendor demo deck because it doesn't look sexy. But it saves pipeline.

Combining data sources correctly: Apollo, Clay, Cognism

A setup that I often find useful for DACH industrial sales: Cognism or a comparable EMEA-strong source for verified business contacts, Clay as an orchestration layer for enrichment and logic, HubSpot or Salesforce as the system of record, plus Reply.io, Salesloft, or Outreach for cadences. Smaller teams can use Apollo plus HubSpot plus lemlist or Smartlead if they master deliverability cleanly. But beware: Instantly and Smartlead are attractive because they are cheap and high-volume – Instantly is shown in comparisons at around $47 per month, Smartlead at around $39. That doesn't mean they fit every industrial sales operation. Cheaply sent wrong emails remain wrong emails.

For a company like Kärcher, Trumpf, or DMG Mori, I would naturally design differently than for a 90-employee supplier from Sauerland. Large organizations need governance, role rights, approval flows, data residency issues, CRM write rules. Smaller companies need speed and clarity. But both need the same fundamental question: Which data can automatically trigger decisions? If the answer is "all," that's not courage. That's laziness with an API key.

Step 3: Separate CRM Automation and AI Execution

In 2026, pricing models are splitting. This is more than just purchasing policy. HubSpot Sales Hub appears in comparisons starting at around $10 per seat per month, Pipedrive Growth at around 39 euros per seat, Salesforce Sales Cloud at entry level around $25 per seat. Microsoft Dynamics 365 Sales, according to Microsoft price lists, is $65 for Sales Professional, $105 for Sales Enterprise, and $150 for Sales Premium. In real enterprise setups, however, you quickly go higher once AI features, add-ons, data volume, integrations, and consulting are added.

On the other side are AI and execution tools with credits, actions, leads, or agent minutes. Reply.io is in market comparisons at $89 to $99 per user per month for multichannel outreach. Clay ranges from free tiers to about $167 to $185 per month, depending on the plan and source. HubSpot Breeze is described in one comparison at $1 per lead. Salesforce Agentforce works with Flex Credits, about $500 per 100,000 credits plus licensing. 11x, according to market overviews, starts at about $3,750 per month, Julian is quoted at around $2,417 per month for chat or $5,333 for voice. Suddenly, the seat price is no longer decisive, but usage. How many leads are enriched? How many actions does an agent perform? How many faulty contacts burn credits?

That's why I separate three responsibilities in projects: The CRM holds truth and history. The data engine researches and evaluates. The engagement layer executes sequences and classifies reactions. If everything is in one tool, it feels convenient at first. Later, no one knows why a contact was created, which source was used, whether an opt-out exists, and which AI text was actually sent. With a Dynamics 365 team in Munich, we saw exactly that: 23 workflows, 11 Power Automate flows, two external enrichment tools, and no central error list. Sales called it "automation." I called it fog.

The clean architecture looks more boring. CRM as the system of record. Clay, Apollo, or Cognism as the data and enrichment layer. Salesloft, Outreach, Reply.io, lemlist, Instantly, or Smartlead as the sequencing layer. Latenode, Make, Zapier, or native APIs for orchestration when error handling, duplicate logic, and API rates become important. Gong can feed back conversation data and objections if the team has enough call volume. Important: Not every signal should automatically write. Some signals should only generate a task.

