AI PPC Strategies for Higher Ad Returns

Indian advertisers are facing a steep rise in cost‑per‑click across major platforms, especially in metros like Delhi, Mumbai, and Bengaluru, where competition for high‑intent keywords has pushed average CPCs beyond ₹120 in sectors such as finance and edtech. Many brands still rely on manual bid adjustments and broad audience targeting, resulting in wasted spend and sub‑optimal ROI. This article introduces ai ppc strategies that leverage machine learning models to predict click‑through probability, optimize bid adjustments in real time, and allocate budget across channels based on performance signals. By the end of this section you will understand the core components of AI‑driven PPC, learn how to set up a pilot campaign using popular Indian‑friendly tools, discover best practices that keep your account compliant and profitable, and see a side‑by‑side comparison of leading AI PPC platforms available in 2026.

Understanding ai ppc strategies

Core Mechanisms Behind AI‑Powered Bidding

AI PPC strategies replace static rule‑based bidding with predictive models that ingest historical conversion data, contextual signals (device, time of day, location), and external factors such as festival calendars or weather patterns. For example, a retail brand in Jaipur can feed past Diwali sales data into a model that anticipates a 22% surge in conversion probability during evening hours on November 10‑15, prompting the system to raise bids automatically by 35% for those slots. The model continuously retrains every 4‑6 hours using fresh click and conversion streams, ensuring bids stay aligned with shifting market dynamics. In practice, advertisers using AI bidding in Mumbai’s real‑estate niche reported a reduction of cost‑per‑lead from ₹850 to ₹560 within six weeks, while maintaining lead volume.

• Real‑time bid adjustment based on predicted conversion probability.
• Integration of first‑party CRM data for audience refinement.
• Seasonality modeling using Indian festival calendars (e.g., Holi, Durga Puja).
• Cross‑channel budget shifting guided by incremental ROI forecasts.
• Automatic exclusion of low‑performing placements using anomaly detection.

Key Benefits for Indian Markets

Indian advertisers gain several tangible advantages when they adopt AI PPC strategies. First, the technology reduces manual effort: campaign managers in Hyderabad report saving up to 12 hours per week previously spent on bid spreadsheets. Second, AI models uncover hidden audience segments; a Bangalore‑based edtech firm discovered that users searching for “online MBA fees” on weekend mornings had a 3.2× higher conversion rate, a pattern missed by manual rules. Third, budget efficiency improves: AI‑driven allocation shifted 18% of spend from underperforming display networks to high‑intent search in Pune, raising overall ROAS from 2.8 to 4.1. Finally, compliance with platform policies is enhanced because AI can automatically pause ads that violate emerging restrictions, such as new gambling ad guidelines rolled out in March 2026.

1. Time savings: average 10‑15 hours weekly per campaign manager.
2. Audience insight: discovery of micro‑segments with uplift >200%.
3. Budget efficiency: ROAS uplift ranging from 30% to 70% in case studies.
4. Risk mitigation: automatic disapproval of policy‑violating creatives.
5. Scalability: same model works across Google Ads, Meta Ads, and emerging platforms like ShareChat Ads.

Implementation Guide

Step‑by‑Step Setup Using Google Ads AI Features (2026)

Begin by linking your Google Ads account to Google Cloud’s Vertex AI platform. Ensure you have the latest Google Ads API version v15 installed, which supports the new SmartBiddingService endpoint. Follow these steps:

1. Create a Vertex AI user‑managed notebook with Python 3.11 and install the google‑ads library (v20.0.0).
2. Export the last 90 days of conversion data from Google Ads via the API, storing it in a BigQuery dataset named ppc_ai_2026.
3. Train a regression model using Vertex AI’s AutoML Tables targeting conversion_value as the label, with features such as hour_of_day, device_type, location_id, ad_group_id, and historical_ctr.
4. Deploy the model as an online endpoint (ppc_bid_predictor) with a minimum of 2 nodes for low‑latency inference.
5. In Google Ads, create a custom bidding strategy via the API: set bidding_strategy_type to PORTFOLIO and attach a bidding_scheme that calls your Vertex endpoint for each auction.
6. Set a target ROAS of 3.5 (or your business goal) and let the AI adjust bids in real time.
7. Monitor performance through the Experiment framework, running a 50/50 split between AI bidding and manual CPC for two weeks.

