Meta Ads management
Facebook and Instagram advertising where the signal is repaired before the budget moves: Pixel and Conversions API sending the same deduplicated events, values on purchases, and creative treated as the variable that actually decides delivery. Structure is chosen to keep learning intact, not to look busy.
- Pixel & Conversions API
- Event deduplication
- Advantage+
- Creative testing
ADVERTISE
Digital advertising & media
A working advertising practice: strategy, media planning, campaign management, creative testing and reporting across search, social, video and app platforms — run against your own conversion data rather than platform defaults.
Meta's delivery system is a matching engine trained on the events you send it. When those events arrive incomplete, duplicated or without a value attached, the system still optimises confidently, just towards a cheaper version of the wrong person. The usual failure is a browser Pixel firing on page loads rather than real actions, no server-side counterpart behind it, and purchase events carrying no currency amount at all. That is measurement work, and it comes before structure.
On Meta, creative is the targeting. Detailed interest layers matter far less than they did, because the system finds the audience once the creative tells it who the product is for. What structure controls is whether learning ever completes: too many ad sets splitting the same budget, edits made mid-flight, and a fresh audience every fortnight all keep campaigns in a permanent learning state where nothing stabilises and nothing can be read.
Meta's reporting and your own analytics will not agree, and neither is lying. Meta attributes on its own click and view windows, includes conversions it models rather than observes, and reports on the day of the click; your analytics usually credits the last non-direct click on the day of the purchase. The job is knowing which gap is structural and which one means an event is broken. We work inside accounts and Business Manager assets that stay yours.
Symptoms of an account running on defaults
Four patterns that appear in accounts left to run the way the interface suggests. Each one has a mechanical cause sitting underneath it, and not one of the four is solved by putting more money through the same day.
- Purchases in Ads Manager, not in the bank
- The account reports conversions the finance side cannot find. Often the Pixel and the Conversions API are both sending purchase events with no shared event id, so one order is counted twice. Sometimes the purchase fires on the checkout page rather than on order confirmation, counting abandoned baskets as revenue. Either way the reported return is inflated, and the bid strategy is learning from it.
- Campaigns never leave the learning phase
- Ad sets sit in learning limited because the budget is spread across too many of them to reach enough weekly conversions each. Every edit to budget, audience, creative or optimisation event resets the count. The account looks busy and performs erratically, because the system is repeatedly restarting an estimate it never finishes. Consolidating the same spend into fewer decisions is usually the fix, and it feels like doing less.
- One creative carries the account, then dies
- A single video does most of the volume until frequency climbs and results fall away sharply. Nothing was in production behind it because testing was informal, so the replacement is a guess made under pressure. Creative fatigue on Meta is not a surprise event. It is a scheduling failure, and it is prevented by having the next angle tested before the current one is exhausted.
- Advantage+ was switched on and left alone
- The campaign takes the budget and the reporting says very little about what inside it worked. Existing-customer spend is uncapped, so a share of the reported revenue is people who would have bought anyway. Creative sits as a mixed bundle with no way to read one idea apart from another, and no exclusion list keeps retention traffic out of the acquisition numbers.
What the work includes
Six areas, and they are not equal. If the event layer is wrong, the first two decide what everything after them is worth. Scope is written down before anything starts rather than assumed into existence.
Pixel, Conversions API and deduplication
The event layer is built or repaired first. Browser and server events are sent for the same actions, matched on a shared event id so Meta counts one conversion rather than two, and verified live in the Events Manager test tool before any budget decision is allowed to rest on them.
- Meta Pixel audited action by action, rather than assumed to be correct
- Conversions API sent server-side, from your backend or a server-side container
- Shared event_id plus fbp and fbc parameters, so browser and server events deduplicate
- Purchase events carrying value, currency, contents and an order identifier
- Customer information parameters hashed and sent only where consent allows
- Event Match Quality treated as a working diagnostic, not a score to admire
- An event map in writing: each event named, the user action behind it, and the campaign bidding on it
Domain verification, AEM and iOS constraints
Web events for iOS traffic pass through Aggregated Event Measurement, which caps how many events a domain can optimise towards and ranks them by priority. Domain verification decides who controls that list. We configure both deliberately, ordered by commercial importance rather than by whatever happened to be set first.