Steps 4 and 5: Advanced Sales Automation Workflows

1. Build an account scoring logic with hard exclusions. Example: A target account gets points for matching industry, employee count 50-500, DACH location, recognizable production, current investment signals, and matching roles in the buying center. But it is immediately blocked if it is an existing customer, active dealer, ongoing opportunity, or suppression hit. For a supplier from Ulm, in May 2025, we placed exactly these blocks before the sequence; the number of embarrassing existing customer contacts dropped to zero.
2. Use AI for research summaries, not for unverified claims. A good prompt structure: Summarize only publicly verifiable information from the website, career section, and press section; state uncertainties; do not invent machines, plants, or projects. For a target account like Wittenstein, the AI can structure roles, business areas, and potential automation topics. But it must not claim that a specific plant is currently replacing a line if there is no source for it.
3. Separate sequence types by industry context. A service upsell sequence to an installed base needs different stop rules than cold outreach to new OEM accounts. Spare parts topics can often be driven more by urgency, new machine business by efficiency, standards, capacity, or quality. A buyer at an automotive supplier reacts differently than a maintenance manager in a plant with shift operations. Sounds trivial. But it's ignored in 80 percent of bad sequences.
4. Deploy human-in-the-loop where relevance trumps volume. For Tier 1 target accounts, large plants, or strategic OEMs, I have humans approve AI drafts. For smaller, well-classified segments, you can automate more heavily. A CSO from Nuremberg recently told me: "My people shouldn't be typing texts, but deciding if an account is worth the effort." That's exactly the right shift.
5. Measure true funnel metrics. Open rates in 2026 are almost worthless because Apple Mail Privacy, security gateways, and bot scans distort the number. Instead, measure verified delivery, positive response rate, qualified meetings, meeting-to-opportunity rate, pipeline per sequence, and cost per qualified meeting. For long sales cycles in mechanical engineering, you should also look at cohorts over 90 to 180 days, not just two weeks after campaign launch.

An advanced workflow I like: Trigger-based outbound waves for service and spare parts. The starting point is not an external contact list, but the installed base in the CRM or ERP. If a machine is older than five years, certain wear parts have been regularly ordered, or a service contract is expiring, the system generates an account task. AI summarizes history and possible next steps. The responsible sales or service employee decides whether an email, a call, or a partner hint makes sense. For manufacturing companies, this is often more profitable than cold new customer acquisition because trust and need already exist.

Another workflow: Distributor and channel management. Many DACH manufacturers work with dealers in France, Italy, Poland, or the USA. The CRM shows revenue but rarely market potential. Here, sales automation can compare dealer territories with company registers, website signals, trade fair exhibitor lists, and existing opportunity data. The output is not an automatic mass email, but a partner list: Which dealers are not covering which segments? Where are there target accounts without a touchpoint since 2022? Which Webasto, Schaeffler, or Bosch suppliers appear in the territory but not in the pipeline? This is automation that interests sales management.

GDPR: Why AI Outbound Isn't the Real Problem

European teams often ask: "Is the AI even allowed to do that?" The better question is: "Are we allowed to process this personal data for this purpose and contact this person in this way?" AI-generated text is not automatically the compliance problem. The real issues are legal basis, legitimate interest, data origin, purpose limitation, transparency, storage periods, and suppression lists. In B2B outreach in DACH, there are different interpretations depending on the country and channel, and I am not a lawyer. Not entirely true: I am very happy to be the person who builds the process so that the lawyer blushes less.

My conservative operating model: Only business contacts, narrow targeting, documented legitimate interest assessment, clear sender identity, easy opt-out, immediate suppression across all systems, and no re-enrichment of unsubscribed individuals. If Cognism is chosen for EMEA and phone verification, that's one component. Not a free pass. If Apollo or Clay are used, data sources and verification must be cleanly documented. And if an AI Agent classifies responses, an opt-out must be weighted more heavily as a stop signal than any score. No "maybe later." Stop means stop.

In a setup for a manufacturer from Bavaria, in June 2025, we built a central suppression table outside the sequencing tool because three systems could generate contacts: HubSpot, Clay, and Reply.io. Every opt-out first wrote to this table, then back to HubSpot, then to Reply.io, then to Clay as an exclusion before future research. Why so complicated? Because a single error is enough to damage trust. A production manager who unsubscribes and is contacted again two weeks later doesn't think: "The API had latency." He thinks: "They don't have their sales under control."

Amplifa Sales Audit Check data quality, CRM logic, outbound processes, and automation potential before buying new sales automation tools.

How to Calculate the ROI of Sales Automation 2026

I don't like ROI calculations that start with "We save two hours a week." In sales, saved time alone doesn't count. What counts is whether better accounts get into real conversations faster and whether salespeople spend less time on research, copy-pasting, and CRM maintenance. The right calculation starts with cost per qualified meeting. Then meeting-to-opportunity. Then weighted pipeline. Then close rate and sales cycle. Only then tool costs.

