Singapore

Better Campaign Leads: Feed Sales Outcomes Back to Bidding

As of November 2026, the answer to "my CPA is on target but Sales says the leads are worse" is that the bidding system is doing its job: it is buying the conversion you defined. If the only conversion you ever sent was a form fill, every lead looks equally good to Google, Meta or Microsoft Advertising, and a lead-quality problem becomes invisible to the algorithm. Improving lead quality starts with handing the platform a deeper outcome such as a sales-qualified lead, an opportunity or a closed deal, plus the plumbing to send it back at reliable volume. Marketers who read this as a copywriting or targeting problem tend to keep buying the same low-intent traffic, which is how a best lead generation programme quietly turns into a lead-count programme.

Key takeaways:

A bidding target only encodes the conversion you gave it; form fills are a proxy, not a business result.
Choose the deepest outcome you can measure reliably, then send it back by offline conversion import or a Conversions API.
Automated bidding needs signal: at least 30 conversions within 30 days per the source this article follows, and roughly 50 conversions to recalibrate after a big budget or target change.
Before adding a feedback loop, check whether Sales actually follows up, whether duplicates inflate the count, and whether the conversion definition is the problem.
Conversion windows bound what can be returned at all, so outcomes that close later than the platform's window never teach the model anything.

Why a healthy CPA and falling lead quality can both be true

The two reports describe different things. The ad platform reports on the conversion event it can see; Sales reports on what happened to the person afterwards. A form submission is complete from the platform's side the moment the thank-you page loads, whether or not the person has budget, authority or a real need.

The platform's optimisation is mechanical rather than malicious. It looks for more of the event that you marked valuable, weighted by the value you attached to it, and it has no way to know that one submission became a strong opportunity and the next went nowhere. When every submission is marked as the same conversion with the same value, the cheapest reachable audience is often an audience that fills forms easily and buying-intent lightly.

The first thing to establish, before asking the platform to get more leads for my business, is whether marketing and Sales agree on what a better lead is. A sales team working from an old impression of which industries convert will keep grading new leads against a stale pattern, and lead quality can shift as targeting changes. Getting the definitions written down before touching bidding or tracking is what separates a fix from a panic.

A marketer and a salesperson sit across a table comparing notes over a printed funnel diagram.

Define the outcome before you touch the bid strategy

The outcome you optimise towards should be the deepest point in the funnel you can measure consistently, not the deepest one you wish you could measure. Writing the definition down as stages makes the choice explicit, and it is the step most best lead generation programmes skip.

A workable model, the one the source column describes, has four levels: a marketing-qualified lead that is not yet ready for Sales; a sales-qualified lead that shows the characteristics of a real prospect and wants a demo or trial; an opportunity that is in active conversation with Sales; and a generated revenue event at close. An inbound-marketing view adds informational and product-qualified leads, which is useful vocabulary but does not change the decision: pick the stage that reliably separates good from bad in your own pipeline.

When most enquiries are sales-qualified and the rest are not, marking both as equal conversions hands the platform a misleading target. Marking the qualified ones as the conversion, and leaving the rest as a softer signal, gives the model a different question to answer.

Outcome targeted	Signal volume	What bidding learns	Typical failure
Form fill	Highest	Who fills forms	Volume of low-intent leads rises
Marketing-qualified lead	High	Who engages past the form	Slightly better, still frictionless
Sales-qualified lead	Medium	Who matches your good-customer profile	Needs Sales to grade consistently
Opportunity	Lower	Who enters real buying conversations	Slow to accumulate volume
Closed revenue	Lowest	Which deals and deal sizes come back	Long sales cycles miss the window

The table is a trade-off, not a ranking. A deeper outcome is a better signal and a rarer one, and the section after next deals with what happens when the rarer one is too rare to learn from.

Send qualification and revenue back as conversion values

Once the outcome exists in your CRM, it has to reach the ad platform. Two mechanisms do that job, and they differ in how the data travels rather than in what the platform does with it.