- Domain verified in Business Manager against the domain that holds the checkout
- Aggregated Event Measurement priorities ordered by commercial value
- Value optimisation enabled only where the purchase event genuinely carries value
- Attribution setting chosen per campaign and stated on every report that uses it
- SKAdNetwork and AdAttributionKit handling where an app is part of the mix
- Consent state passed through, so a declined visitor is suppressed rather than mislabelled
Campaign structure that protects learning
Structure follows the choices the business actually has to make, then consolidates until each of those choices sits behind enough weekly conversions to finish learning. Fewer ad sets, budget held higher up and changes batched into releases give each decision point the conversion volume it needs, which a detailed hierarchy in permanent learning cannot. Every change is logged with a date and a reason for it.
- Prospecting and retention kept in separate campaigns, so acquisition is never propped up by returning buyers
- Budget consolidated at campaign level where that serves the goal better than per-ad-set control
- Ad set count kept low enough for each one to clear its weekly conversion threshold
- Edits batched into a release rather than made daily out of nervousness
- Naming conventions carried across campaigns, ad sets, ads and UTM parameters
- Customer exclusions applied to prospecting and re-checked after every audience refresh
Creative direction and structured testing
On Meta the creative decides more of the outcome than the settings do, which is why it arrives on a plan instead of on request. Angles are written down first, each variant isolates a single idea, and every test carries a read-out point and a floor on spend fixed before it begins. Direction, briefing and assembly happen here; there is no shoot floor behind this business, so anything requiring one is scoped separately rather than quietly promised. Retired assets keep their results attached, which stops a losing angle reappearing later under a fresh name.
- A written angle list: problem, proof, offer, objection, and the context the ad appears in
- Hooks tested in the opening seconds, where the scroll decision is actually made
- Static, video, carousel and collection formats produced for the same angle
- Placement-aware crops and safe zones for Feed, Reels, Stories and in-stream ratios
- One variable changed per variant, so a result can be explained rather than admired
- Product footage and creator-style formats briefed separately instead of blended
- A creative library recording spend, result and the reason each asset was retired
Audiences, catalogue and dynamic ads
Audience work here is mostly about refusing to fragment delivery. Broad prospecting behind strong creative can beat stacked interest layers outright, so the two are run against each other rather than assumed either way, and retention audiences stay in their own campaigns instead of being folded into acquisition. Where a catalogue exists, feed health decides whether dynamic ads can function at all, which is why the feed is treated as infrastructure rather than as a marketing asset.
- Broad targeting tested against interest stacking, with the same creative on both sides
- Customer lists uploaded hashed, refreshed on a schedule and matched to consent status
- Website and engagement custom audiences built only from events we have verified
- Lookalike sources chosen for customer value rather than for list size
- Catalogue feed checked for availability, price, image and identifier errors
- Dynamic and catalogue ads configured where the feed is genuinely good enough to carry them
Reporting and reconciliation you can check
Reporting puts Meta's figures and your own next to each other, names the attribution setting on both, and explains the difference instead of smoothing it away. Where the gap is structural we say so. Where it points at a broken event or a blocked request, we fix the cause rather than restate the number.
- Ads Manager, GA4 and order data compared on the same date basis
- Click and view attribution windows reported explicitly, never blended silently
- Modelled and observed conversions distinguished wherever the platform discloses the difference
- UTM parameters standardised so your analytics can see Meta traffic properly
- Incrementality tested by geo holdout or a planned spend pause where budget allows
- A plain-language read each period, including the questions the data cannot yet answer
How a Meta engagement runs
Sequence matters more than the list. Signal is repaired first, structure is settled second, and creative volume only becomes worth funding once the account can say which idea did the work.