A simple example from mechanical engineering: 6 sales employees, 2,000 target accounts, average deal 80,000 euros, close rate from qualified opportunity 22 percent, gross margin 35 percent. If a clean stack generates 18 additional qualified meetings per month, resulting in 6 opportunities and ultimately 1.3 deals won, we're talking about 104,000 euros in revenue and 36,400 euros in gross margin. If the stack, including CRM add-ons, Clay, data source, Reply.io, and implementation, comes to 7,000 euros in monthly full costs, that's interesting. But if those same 18 meetings are not qualified, the calculation is garbage. Then you've only bought calendar fillers.

That's why I have teams separate three cost categories: platform costs, usage costs, and process costs. Platform costs are seats and licenses. Usage costs are credits, leads, AI actions, mailboxes, data retrievals. Process costs are human reviews, data maintenance, deliverability, legal review, and CRM administration. Especially in 2026, budgets are shifting from seat to usage. Salesforce Agentforce with Flex Credits, HubSpot Breeze per lead, Clay credits, or AI SDR monthly fees make volume expensive if data is poor. The cheapest lead is the one you don't even enrich because the account doesn't fit.

Benchmark Logic for Medium-Sized Manufacturers

For companies with 50 to 500 employees in DACH, I like to set up a 90-day test logic. Not 14 days. Industrial sales needs time. In the first 30 days, we measure data quality, deliverability, and positive response patterns. In the next 30 days, we measure qualified meetings and objections. In the last 30 days, we check opportunity creation, sales acceptance, and follow-up discipline. If after 90 days only open rates are celebrated, I mentally stop the experiment. Opens are confetti. Pipeline is work.

A realistic target value depends heavily on the segment. For narrow account-based prospecting campaigns in DACH industry, I prefer 3 to 8 percent positive response rate on small, well-researched segments over 0.7 percent on huge lists. For existing customers or service upsells, the meeting rate can be significantly higher if data from the installed base is accurate. For completely cold international target segments, it decreases. And yes, some months are ugly. If all target accounts have budget freezes or the industry is struggling, no prompt in the world will save your forecast.

Which Tools Fit Which Sales Model?

For a CEO, the tool question is often the most visible. For me, it's the last. Nevertheless, guidance is needed. HubSpot often works well for medium-sized teams that want to start quickly, bring marketing and sales closer together, and don't immediately have an enterprise RevOps team. Salesforce fits when complex roles, regions, products, partners, forecast logic, and integrations need to be cleanly mapped. Microsoft Dynamics 365 is obvious if the company is already deeply embedded in Microsoft environments, ERP proximity, and Power Platform. Pipedrive can be sufficient for smaller, direct sales teams if processes are clear and integrations are manageable.

For engagement tools, I differentiate strictly by maturity level. Salesloft and Outreach are strong for structured, larger sales teams with coaching, governance, and multiple roles. Reply.io is often a pragmatic multichannel middle ground. lemlist can work well for personalized email and LinkedIn-focused campaigns. Instantly and Smartlead are popular for mailbox and volume setups but require particularly clean deliverability governance. Salesforge and 11x lean more towards AI execution. I would never evaluate them in isolation. Always with data source, CRM write rules, and human control.

For a company that sells components to machine builders, my starting stack would often be: HubSpot or Salesforce as CRM, Cognism or Apollo for contacts, Clay for enrichment and scoring, Reply.io or Salesloft for sequences, Gong only when enough conversations are happening and coaching is a bottleneck. For a service-heavy manufacturer with an installed base, I would invest less in cold data and more in CRM/ERP signals, lifecycle triggers, and account plans. For a dealer sales force, I would first clarify channel data and partner performance. Tool names don't impress buying committees. Relevant occasions do.

Amplifa Product Amplifa connects GTM workflows, data logic, and AI-powered sales processes for B2B teams that don't just want to send more emails.

Sales Automation 2026 for Account-Based Prospecting

Account-Based Prospecting in industrial sales is not a buzzword, but almost always the most sensible mode of operation. Large buying committees, long sales cycles, multiple plants, technical specifications, supplier approvals – generic sequences rarely win here. A production manager is interested in downtime and throughput. A purchasing manager in risk and total cost. A development manager in specifications, standards, and feasibility. A CEO in market opportunity and capacity. If everyone receives the same AI text, you've automated. But you haven't sold.