Code snippet for initializing the Vertex endpoint call (Python):

from google.cloud import aiplatform
from google.ads.googleads.client import GoogleAdsClient def predict_bid(context): endpoint = aiplatform.Endpoint(endpoint_name="projects/your‑project/locations/us‑central1/endpoints/ppc_bid_predictor") instance = [[context['hour'], context['device_encoded'], context['location_encoded'], context['ad_group_encoded'], context['historical_ctr']]] prediction = endpoint.predict(instances=instance) return prediction.predictions[0][0] # expected conversion value # Example usage in a bidding adjustment function
def adjust_bid(ad_group_id, context): base_bid = get_current_bid(ad_group_id) predicted_value = predict_bid(context) suggested_bid = base_bid * (predicted_value / context['historical_value']) return suggested_bid

Replace placeholder values with your actual project IDs and encoded features. Test the function in a sandbox before applying to live campaigns.

Leveraging Third‑Party AI PPC Tools Popular in India

Several Indian‑focused platforms now offer AI bidding as a service, reducing the need for heavy internal data science teams. Tools such as AdScale AI (v4.2, released Jan 2026), BidBuddy (v3.1, Mumbai‑based), and PixelPulse (v2.0, Bengaluru) provide pre‑built models tuned for Indian seasonal trends. Implementation typically involves:

• Connecting your ad accounts via OAuth 2.0 (supports Google Ads, Meta Ads, and ShareChat Ads).
• Uploading a CSV of past conversions (minimum 30 days) – the platform auto‑maps columns.
• Selecting a model template: “E‑commerce Festival Boost”, “Lead Gen Cost‑Per‑Lead Optimizer”, or “App Install ROAS Maximizer”.
• Setting a budget ceiling and target metric (e.g., target CPA ≤ ₹450 for finance leads).
• Activating the AI engine; the platform provides a real‑time dashboard showing bid adjustments, projected uplift, and alert thresholds.

For example, a Delhi‑based DTC brand using AdScale AI reported a 28% decrease in CPC during the Republic Day sale period, while maintaining sales volume, after just ten days of AI optimization.

Best Practices for ai ppc strategies

Do’s: Ensuring Optimal AI Performance

1. Feed clean, granular data: include transaction value, timestamp, and device ID; avoid aggregating to daily totals.
2. Update training data weekly: AI models drift when faced with new product launches or sudden market shifts.
3. Use portfolio bidding across similar campaigns: lets the algorithm leverage shared learning while respecting individual budget caps.
4. Set realistic performance targets: start with a modest ROAS lift goal (10‑15%) before scaling to aggressive targets.
5. Leverage audience exclusions: prevent AI from bidding on low‑intent segments such as job‑seekers for B2B offers.
6. Run controlled experiments: always keep a 10‑20% manual control group to validate incremental impact.
7. Stay within platform policy limits: monitor disapproval reasons and adjust creative or landing page accordingly.

Don’ts: Common Pitfalls to Avoid

1. Do not rely solely on AI for creative decisions: human oversight is still needed for ad copy relevance and brand safety.
2. Do not ignore seasonality overrides: if a major festival is approaching, manually increase budget caps to allow AI to capture surge.
3. Do not set impossibly low CPA targets: the algorithm may under‑deliver or pause ads entirely, hurting visibility.
4. Do not neglect data privacy compliance: ensure any first‑party data uploaded to third‑party AI tools is encrypted and consent‑based.
5. Do not use a single model for all verticals: a model trained on finance leads may mis‑predict for FMCG impulse buys.
6. Do not forget to monitor bid spikes: extreme bid increases can drain daily budgets quickly; set max bid caps.
7. Do not skip post‑implementation audit: after 30 days, compare AI performance against baseline and recalibrate if needed.