- 01
Access to your assets, and a written baseline
You grant access through Business Manager to the ad account, Pixel dataset, catalogue and pages you already own. We read the structure, the event setup, the creative history and the spend record, and the first thing you get back is a written baseline: what exists today, which parts of it are wrong, and the order we would correct them in. Nothing is created under our own identity along the way.
- 02
Fix the signal before spend is touched
Pixel events are traced back to the actions they claim to represent, the Conversions API is implemented or corrected, deduplication is proved in the test tool, and domain verification and event priorities are set. Value and currency go onto purchase events. Until an order placed on a real device can be followed the whole way through, structural work waits.
- 03
Consolidate the structure, then let it settle
Campaigns are rebuilt around the choices that genuinely get made each month, and the ad set count comes down until every one of them can clear its learning threshold. Prospecting is separated from retention and exclusions are applied. The account is then left alone for a defined hold period, because batching changes feels slower and produces results that can actually be read.
- 04
Test creative on a cadence, not on a bad week
Angles are briefed and variants produced to a schedule rather than in reaction to a poor fortnight. Each test runs to a stated read-out point with a minimum spend behind it. Winners are scaled inside the structure that already exists, losers are retired with the reason recorded, and the following angle is already in production.
- 05
Reconcile the numbers, then hand back cleanly
Each period closes with Meta's reporting and your own figures placed side by side, the attribution settings named, and any difference explained rather than argued about. Should the engagement stop, the event map and the creative library are handed across, our access is revoked and the revocation confirmed to you in writing, and every asset carries on sitting exactly where it already sits, inside your Business Manager.
Two ways to run a Meta account
The difference is rarely budget. It is whether the account is being told something true, and whether anyone can read what happened afterwards.
Left to defaults
- One browser Pixel, installed once and trusted forever, with nothing server-side behind it
- Ad sets multiplied until every one of them is stuck in learning limited
- Creative delivered when somebody remembers, judged on which version felt strongest
- Advantage+ switched on with existing customers included and no exclusion list applied
- Ads Manager figures reported as fact, with the analytics gap left unexplained
Run deliberately
- Browser and server events sent together, deduplicated by event id and verified live
- Fewer ad sets, consolidated budget, and changes batched so learning completes
- Angles briefed on a schedule, one variable per variant, read-out point agreed first
- Advantage+ used where it earns its place, with exclusions and spend caps in force
- Both sets of numbers reported side by side, with the reason for the gap written down
Signal quality, before any budget moves
None of this is glamorous, and all of it decides whether the delivery system is being trained on something real. It is completed and evidenced before a bid strategy or a budget is worth arguing about.
- The Pixel fires on completed actions and not on page loads, checked one action at a time in the browser
- The server stream leaves your backend or a container carrying the same payload the browser reported
- A shared event id proves the two streams collapse: one order reads as one conversion, never as two
- The domain holding the checkout is verified under the Business Manager belonging to the trading entity
- Aggregated Event Measurement slots are ranked by commercial worth, not by whichever order they were first saved in
- Every purchase event carries a currency amount, its line items and an order reference the finance side can look up
- Matching parameters are hashed before they leave the page, and travel only where consent and law both permit
- An order placed on a real handset arrives in the Events Manager with each of its parameters intact
- Where a catalogue is in play, its feed clears availability, price, imagery and identifier checks
- A declined consent state suppresses the visitor outright instead of sending an unlabelled event onward
- Documentation is current enough for somebody new to trace a reported conversion back to the action behind it
- No line here is ticked on assurance; each one has a screenshot, a log line or a test result attached to it
Independence from Meta
MNF Infotech works on Meta's advertising products as an outside supplier and nothing beyond that. No Business Partner status sits behind this page, no Meta certification, no sponsorship, no affiliation and no authority to act for Meta Platforms, Inc. We never present ourselves as Meta to you or to anybody else, and nobody outside the platform can steer a review, lift a restriction, overturn a disapproval or hurry an appeal along; those calls belong to Meta alone. The names Meta, Facebook, Instagram and Advantage+ belong to Meta Platforms, Inc. and appear here only to identify the products this work is done inside.