The workflow: Select target account, understand plant or location, identify roles in the buying center, search for triggers, formulate individual hypothesis, start sequence, classify reaction, update CRM. AI helps with research and drafting. The human verifies the hypothesis. Example: A supplier wants to land with manufacturers of packaging machines. Clay finds companies, Apollo or Cognism adds contacts, the AI reads career pages for terms like "PLC," "servo," "retrofit," "OEE," or "CE conformity," HubSpot stores clusters and triggers, Reply.io starts a sequence with a maximum of four touches over 21 days. No daily nagging. No "I just wanted to follow up" after 48 hours. Adults don't buy complex industrial goods because a bot is persistent.

A good first email is short and specific. Not obnoxiously personalized. I don't want to read a line like "Congratulations on your inspiring LinkedIn post" if the post is three years old. Better: "I saw that you are currently looking for mechatronics engineers for commissioning at your Crailsheim site. We are currently speaking with several special machine builders who want to reduce service deployments because commissioning capacity is scarce." That is verifiable. That is relevant. And if it doesn't fit, the recipient can cleanly object.

How Much Automation is Too Much in Industrial Sales?

More automation is not automatically better. I know that sells worse. But in DACH mid-sized companies, trust is a factor of production. If a sales manager from Augsburg tells me his best salespeople know many target customers from trade fair talks at Motek 2019 or Hannover Messe 2023, then I don't build a machine that overwrites these relationships with generic follow-ups. I build a machine that creates memory: last touchpoint, topic, product interest, plant, partner, next sensible occasion.

You recognize too much automation by four symptoms. First: Sales no longer reads AI texts before they go out. Second: Opt-outs land in one tool, but not in all. Third: The CRM fills up faster, but forecast quality doesn't increase. Fourth: People discuss open rates because no one wants to talk about opportunities. If two of these symptoms appear, stop new sequences and check the process. Don't buy another tool. Really not.

FAQ: What Sales Managers Ask About Sales Automation 2026

Do we need an AI SDR in 2026?

Maybe. If you have clean target segments, good data sources, clear stop rules, CRM governance, and enough volume, an AI SDR can accelerate research, drafting, routing, and follow-ups. If your ICP is unclear and the CRM contains duplicates, an AI SDR will only amplify chaos. For many manufacturing companies, a human-in-the-loop setup with Clay, Cognism or Apollo, a sequencing tool, and clean CRM writeback is the better first step.

Is HubSpot, Salesforce, or Dynamics 365 better for Sales Automation?

That depends less on the logo than on process maturity and IT landscape. HubSpot is often faster for medium-sized teams, Salesforce stronger for complex enterprise governance, Dynamics 365 makes sense with deep Microsoft integration. Crucially, the CRM must not become a dumping ground for automatically generated contacts. Mandatory fields, duplicate checking, role rights, and clear lifecycle stages are more important than the prettiest AI feature.

How do we remain GDPR compliant with AI outreach?

Work with narrow B2B target groups, documented legal basis, clear sender identity, easy unsubscribe, verified data origin, and central suppression lists. Check country-specific rules with legal advice. AI is not automatically forbidden, but it does not exempt you from data protection obligations. Particularly critical are unclear data sources, missing opt-out synchronization, and automated re-engagement of unsubscribed contacts.

Amplifa GTM Workflow Check Analyze where your sales automation is currently breaking: ICP, data quality, sequences, CRM writeback, reporting, or GDPR process.

Summary: Three Takeaways for Your 2026 Stack

1. Data quality beats tool hype. Apollo, Clay, Cognism, ZoomInfo, or Lusha are only valuable if ICP logic, verification, duplicate checking, and suppression matching happen before the sequence.
2. CRM automation and AI execution must be considered separately. HubSpot, Salesforce, Dynamics 365, or Pipedrive hold truth and pipeline; AI and sequencing tools execute actions. Do not blindly mix them.
3. ROI comes from qualified meetings and pipeline, not from sent emails. Factor in credits, actions, leads, human reviews, and data maintenance. In 2026, the team with the cleanest relevance engine wins, not the one with the loudest outbound.

When I speak with CEOs from the manufacturing mid-market today, I rarely hear fear of AI. I hear fatigue from tools that promised more than the process could deliver. On the iPad in the Stuttgart hotel lobby, in the end, 1,842 leads weren't the problem. It was 1,842 unresolved decisions that no one wanted to make beforehand.