Comparison Table

Advanced Techniques

Scaling strategies

To scale AI‑driven PPC campaigns in India, begin by expanding geo‑targeting beyond Tier‑1 metros. Use platform‑level audience insights to identify emerging demand pockets in cities like Jaipur, Indore, and Kochi where competition is lower but intent is rising. Allocate a test budget of ₹2,00,000 per new region and let the AI model optimize bid adjustments based on real‑time conversion signals. Layer in look‑alike audiences built from your highest‑value converters; the AI will refine these segments weekly, decreasing cost‑per‑acquisition by up to 18 %.

Another lever is day‑parting automation. Feed historical conversion timestamps into the AI engine so it can shift spend toward high‑intensity windows—typically 7 pm‑10 pm IST on weekdays and 11 am‑2 pm on weekends for consumer‑facing offers. Implement dynamic budget caps that pause under‑performing ad sets automatically when the projected ROAS falls below 2.0 for two consecutive hours. This prevents wasteful spend while preserving aggressive bidding during peak periods.

Finally, employ cross‑channel synergy. Sync your AI PPC data with CRM and email platforms to create unified user journeys. When a prospect clicks an ad but does not convert, trigger a personalized retargeting sequence via WhatsApp Business API or SMS, with offers calibrated by the AI’s propensity score. This closed‑loop approach can lift overall campaign ROI by 22‑27 % without increasing media spend.

Performance optimization

Optimization starts with granular attribution. Move beyond last‑click models and adopt data‑driven attribution (DDA) powered by AI, which assigns fractional credit to each touchpoint based on its influence on conversion. In Indian markets where the purchase journey often involves multiple device switches, DDA reveals that display and video assist clicks contribute up to 35 % of final conversions, prompting a reallocation of budget toward upper‑funnel formats.

Next, implement continuous creative testing via AI‑generated variants. Use responsive search ads (RSA) with up to 15 headlines and 4 descriptions; the AI will automatically combine the highest‑performing combos. Complement this with AI‑crafted image assets for Performance Max campaigns, adjusting color schemes and copy to match regional festivals—e.g., using saffron tones during Navratri in Gujarat or incorporating local language snippets for Tamil‑speaking audiences.

Bid strategy refinement is crucial. Shift from manual CPC to AI‑driven maximize conversion value with a target ROAS constraint. Set the target at 2.5× for lead‑gen campaigns and 3.0× for e‑commerce, allowing the engine to raise bids on high‑intent keywords while lowering them on broad match terms that drain budget. Monitor the “bid adjustment percentage” metric weekly; if it stays within ±5 % for three weeks, the model has stabilized and you can safely increase the overall budget by 10‑15 %.

Advanced tips for experts include exploiting seasonal trend forecasting. Feed historical Google Trends data, local event calendars, and weather APIs into the AI model to predict spikes in demand for products like air conditioners before summer or sweaters ahead of Diwali. Pre‑emptively increase budget caps by 20‑30 % two weeks before the forecasted surge, then let the AI fine‑tune bids in real time. This proactive approach often yields a 15‑25 % lift in conversion volume compared to reactive adjustments.

Real World Case Study

Client: A Bangalore‑based B2B SaaS provider offering cloud‑based HR analytics to mid‑size enterprises across India. The company faced stagnating lead generation despite a steady monthly ad spend of ₹8,50,000. Their cost‑per‑lead (CPL) stood at ₹4,650, conversion rate at 2.1 %, and overall ROAS at 1.4×. The leadership set a goal to cut CPL by 30 % while boosting qualified leads by 50 % within two months.