Offline conversion imports upload outcomes that happened after the click, such as a lead qualified two days later or a deal signed six weeks later, and match them back to the original click. The core fields are the click identifier, the conversion time and the conversion goal name, which is why the click identifier has to survive from the ad click into your CRM.

A Conversions API, usually called CAPI, moves the same events server to server instead of relying on manual uploads, and can carry revenue, product details, click identifiers, hashed identifiers and consent status. Microsoft Advertising's Conversions API sends conversion and engagement events through Universal Event Tracking. The point of CAPI is freshness and completeness, not a different optimisation logic.

Values matter as much as the event itself. A click that produced a sales-qualified lead worth a predictable amount should carry that amount, or a directional value, so budget moves towards the audiences that produce those leads. If you cannot yet measure real value, conversion value rules that adjust the value by probability still beat sending every lead as identical.

The mechanism is not the strategy. If the CRM marks poor leads as successful, a faster pipe sends the wrong definition around the loop more quickly. Building offline conversion tracking before the CRM grading is trusted usually produces a confident version of the same mistake.

An analyst at a desk compares a CRM pipeline screen with an ad platform dashboard on two monitors.

Give automated bidding the signal and time it needs

Automated bidding is a learner, and learners need examples. The source this article follows puts a useful benchmark at 30 conversions within 30 days per campaign as a starting point for automated bidding, and notes that after a significant change to budgets or targets, conversion-based bidding can take roughly 50 conversions to calibrate according to multiple platforms.

Read those numbers as a curriculum requirement if your goal is to get more leads for my business from paid search rather than from referrals. If the deeper outcome you chose produces a handful of events a month, the campaign will not learn from it in any practical timeframe. The usual fixes are to target one level earlier, a marketing-qualified or sales-qualified lead rather than the closed deal, and to keep the ultimate outcome as the reporting truth rather than the bidding signal.

Periods of instability reset the learning. A budget doubled to hit a quarterly target, or a target CPA tightened after a good month, restarts the calibration, and results during that window say more about the reset than about the audience. Resisting aggressive changes while outcomes are still catching up is a control, not timidity.

Conversion windows decide which outcomes can ever be fed back

This is the part that gets missed when a feedback loop is first designed. A platform can only attribute a conversion inside the window it supports, so an outcome that closes after the window is discarded by the algorithm even if you upload it.

Microsoft Advertising conversion tracking supports configurable conversion windows depending on the goal, from one minute up to 90 days. Its CAPI currently has a 90-day look-back window. Google and Meta run their own windows, which the source does not specify and which change, so check them in the platform documentation rather than assume parity.

What follows is a sequencing decision. Map your sales cycle against the window. If most deals close inside it, closed revenue is a realistic signal. If the tail of the cycle runs past it, the honest choice is a mid-funnel outcome such as a sales-qualified lead or an opportunity, with revenue reported internally and used to steer budget reviews rather than bidding.

Audit the loop before you blame the platform

A feedback loop inherits every defect upstream of it, which is why the best best lead generation check starts with the data, not the platform. Before changing bidding strategy, run a short audit on the list below; each item is a cause that produces exactly the symptom of a met CPA with worsening leads.

Audit item	Question to answer	If it fails
Conversion definition	Is each tracked action the right one, and does a micro-conversion carry directional value?	Rebuild the conversion set before bidding changes
Lead qualification	Where do enquiries fall between marketing- and sales-qualified, and why?	Agree grading criteria with Sales
CRM progression	Do leads move between stages cleanly and is each move documented?	Record stage changes as events; they may become your signal
Form and landing page	Is it easy to hand over details, and does the page guide rather than confuse?	Test the form before adding spend
Duplicates and invalid leads	Are conversions counted twice by multiple pixels or manual entry?	Deduplicate; inflated counts mislead bidding
Sales follow-up	Are paid leads contacted within 90 minutes of submission?	Fix follow-up; a slow reply loses prospects to a competitor

The last row is the one sales and marketing tend to argue about, and it is measurable. The source's question is whether every lead you pay for gets a follow-up within 90 minutes of submission; where it does not, prospects simply find their solution elsewhere, and no bidding change will recover them. Closing the loop is as much an operations audit as a tracking project.