Tools we work in
Everything listed below belongs to somebody else and is named so you can see which surfaces the work touches. A tool appearing on this list claims nothing at all: no partnership, no certification, no authorisation, no reseller arrangement, no sponsorship, no endorsement and no relationship whatsoever with whoever holds the trademark.
- Campaign platforms
- Meta Ads Manager
- Meta Business Manager
- Meta Events Manager
- Meta Commerce Manager
- Advantage+ campaigns
- Meta Ad Library
- Instagram professional tools
- Signal and tagging
- Meta Pixel
- Conversions API
- Conversions API Gateway
- Google Tag Manager
- Server-side Tag Manager
- Shopify
- WooCommerce
- Consent Mode
- Measurement and reporting
- Google Analytics 4
- BigQuery
- Looker Studio
- Google Sheets
- SQL
- Meta Ads reporting exports
- Creative direction and editing
- Figma
- Adobe Premiere Pro
- Adobe After Effects
- CapCut
- Canva
- Meta Creative Hub
What you end up owning
- An ad account and Business Manager assets that remain under your ownership
- Browser and server events deduplicated, valued and verified in the Events Manager
- Campaign structure consolidated enough for learning to finish
- A creative testing schedule with angles, read-out points and retired results recorded
- Meta's numbers and your own reconciled, with the gap explained in writing
What people ask before starting
01Are you a Meta Business Partner, or Meta certified?
Neither. There is no Business Partner badge attached to this business and no Meta certification behind it, and MNF Infotech carries no affiliation with Meta Platforms, Inc. and no endorsement from it. Every platform name on this page identifies a product, never a relationship. Where your procurement rules make a badge mandatory, that rules us out, and both sides are better off establishing it before a contract exists than discovering it afterwards.
02Who owns the ad account, the Pixel and the audiences?
You do, all of it. The ad account, the Business Manager itself, the Pixel and its dataset, the catalogue, the pages and every custom audience built from your own data are registered to your business; we hold assigned roles inside those assets, and you can withdraw them at any hour without asking us first. Assets are never spun up under an identity of ours, so when the work stops there is nothing to migrate and nothing to reclaim.
03Why does Meta report more purchases than our own system?
Because they are counting different things. Meta credits a conversion to the day of the click or view inside its attribution window, and includes conversions it models rather than observes; your analytics usually credits the last non-direct click on the day the order happened. A gap is normal. A gap that widens, or that appears on one event only, usually means the signal is broken.
04Should we just use Advantage+ and stop structuring campaigns?
Sometimes. Advantage+ shopping campaigns earn their place where there is a healthy catalogue, enough conversion volume for the system to learn from, and creative variety to draw on. They are a poor choice for thin accounts, for offers that need a specific audience, or where existing-customer spend is left uncapped and inflates the reported return. We test it against a manual structure rather than assuming either answer.
05Our ad account was restricted. Can you get it back?
We can help you prepare and submit an appeal, correct whatever policy issue is visible in the ads, the page or the destination, and keep the rest of the account compliant while review runs. We cannot influence the outcome, expedite a review, or reach anyone inside Meta on your behalf, and any provider claiming otherwise is selling access it does not have.
06How are you paid, and what happens if it does not work?
Management is quoted after the audit, as a monthly fee for the scope agreed, and media spend is billed by Meta to your own payment method rather than passing through us. We do not guarantee a return, a cost per purchase or an approval outcome. Where the evidence points at pricing, at the proposition itself or at the page the traffic lands on instead of at the campaigns, that goes into the written read as well, including on the occasions when it argues for cutting the budget we are paid to manage.
What this work usually connects to
Start with the signal, not the budget
Send the ad account ID, the domain your Pixel sits on and a line about what you sell. We will come back with the specific checks we would run first and the order we would run them in.
Directcontact@mnfinfotech.com