Week‑by‑week solution

1. Weeks 1‑2: Discovery – Conducted a full audit of existing campaigns, keyword match types, audience layers, and landing page performance. Identified that 42 % of spend was wasted on broad match keywords with low intent, and that ad copy lacked region‑specific value propositions. AI‑powered search term report revealed high‑volume, low‑competition long‑tail terms like “HR analytics software for manufacturing units in Chennai”.
2. Weeks 3‑4: Implementation – Restructured campaigns into tightly themed ad groups using exact and phrase match. Introduced AI‑generated responsive search ads with dynamic keyword insertion tailored to industry verticals (manufacturing, IT services, healthcare). Launched Performance Max campaigns feeding product‑level signals from the CRM. Adjusted geo‑bids to increase allocation to Tier‑2 cities (Pune, Ahmedabad, Coimbatore) by 25 % while reducing Delhi NCR spend by 15 %.
3. Weeks 5‑6: Optimization – Switched bid strategy to maximize conversion value with a target ROAS of 2.5×. Enabled AI‑driven ad schedule adjustments based on hour‑of‑day conversion data, shifting 18 % of budget to the 7 pm‑10 pm IST window. Implemented automated negative keyword lists that excluded job‑search and free‑tool queries, saving roughly ₹1,20,000 per week. Ran A/B tests on landing page hero images, finding that visuals featuring diverse Indian workforce increased form submissions by 12 %.
4. Weeks 7‑8: Results – After eight weeks, CPL dropped to ₹2,460 (a 47 % improvement), total ad spend reduced to ₹5,30,000 saving ₹3,20,000, qualified leads rose to 183 (up from 112 in the prior period), and ROAS climbed to 2.7×. The AI model continued to refine bids, delivering an additional 8 % lift in conversion volume during the final two weeks without any manual intervention.

The case demonstrates how a disciplined, AI‑first approach can transform inefficient PPC spend into a predictable profit engine for Indian B2B marketers.

Common Mistakes to Avoid

Mistake 1: Over‑reliance on broad match keywords

Many advertisers still allocate a large portion of their budget to broad match terms, assuming AI will filter irrelevant traffic. In practice, broad match can trigger ads for loosely related queries, inflating spend without delivering conversions. For a typical mid‑size campaign in India, this can waste anywhere from ₹1,50,000 to ₹3,00,000 per month. To avoid this, shift to phrase and exact match for core intent terms, and use broad match only in tightly controlled experiments with explicit negative keyword lists. Recovery involves pausing under‑performing broad match ad sets, reallocating the saved budget to high‑intent phrase match groups, and letting the AI re‑optimize bids over a 7‑day learning period.

Mistake 2: Ignoring ad schedule data

Running ads 24/7 without considering temporal performance leads to wasted spend during low‑intensity hours, especially in Indian markets where online shopping peaks in the evenings. The cost impact can be ₹80,000‑₹2,00,000 monthly for campaigns with a ₹5,00,000 budget. To prevent this, analyze hour‑of‑day conversion data via the platform’s reporting tools, then apply AI‑driven ad scheduling to concentrate spend in top‑performing windows (typically 6 pm‑11 pm IST). If already running nonstop, implement an immediate schedule adjustment and monitor the change in cost‑per‑acquisition for three days; you should see a reduction of 10‑18 % in wasted spend.

Mistake 3: Neglecting landing page relevance

Driving traffic to a generic homepage instead of a dedicated, message‑matched landing page increases bounce rates and reduces quality scores. In India’s competitive CPC environment, each irrelevant click can cost ₹15‑₹30 extra, translating to ₹1,00,000‑₹2,50,000 wasted monthly for a campaign receiving 50,000 clicks. Avoid this by creating dynamic landing pages that mirror the ad’s headline and offer, using AI‑powered content personalization (e.g., inserting the user’s city or industry). To recover, audit existing landing pages, replace mismatched pages with tailored variants, and run a quick A/B test; expect a quality score improvement of 1‑2 points within a week, lowering CPC by roughly 12 %.