A salesperson at a desk answers a call quickly, a wall clock behind the laptop of new leads.

What Topkee's systems do with sales-stage data

Topkee's own stack is built around the same loop, which makes it easy to see what the mechanism looks like when it is wired end to end. Every lead carries a unique identifier (ttoclid) with its UTM parameters, visit time, time on page and browser information; when a salesperson changes a lead's stage in Leadbox, that stage event goes back through TTO Tag to Google Analytics, the Facebook Pixel, Google Ads Conversions and Bing UET. When a lead moves from discovered to created, an ImportedLead event is matched automatically and returned to the ad platforms, which then use their attribution models to look for similar prospects. Stage changes, not form fills, are what get fed back.

That chain, from click through to revenue, is held in the client's own tenant and TTO CDP, so the data the bidding system receives is the client's own record of what happened rather than a platform estimate. The distinction shows up in Topkee's own-account Campaign Intelligence Report, whose published sample runs a six-layer funnel from impression to signed deal and tracks spend against signed-deal value by platform. Across its funnel, the volume of form activity and the volume of deals that leave for the platform are plainly different quantities, which is the gap the feedback loop exists to close. That sample is a demonstrative own-account report, not a customer average or an industry benchmark.

The related offers follow from the same design. The ROAS Engine is measured to closed deal and includes CAPI and offline conversion setup with attribution reconciliation; the TTO Funnel Tuning Agent runs a full-funnel diagnosis in a single working session and states a limit that matches the window problem: a single platform's AI cannot allocate budget across platforms, and conversions outside the official windows are discarded. For an advertiser in Singapore spending across Google, Meta and TikTok, that is the argument for returning revenue rather than clicks.

Recommended next action

Start with the audit table above and grade last month's leads against the stage definitions before changing anything. Then choose the deepest stage that clears the volume test and wire it back. If you want the plumbing and the reporting built in one place, the team's AI-driven Google Ads automation work is the closest starting point, and the guide to Google Ads and first-party data explains how the CRM side of the loop is connected.

A checklist to run this week:

Write down the stage definitions Sales already uses and get them signed off.
Count how many leads last month reached each stage, and compare with the count the ad platform reported.
Check the follow-up time on the last month of paid leads against the 90-minute test.
List every conversion action in each ad account and mark each as predictive or directional.
Compare your sales cycle length with each platform's conversion window.
Pick one deeper conversion to send back, and connect it by import or CAPI.

FAQ

Why does my CPA stay on target while lead quality falls?

Because the platform is optimising for the conversion you defined, not for the outcome you want. If form fills are the only conversion, the system buys cheap form fills, and the CPA it reports stays healthy while the share of those leads that Sales can work falls.

Which conversion should I bid towards instead?

The deepest stage you can measure consistently and reach often enough for the algorithm to learn from. If closed deals are too rare, target a sales-qualified lead or an opportunity, and keep revenue as the internal reporting measure.

How many conversions does automated bidding need?

The source this article follows puts a useful starting benchmark at 30 conversions within 30 days, and roughly 50 conversions to recalibrate after significant changes to budgets or targets according to multiple platforms. Treat those as the minimum curriculum, not a guarantee of performance.

What is the difference between offline conversion imports and CAPI?

Both return post-click outcomes to the ad platform. Offline conversion imports upload the events, typically in batches, matched by click identifier, conversion time and goal name. A Conversions API sends the same events server to server as they happen and can carry revenue, identifiers and consent status.

Why does my feedback loop not change results?

Three common causes: the volumes are too small for the algorithm to learn from, the outcomes close after the platform's conversion window so they are discarded, or the CRM feeding the loop marks poor leads as successes, which simply speeds up the wrong definition.