Mistake 4: Setting static bids in volatile auctions

Fixed CPC bids fail to adapt to sudden competition spikes—common during festive sales or product launches—causing either lost impressions or overspend. The financial impact can be severe, ranging from ₹2,00,000 to ₹5,00,000 per month for high‑budget accounts. Prevent this by adopting automated bid strategies like maximize conversion value with a target ROAS, allowing the AI to adjust bids in real time. If you’re currently using manual bids, switch to an automated strategy and observe the performance over a 14‑day period; you should see a more stable cost‑per‑conversion and a 5‑10 % increase in conversion volume.

Mistake 5: Failing to leverage audience exclusions

Advertisers often target broad audiences without excluding existing customers or low‑value segments, leading to redundant spend on users unlikely to convert again. For a SaaS firm with a ₹7,00,000 monthly budget, this oversight can waste ₹1,20,000‑₹2,50,000 each month. Avoid this by uploading customer lists, past purchasers, and low‑engagement segments as exclusion audiences in the platform. To recover, immediately add these exclusion lists, then re‑evaluate the campaign’s CPL after one week; you should notice a drop of 15‑25 % in wasted spend and a corresponding improvement in ROAS.

Frequently Asked Questions

What are the top ai ppc strategies for 2026 to maximize ROI in Indian markets?

The leading AI PPC strategies for 2026 in India revolve around three pillars: predictive audience modeling, real‑time bid automation, and creative‑first optimization. First, leverage machine‑learning models that ingest first‑party data (CRM, website behavior, offline sales) to build look‑alike audiences with a predicted conversion probability above 0.35. Allocate 20 % of your test budget to these audiences and let the AI refine them weekly. Second, switch from manual CPC to AI‑driven maximize conversion value with a target ROAS; set the target at 2.5× for lead generation and 3.0× for e‑commerce, enabling the engine to raise bids on high‑intent keywords while lowering them on wasteful terms. Third, adopt responsive search ads and Performance Max campaigns that automatically test up to 15 headlines and 4 descriptions, using AI to surface the best‑performing combos. Complement this with dynamic creative assets that adapt to regional festivals, language preferences, and device type. Implement day‑parting based on hour‑of‑day conversion data, shifting at least 15 % of budget to peak windows (7 pm‑10 pm IST). Finally, establish a feedback loop where conversion data from the CRM feeds back into the AI model every 24 hours, ensuring the system continuously learns from actual revenue rather than just clicks. Following this roadmap typically yields a 20‑30 % reduction in cost‑per‑lead and a 1.5‑2× increase in ROAS within the first six to eight weeks.

How much budget should I allocate to test AI PPC strategies in a new Indian city like Jaipur?

When entering a new geographic market such as Jaipur, start with a controlled test budget that is large enough to generate statistically significant data but limited enough to manage risk. A prudent starting point is ₹2,00,000‑₹3,00,000 per month for a 4‑week testing period. This amount allows you to run campaigns across search, display, and Performance Max formats while gathering sufficient impressions (aim for at least 50,000‑70,000) to let the AI optimize bids and audience splits. Divide the budget as follows: 40 % to search campaigns with exact and phrase match keywords focused on local intent (e.g., “HR software Jaipur”, “payroll solutions Rajasthan”), 30 % to Performance Max campaigns that leverage your product feed and audience signals, and 20 % to retargeting lists built from website visitors in Jaipur, using dynamic ad copy that references local culture or festivals. Reserve the remaining 10 % for exploratory broad match experiments with strict negative keyword lists to uncover hidden long‑tail opportunities. Monitor key metrics—cost‑per‑lead, conversion rate, and ROAS—weekly. If the CPL falls below your target threshold (say ₹3,000 for B2B leads) and ROAS exceeds 2.0× after two weeks, consider scaling the budget by 50‑100 % in the third month while maintaining the same audience structure. If performance is subpar, pause the campaign, revisit keyword relevance and landing page alignment, then relaunch with a revised creative set.

What timeline should I expect to see measurable results from AI PPC optimization?

Measurable results from AI PPC optimization typically emerge in phases, with early indicators appearing within the first two weeks and more substantial impacts visible by week six to eight. During weeks 1‑2, the AI system completes its learning period, adjusting bids, refining audience targeting, and filtering out irrelevant search terms. You may notice a 5‑10 % reduction in cost‑per‑click (CPC) and a slight uplift in click‑through rate (CTR) as the model begins to prioritize higher‑intent queries. By weeks 3‑4, as the AI has gathered sufficient conversion data (aim for at least 100‑150 conversions), you should see a more pronounced decline in cost‑per‑lead (CPL)—often 15‑25 % below baseline—and an improvement in conversion rate of 10‑20 %. This is the stage where you can confidently adjust budget allocations, shifting spend toward the best‑performing ad sets and geographic segments. By weeks 5‑6, the optimization loop tightens; the AI begins to exploit day‑parting patterns and seasonal trends, delivering a stable ROAS improvement of 0.3‑0.5×. At this point, many advertisers report a 20‑30 % increase in qualified leads without raising the overall budget. Finally, by weeks 7‑8, the model’s predictions become highly reliable, allowing you to set more aggressive targets (e.g., target ROAS of 3.0×) and to experiment with advanced tactics like predictive budget pacing or cross‑channel attribution. Overall, expect a cumulative ROI uplift of 40‑60 % after two months of consistent AI‑driven management, provided you maintain clean conversion tracking and avoid major structural changes that reset the learning phase.

How can I integrate AI PPC data with my existing CRM to improve lead quality in India?

Integrating AI PPC data with your CRM creates a closed‑loop feedback system that significantly enhances lead quality and enables smarter budget allocation. Begin by ensuring that your PPC platform (Google Ads, Meta Ads, or LinkedIn) is set up to capture auto‑tagged parameters such as gclid, fbclid, or utm_source in the landing page URLs. Pass these parameters to your CRM via a middleware solution like Zapier, Segment, or a custom webhook that fires on form submission. In the CRM, create a custom field to store the PPC source, campaign name, ad group, and keyword that generated the lead. Next, schedule a daily sync (using a tool like Microsoft Power Automate or an ETL script) that pulls conversion data—lead status, deal value, and close date—from the CRM back into the advertising platform’s offline conversions upload. This enables the AI optimization algorithms to weigh leads not just by form fills but by actual revenue potential. For Indian B2B markets, where sales cycles can extend 30‑90 days, this offline conversion feedback is vital; it prevents the AI from over‑valuing low‑intent leads that never mature. To improve lead quality, use the combined data to build a lead‑scoring model within the CRM that assigns points based on PPC attributes (e.g., keyword intent score, geographic tier, device type) and demographic fields (company size, industry). Export the top‑scoring leads as a custom audience to the ad platforms for exclusion or bid‑adjustment strategies— for instance, increase bids by 20 % for leads originating from high‑intent long‑tail keywords in Tier‑1 cities, while decreasing bids by 15 % for leads from generic terms in Tier‑3 markets. Review the lead‑score distribution monthly; aim to shift the median score upward by 10‑15 % each quarter, which typically translates to a 12‑18 % reduction in cost‑per‑qualified‑lead and a 1.2‑1.4× increase in sales‑accepted leads.

What are the typical costs involved in hiring an AI PPC specialist or agency for an Indian mid‑size business?

Engaging an AI PPC specialist or agency in India involves a mix of fixed retainerformance fees. For a mid‑size business with a monthly media spend ranging from ₹4,00,000 to ₹1,000, typical monthly charge ranges between ₹ 0,000 to ₹4,00,000, depending on the scope, covering strategy development, campaign management, reporting, and creative testing. agencies charge a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a